Each record describes something a country built that the indicators do not capture. The source provides the published number; the mechanism is our interpretation.
Scores describe outcomes. These records describe how a country acted, what it needed and what happened. 742 of 742 records include a mechanism. They do not affect scores.
Brazil created a public agricultural research corporation in 1973 to make acidic tropical soils productive, kept it funded across every change of government since, and cereal production rose from 46.5 million tonnes in 2000 to 155.9 million in 2023.
155,877,082 tonnes, cereal production, 2023. World Bank, from FAO, retrieved 2026-08-26.
The move, and what it needed
A public research corporation with its own budget line, staffed by researchers sent abroad to train and then brought back, mandated against one specific national constraint rather than research in general.
Preconditions
A named constraint worth fifty years of work
Budget continuity across governments
A private sector able to absorb the results
Where else it has been used: The model of a mission-specific public research corporation is highly transferable. The fifty years of funding continuity is the part nobody can procure.
Limits: Production is an outcome and the credit is shared with land expansion, subsidised credit and private plant breeding, so this number cannot separate the research from the rest. It also cannot see the cost: much of the expansion came from converting cerrado and forest. What the record actually evidences is fifty years of continuity in a public research mandate, which is the part no dataset measures.
๐ณ๐ฑNetherlandssince 2010still operatingbears on Government foresight capacity
The Netherlands legislated a standing programme against flooding and fresh-water shortage, run by an independent Delta Commissioner and paid from a dedicated Delta Fund that expects 27.4 billion euros to be available through 2050.
Parliament wrote the Delta Act: a permanent commissioner proposes a rolling programme each year, statutory safety standards make the work non-optional, and a ring-fenced fund keeps the money out of annual budget fights.
Preconditions
A shared national threat no party disputes
Water authorities with their own tax base, older than the state
Fiscal room to ring-fence a billion a year for decades
Where else it has been used: Bangladesh's Delta Plan 2100 was written with Dutch help and copies the programme form. The statutory fund and the commissioner's independence are the parts that travel least, because they require a legislature to give up annual control.
Limits: A budget is an input: the number shows money committed, not risk reduced, and the same page says 30.8 billion is needed, so the fund is currently short of its own task. Spending on water at this scale is also ordinary for the Netherlands, which has run such works for a century. What the record evidences is the institutional form, a legally required programme, commissioner and fund that outlive governments, not the size of the number.
๐ง๐ทBrazilsince 1991still operatingbears on Long-horizon research commitments
Law 8.248/1991 tied information-technology incentives to annual research, development and innovation investment of at least 5 percent of eligible domestic gross revenue, creating a predictable regime for sectoral capability.
The state exchanges a sector-specific incentive for commitments to domestic research, development, training and innovation, giving firms a reason to make longer-horizon plans within a stable policy frame.
Preconditions
A government able to target fiscal incentives at a defined industrial sector
Firms with enough domestic revenue to sustain the investment requirement
Administrative capacity to review projects and monitor the obligation
Limits: The legal investment requirement is an input, not evidence of research quality, exports or firm growth. The regime is narrow, has been amended repeatedly and creates a renewal pressure that makes continuity part of the case rather than a settled result.
๐ง๐ทBrazilsince 2001still operatingbears on Government foresight capacity
Created in 2001, the Centro de Gestรฃo e Estudos Estratรฉgicos was designed to produce prospective studies and long-term recommendations for Brazilian science, technology and innovation policy, with 273 researchers and specialists signing its creation record.
Government creates a distinct organization to turn evidence and expert networks into long-horizon options, giving policymakers a place to ask what future conditions require before a crisis sets the agenda.
Preconditions
A ministry willing to give a foresight body room to examine its own assumptions
A network of researchers and institutions that can contribute expertise
A route from prospective studies into planning and policy decisions
Limits: The creation record and a mandate for prospective studies show institutional intent, not whether foresight changes decisions or survives political pressure. CGEE operates within the science and technology policy system, so its reach does not cover every part of government.
๐ธ๐ฌSingaporesince 2009still operatingbears on Government foresight capacity
Singapore established the Centre for Strategic Futures in 2009 inside the Prime Minister's Office strategy system to study blind spots, run long-term futures research and develop foresight methods.
A central strategy unit gives government a standing place to examine emerging risks and challenge assumptions before those risks become immediate policy demands.
Preconditions
A centre of government willing to receive uncomfortable analysis
Access to policy teams that can act on long-horizon findings
Methods and networks that keep foresight connected to evidence
Limits: Creating a foresight unit shows a formal capability for examining uncertainty, not whether its work changes budgets, policies or decisions. The centre describes its own mandate and does not publish a counterfactual for decisions made without it.
The United Kingdom's 2008 Climate Change Act made five-year emissions budgets a standing legal cycle set twelve years ahead, and by July 2026 seven budgets covered 2008 through 2042 while the first three had been met.
Parliament requires each government to set a quantified five-year ceiling twelve years in advance, receive independent advice, publish a delivery plan and report the final account, converting a distant target into a recurring decision and scrutiny cycle.
Preconditions
A national emissions inventory stable enough to support legal accounting
An independent body with the standing and expertise to challenge government plans
A legislature willing to bind successive governments to decisions beyond one term
Where else it has been used: France and several other countries later adopted carbon-budget frameworks. What travels is the rolling statutory cycle; a target without independent advice, advance budgets and final accounts keeps the name but drops the institutional work.
Limits: A statutory cycle and three budgets met show continuity and compliance with a national accounting framework, not that the framework caused the emissions reductions or that budgets four to seven will be delivered. Territorial accounts omit emissions embodied in imports, and government reports its own compliance even though an independent Climate Change Committee advises and scrutinises it.
๐ฏ๐ตJapansince 2011still operatingbears on Government foresight capacity
After the 2011 Fukushima Daiichi accident, Japan adopted a government-led mid-and-long-term decommissioning roadmap and set completion at roughly 30โ40 years after cold shutdown, treating the work as continuous risk reduction rather than a one-off rebuild.
Japan put the government in front of the operator, broke an unprecedented nuclear clean-up into staged milestones, invited international review and treated each step as risk reduction while preserving a long horizon for work that cannot be rushed safely.
Preconditions
A state capable of holding responsibility across electoral cycles and coordinating the operator, regulator and local communities
Technical institutions and international reviewers able to test milestones under deep uncertainty
A funding and monitoring system that can keep safety work moving while revising methods as evidence changes
Where else it has been used: Long-horizon decommissioning roadmaps travel to other countries facing nuclear accidents or legacy hazards, but Fukushima makes the precondition explicit: a plan must be a living risk-reduction system with independent review, not a distant completion date.
Limits: A roadmap horizon is a planning commitment, not evidence that decommissioning will finish on time, within cost or with public consent. The technical condition of fuel debris and contaminated water remains uncertain, and the government and operator report progress within their own programme. The case measures willingness to plan across decades and revise stages, not successful completion.
๐จ๐ณChinasince 2023still operatingbears on Long-horizon research commitments
China's National Bureau of Statistics, Ministry of Science and Technology and Ministry of Finance reported 225.91 billion yuan in basic-research expenditure in 2023, equal to 6.77% of national R&D spending.
China's statistical and science ministries separate basic, applied and experimental R&D in a recurring national account, while public funders protect a basic-research tranche alongside mission-oriented programmes. That makes the long-horizon allocation visible in the same frame as industrial development spending.
Preconditions
A national R&D classification that distinguishes basic research from applied work and experimental development
Stable public funders and multi-year grants that can sustain uncertain work beyond an annual policy cycle
A statistical system able to collect comparable expenditure returns from firms, universities and government institutes
Where else it has been used: A basic-research share travels only with a classification and funding system behind it: a percentage without stable grant institutions can be re-labelled spending rather than a durable long-horizon capability.
Limits: The share measures spending classification, not research quality, scientific independence, peer-review quality, frontier breakthroughs or the eventual use of results. The communiquรฉ uses national administrative statistics, reports preliminary GDP inputs and excludes Hong Kong, Macao and Taiwan, so it is not a complete measure of every Chinese research activity.
๐ต๐นPortugalsince 2022still operatingbears on Government foresight capacity
Portugal's PLANAPP operates as a Centre of Government body for evidence-based foresight, planning, monitoring, evaluation and impact analysis; its REPLAN network lists five multisectoral team remits and held its inaugural meeting on 23 November 2022.
5 team remits, named multisectoral replan team remits, 2022-11-23. PLANAPP, REPLAN presentation, retrieved 2026-09-02.
The move, and what it needed
Portugal placed foresight, planning, monitoring and evaluation in a centre-of-government body, then connected ministry planning units through REPLAN so cross-cutting work could be commissioned and reviewed across departmental boundaries.
Preconditions
A centre-of-government mandate that can convene ministries and set common planning instruments
Planning and foresight units inside each governmental area with staff able to join joint work
Shared indicators, evaluation practice and a route from foresight outputs into budget and policy decisions
Where else it has been used: A foresight network travels when it has a convening mandate and a decision route; a named unit without ministry participation or a budget link is only an advisory label.
Limits: The five is a count of named remits on PLANAPP's presentation page, not a verified staffing or output measure; the same page describes the teams as four before listing five. The record therefore evidences a formal foresight-and-coordination architecture and its inaugural date, not continuity across governments, the quality of forecasts or whether advice changed decisions.
๐ฟ๐ฆSouth Africasince 2012still operatingbears on Government foresight capacity
South Africa's Department of Higher Education and Training ran a Labour Market Intelligence Programme from 2012, published the fifth iteration of its national occupations-in-high-demand list by April 2024, and identified 138 occupations for the emerging hydrogen-economy value chain.
DHET institutionalised a research cycle that combines labour-market data, sector studies and consultations with business, labour and government, then turns the findings into recurring occupations-in-demand lists and skills assessments for education planning.
Preconditions
A department with authority to connect labour-market evidence to post-school enrolment and funding decisions
Stable research financing and partnerships with public research organisations and universities
Stakeholder channels that let employers and labour organisations challenge or refine the forecast
Where else it has been used: Foresight travels when a recurring evidence cycle changes a budget, curriculum or visa rule; publishing another list without tracing decisions that used it is only a report series.
Limits: The programme's own ministerial account reports its outputs, and a list of occupations is a planning instrument rather than proof that enrolment, funding or curricula changed. The source describes the second phase and a third phase planned for 2025; the record therefore establishes a recurring foresight function and delivered reports, not the quality or current use of every forecast.
๐ซ๐ฎFinlandsince 1993still operatingbears on Government foresight capacity
Every Finnish government since 1993 has submitted a Report on the Future to Parliament, answered by a standing Committee for the Future; Part I of the Orpo government's report, a 20-year scenario process prepared jointly by all ministries and informed by 40 dialogues on Finland's future, was submitted on 18 September 2025.
The Prime Minister's Office coordinates an interministerial foresight working group that drafts a scenario report once per electoral term, and Parliament's Committee for the Future answers it, so foresight is a recurring duty of both branches rather than a project.
Preconditions
Cross-party agreement in Parliament that long-range questions deserve a standing committee
A Prime Minister's Office strategy department with a secretariat to run the interministerial process
A consensus political culture in which coalition changes do not discard the previous government's institutions
Where else it has been used: Several parliaments have set up committees for the future modelled on Finland's; the retelling usually keeps the committee and drops the government's matching duty to report, which is the half that makes the dialogue run.
Limits: Both numbers describe the process, not its effect: a 20-year horizon and 40 dialogues show the report was made, not that any decision changed because of it. The construct asks for existence, mandate and continuity, and the strongest evidence of continuity, one report every electoral term since 1993, is stated by the Prime Minister's Office as a practice; the source does not list each report, so the count of reports since 1993 is not given here. The figures are the government describing its own work. The parliament page states that the Committee for the Future is standing and was established in 1993, and it renders by script, so a reader without JavaScript sees an empty page although the text is in the delivered data. Finnish foresight is widely cited abroad, but praise is not evidence and is not used here.
๐ช๐ชEstoniasince 2016still operatingbears on Government foresight capacity
Estonia's parliament set up a statutory Foresight Centre within the Chancellery of the Riigikogu under the Foresight Act of 2016. In 2025 the centre published four research-direction reports and 17 short reports, drawing on more than 30 researchers and experts.
The Riigikogu legislated a foresight centre inside its own chancellery and gave it a separate budget line and a duty to report to the Economic Affairs Committee every year. Its five-member council, from which MPs and ministers are barred, is appointed by the Economic Affairs Committee, a business organisation the committee names, the President, the university rectors and the Academy of Sciences (ยง 4(2)-(3)), which puts most of the council outside parliament.
Preconditions
A parliamentary committee willing to adopt the centre's yearly plan and defend its budget across governments
A pool of domestic researchers the centre can commission, since it buys most of its studies
An earlier foresight programme (the Development Fund) whose staff and practice could be moved rather than started from zero
Where else it has been used: Finland's Committee for the Future is the usual comparison. It is a standing committee of MPs, while Estonia's centre is a staffed analytical unit with an outside council. The Estonian form needs less political time but depends more on one budget line.
Limits: The two counts measure output, not influence. Nothing in the source shows that a scenario changed a budget, a law or a programme. The publisher is the Riigikogu Chancellery reporting on a unit it houses and services, so this is self-reporting with no external evaluation. The 34% spontaneous awareness figure in the same chapter comes from a Kantar Emor survey reported in the same chapter. Report counts are not comparable across countries or years: research directions vary in scope, and the 17 short reports run from agriculture competitiveness to private tutoring. The centre advises parliament, not the executive, so it says nothing about whether the government runs foresight inside ministries. It is small: in January 2025 the chair of the Economic Affairs Committee, Jaak Aab, said the committee would do its best to keep the centre's budget intact through spending cuts (Riigikogu press release, https://www.riigikogu.ee/pressiteated/majanduskomisjon-kinnitas-arenguseire-keskuse-2025-aasta-tegevuskava/), which shows continuity is not guaranteed. Continuity is real but rests on a redesign. The 2016 Act dissolved the Estonian Development Fund, which had housed the earlier foresight work, and the record does not show whether capacity was lost in that transfer. The Riigi Teataja page that carries the founding dates renders only with JavaScript.
Congress gave the Office of Technology Assessment, created under the Technology Assessment Act of 1972, the statutory function to provide early indications of the probable impacts of technology; it appropriated $21,970,000 for the office with a cap of 143 staff for fiscal year 1995, then appropriated $3,615,000 for fiscal year 1996 to carry out its orderly closure, with no more than 17 employees paid except for severance.
Congress built its own analytic office, governed by a bipartisan board of members, to give legislators early warning on technology choices, and later closed it through the appropriations bill rather than by repeal, cutting its funding by roughly five sixths in one year and capping staff for the wind-down.
Preconditions
A legislature with its own budget for in-house analytic agencies
Bipartisan support strong enough to protect a small agency through spending rounds, which failed here
A statutory mandate that survives defunding, so the function can be revived without new law
Where else it has been used: European parliaments built technology assessment offices on the model, including the German Bundestag's TAB, which cites the US office as its model; the US original closed while the copies continued.
Limits: Technology assessment is adjacent to strategic foresight, not the same thing: the office studied technologies on request from congressional committees rather than running whole-of-government scenarios. It is filed here because its statute, 2 U.S.C. 472(c), states an anticipatory function, to provide early indications of the probable beneficial and adverse impacts of the applications of technology (govinfo, US Code 2023 edition, https://www.govinfo.gov/content/pkg/USCODE-2023-title2/html/USCODE-2023-title2-chap15-sec472.htm). Both figures are appropriations, not spending or output: they show the money Congress gave, not what the office produced, and the 143 and 17 figures are statutory staff caps, not headcounts. The FY1995 figure comes from a different source, Public Law 103-283 in the Statutes at Large (108 Stat. 1423), inside a PDF on page 11 of the file (https://www.govinfo.gov/content/pkg/STATUTE-108/pdf/STATUTE-108-Pg1423.pdf). The authorising statute was not repealed and still appears in the US Code, so the office was ended by withdrawing funds rather than by abolishing its legal basis; it could be refunded without new legislation. The 1996 act also carried $2,500,000 of FY1995 funds forward to September 1996, so the FY1996 closure figure is not the whole of the wind-down money. Started is the year of the authorising Act, not the year the office first reported.
๐ต๐ฑPolandsince 1996dismantledbears on Government foresight capacity
Poland's Government Centre for Strategic Studies, attached to the Prime Minister, with tasks of strategic programming and forecasting of economic and social development set in article 14k of the Act of 8 August 1996 on the Council of Ministers, was liquidated as of 31 March 2006 by the Act of 17 February 2006, in the year its budget had been planned at PLN 10,016 thousand.
Parliament gave a strategic studies centre attached to the Prime Minister its own statutory tasks and budget part, then a new government folded it into the Prime Minister's Chancellery and a line ministry as part of an administrative savings programme.
Preconditions
A statute giving the unit its own budget part, which made it visible as a line to cut
A government programme of administrative consolidation that counted a separate agency as overhead
Receiving ministries willing to absorb the tasks, which let the unit be closed without a visible gap
Limits: The figure is planned spending in the government's draft 2006 budget, quoted in the bill's regulatory impact assessment, not an outturn, and the same assessment expected savings of about half of it. The bill proposed liquidation on 28 February 2006; the act as passed (Dz.U. 2006 nr 45 poz. 319, ISAP, https://isap.sejm.gov.pl/isap.nsf/DocDetails.xsp?id=WDU20060450319, PDF) set 31 March 2006, and the claim uses the enacted date. The centre's statutory work was forecasting and long-term strategic programming, which sits closer to development planning than to scenario-based foresight. Its tasks did not vanish: the act moved impact assessment, foreign policy concepts and analyses ordered by the Prime Minister to the Chancellery of the Prime Minister, and the bill moved strategy and forecasting to the minister for regional development, so what ended is a dedicated unit with its own budget line, not every forward-looking task. Started is 1996, the year of the Act on the Council of Ministers whose article 14k held the centre's tasks; neither source gives the centre's founding date or first year of operation.
๐ฉ๐ชGermanysince 1990still operatingbears on Government foresight capacity
Since 1990 the German Bundestag has kept an Office of Technology Assessment, run under contract by the Karlsruhe Institute of Technology, whose tasks include horizon scanning of scientific and technological trends at early stages; by 2026 it reported more than 200 projects and over 400 publications.
The Bundestag contracts a university technology assessment institute to run a dedicated office for parliament, steered by its research committee, which picks topics and approves reports before publication.
Preconditions
A national research institute with an established technology assessment group able to host the office
A standing committee willing to act as steering client across electoral terms
A contract model that lets parliament buy analytic capacity without creating a civil service unit
Where else it has been used: The office names the US Congress's OTA as its model; Germany kept the parliamentary office when the US closed its own in 1995, but placed it with a contractor rather than inside parliament.
Limits: Technology assessment is adjacent to strategic foresight: most TAB projects assess specific technologies on request from the Bundestag committee for research and technology assessment. It is filed here because the same page lists horizon scanning, the observation of trends in early stages of development, among its tasks and describes its aim as forward-looking analysis, and the site carries a separate foresight section. Both figures are cumulative lower bounds (more than 200, over 400) with no date stated, so asOf is the retrieval date. TAB reports on itself, and it is operated by a university research institute (ITAS at KIT) under a contract with the Bundestag rather than by parliamentary staff, so continuity depends on contract renewal. The counts show output, not influence on legislation, and they serve parliament, not the federal executive.
๐ฐ๐ทSouth Koreasince 2018still operatingbears on Government foresight capacity
Korea's National Assembly enacted the National Assembly Future Institute Act on 12 December 2017 and opened the institute on 28 May 2018 with a statutory duty to predict and analyse changes in the future environment and derive mid to long term national development strategies; its English-language research report archive listed 203 reports in 2026.
The National Assembly legislated a futures research institute attached to the National Assembly and gave it a statutory mandate for long-range national strategy, so parliament holds its own foresight capacity independent of the government of the day.
Preconditions
Cross-party support for a statute rather than a Speaker's initiative that could lapse
A parliamentary research ecosystem (budget office, research service, library) to house it beside
A domestic research community the institute can recruit from
Limits: The institute reports on itself. The count is a running total shown on an archive board ("Total 203"), not a dated annual figure, so asOf is the retrieval date, and it covers the English-language listing only; the Korean site organises reports differently and may hold more. The newest report in the English listing is dated 30 December 2024, so the count shows nothing published in English since then. A report count measures output, not whether any law or budget changed. The institute serves the National Assembly, not the executive, so it says nothing about foresight inside ministries. The founding dates and the Article 1 purpose come from the institute's own History and Vision pages (https://nafi.re.kr/home/eng/contents.do?menuPos=10 and menuPos=11), not from the statute text itself.
๐จ๐ญSwitzerlandsince 2007still operatingbears on Government foresight capacity
Switzerland's Federal Chancellery runs a long-term, continuous situation and environment analysis for the Federal Council that looks 10 to 15 years ahead; it produced Herausforderungen in 2007 with expertise from 30 federal offices, Perspektiven 2025 in 2011, Perspektiven 2030 in 2014, the Schweiz 2030 and Schweiz 2035 reports and the scenario report Perspektiven Schweiz 2040, which feed the Federal Council's legislature planning.
6 reports, reports in the chancellery situation and environment analysis series named on the page, 2026-10-02. 30 federal offices, federal offices whose expertise fed the 2007 report herausforderungen, 2007. Swiss Federal Chancellery, Lage- und Umfeldanalyse, retrieved 2026-10-02.
The move, and what it needed
The Federal Chancellery, as staff office of the Federal Council, pools experts from across the federal offices into a standing interdepartmental staff that writes a futures report timed to each legislature planning cycle.
Preconditions
A collegial executive whose four-year legislature planning gives the report a fixed customer and date
A neutral staff office above the departments that can convene them without belonging to one
Federal offices willing to lend analysts to a cross-cutting exercise
Limits: The report count is ours, made from the six titles the page names (Herausforderungen, Perspektiven 2025, Perspektiven 2030, Schweiz 2030, Schweiz 2035, Perspektiven Schweiz 2040). Both numbers describe the process, not its effect: six reports and 30 contributing offices show the analysis was made, not that a decision changed because of it. The Chancellery describes its own work and states that the reports serve as a basis for legislature planning; the source does not show how. The page gives publication years only up to 2014 and does not date Schweiz 2030, Schweiz 2035 or Perspektiven Schweiz 2040, so the series after 2014 is shown as existing, not as dated. Started is the year of the earliest report the page names, not the founding of the Perspektivstab, which the page does not date. The 30 offices figure belongs to the 2007 report only. The Chancellery is the Federal Council's staff office, so this is a staff function, not a minister's unit.
๐ซ๐ทFrancesince 2009still operatingbears on Government foresight capacity
The Bureau of the French Senate created a foresight delegation in April 2009, succeeding the planning delegation set up by the 1982 planning reform law, with a mandate to study changes in society and the economy and to build scenarios; it has 36 senators appointed in proportion to the political groups and adopted reports in every year from 2010 to 2022 listed on its page.
A chamber's governing bureau turned an existing planning delegation into a standing foresight body of members with free choice of topics and a duty to write scenarios, which keeps it outside the legislative calendar.
Preconditions
An earlier planning delegation whose seat and staff could be repurposed
A chamber with longer terms and partial renewal, which suits long-horizon work
Freedom from bill work, which lets members choose topics without committee pressure
Where else it has been used: The page names Finland's and Lithuania's committees for the future as the only European parliamentary equivalents; the French version is a delegation of senators served by Senate staff rather than a research institute.
Limits: Membership measures size, not influence: 36 senators on a delegation say nothing about whether a report changed a law, and the delegation has no legislative power by design. It informs the Senate only; the National Assembly has no equivalent, as the page itself says, and the figure says nothing about foresight in the executive. The page's report list stops at 2022, although the delegation's main page shows a report dated June 2026, so the list is incomplete and no report count is given here. After the 2026 Senate renewal the main page states the delegations will be reconstituted, so the current membership is pending. It is filed against the foresight construct because its rules instruct it to draw up scenarios of change, which is foresight practice rather than technology assessment (France's technology assessment office, OPECST, is a separate body).
๐จ๐ฑChilesince 2012still operatingbears on Government foresight capacity
Chile's Senate created the Commission on Future Challenges, Science, Technology and Innovation in April 2012, modelled on Finland's Committee for the Future, with the Academy of Sciences, National Prize winners and university rectors as permanent members; by July 2024 more than 700 scientists had worked in more than 20 of its expert tables.
A Senate committee co-opts the national science establishment as permanent members and runs expert tables that turn scientific frontiers into draft strategies and bills, giving parliament a standing futures agenda without a research staff of its own.
Preconditions
A science academy and university rectors willing to sit with legislators
A senator who chaired it for a decade, which gave it continuity
A public outreach event, Congreso Futuro, that keeps the agenda visible
Where else it has been used: It was modelled on Finland's Committee for the Future; Chile kept the committee and replaced the Finnish government's duty to report with expert tables drawn from academia.
Limits: Both figures are lower bounds (mรกs de) and sit inside a four-page PDF, on page 3. They count participation, not influence: the same note lists five bills in the commission's legislative work, four members' motions and one government bill, all still in their first or second constitutional stage in July 2024, including two to create a national strategic foresight agency (Boletรญn 16.061-07) and a new foresight institution (Boletรญn 16.441-19). That means Chile had a parliamentary foresight committee but, as of the note's date, no statutory executive foresight body, and this record does not claim one. The commission is a standing Senate committee; the note does not count its sessions or reports. The figures are compiled by the Congress library at a director's request and are not independently audited. The Senate's own committee page (https://www.senado.cl/actividad-legislativa/comisiones/941) still lists the commission in 2026.
๐จ๐ฆCanadasince 1996still operatingbears on Government foresight capacity
Policy Horizons Canada, which describes itself as the Government of Canada's centre of excellence in foresight and reports through the Deputy Minister of Employment and Social Development Canada, published Disruptions on the Horizon 2024, which identified and assessed 35 possible disruptions using survey input from around 500 stakeholders, colleagues and foresight experts across the Government of Canada and beyond.
A small dedicated foresight unit inside the federal public service, governed by a deputy minister steering committee, runs recurring horizon scans and surveys hundreds of officials and outside experts to rate disruptions by likelihood, impact and timing, and trains and networks foresight practitioners across departments.
Preconditions
A permanent home and budget line inside the public service that survives changes of government
Senior sponsorship at deputy minister level so departments answer the survey and read the result
A network of departmental foresight practitioners to carry the work into policy files
Where else it has been used: Policy Horizons' published foresight training manual is listed as a toolkit by the OECD Observatory of Public Sector Innovation; this was seen in search results only and not verified at source.
Limits: The unit is reporting on its own work, and the 500 is stated as 'around 500', an approximate count of survey inputs rather than a measure of whether any department used the scan; the report itself says its content does not necessarily represent the views of the Government of Canada. The description as centre of excellence is in the report's disclaimer; the reporting line through the Deputy Minister of Employment and Social Development Canada is stated on the unit's about page (https://horizons.service.canada.ca/en/about-us/index.shtml), not in the report. Both figures appear on the web version at the source URL and in the PDF edition (pages 5 and 7 of https://horizons.service.canada.ca/en/2024/disruptions/Disruptions_on_the_Horizon_2024_report.pdf). No published budget, staff count or uptake figure was found: Employment and Social Development Canada's 2024-25 Departmental Results Report does not mention the unit, so the record evidences a standing federal function producing a dated whole-of-government scan with a deputy minister steering committee named in its acknowledgements, not the influence of that scan on policy. Whether the Disruptions report is annual was not verified: no 2023 or 2025 edition was found at the corresponding addresses. Started is 1996, when secondary histories place the unit's predecessor, the Policy Research Secretariat, in the Privy Council Office, renamed Policy Research Initiative in 2000 and Policy Horizons Canada around 2011 to 2012; that lineage was not verified on an official page today and the foresight mandate as now worded dates from the later renaming. The unit sits inside a line department rather than at the centre of government, which limits its reach.
๐ธ๐ชSwedensince 2011delivered and closedbears on Government foresight capacity
The Swedish government appointed Framtidskommissionen, chaired by Prime Minister Fredrik Reinfeldt, in November 2011 to identify the societal challenges Sweden faces in the longer term, and by its final report in March 2013 it had held eight meetings, organised 40 open meetings, seminars and workshops, published twelve background reports and an anthology, and carried out four sub-inquiries.
The head of government chairs a time-limited commission of ministers, social partners and researchers that commissions background reports and runs public meetings across the country to name long-term challenges, then publishes one report and disbands.
Preconditions
A head of government willing to chair it personally
A secretariat inside the Government Offices for the life of the commission
A successor structure to carry the work on, which was missing here
Where else it has been used: The format recurred under the following government as Mission: The Future (2015 to 2016), which was also time-limited.
Limits: The commission reports on itself, and the figures count activities, not influence on policy; the foreword says identifying challenges, not proposing how to meet them, was its purpose. The figures are in a 316-page PDF, on PDF page 17 (printed page 14) and repeated on PDF page 310 (printed page 291), where the November 2011 appointment and the membership are also stated. It was a time-limited commission of ministers from one coalition and outside members, and it ended with the report: it evidences the existence and mandate of a centre-of-government foresight exercise and, for the construct's third element, its lack of continuity. The next government ran its own time-limited project, Mission: The Future, from February 2015 to April 2016, with a Minister for Strategic Development from 3 October 2014 to 25 May 2016 (government.se, https://www.government.se/government-of-sweden/prime-ministers-office/mission-the-future/, read in a browser because the site returns 403 to scripts), and Sweden's innovation agency Vinnova wrote in 2024 that in strategic institutionalisation of foresight for policy Sweden is 'a weak case' (Strategic Foresight for Innovation, PDF page 32). Status is concluded rather than dismantled because both vehicles were set up as time-limited and no source read today shows either was cut short.
๐ฒ๐ผMalawisince 2020still operatingbears on Government foresight capacity
Malawi's National Planning Commission, which signed a UNDP initiation plan in November 2020 to create a foresight unit and horizon scanning capability, reported for 2022/23 that it produced a foresight modelling report with modelled scenarios for the Malawi 2063 vision, policy briefs on three of the vision's pillars with UNDP and the finance ministry, and foresight modelling training for planners from ministries and academia.
The statutory planning commission that owns the national long-term vision built a small foresight and modelling capability with UNDP support and used it to test the vision's targets against scenarios and to train planners in ministries.
Preconditions
A statutory planning commission with a mandate over a long-term vision
Donor funding and technical partners to seed the capability
Ministry planners and academics willing to adopt the methods
Where else it has been used: The commission organised a learning visit on foresight planning to South Africa in the same year, so the method was borrowed from a regional neighbour as well as from UNDP.
Limits: The commission reports on itself, in a PDF (section 3.3, printed pages 38 to 39, PDF page 42). The metric is a single report, and the text does not say how many policy briefs were written, only that they covered the urbanisation, industrialisation and agricultural productivity pillars, so the delivery figure is thin. The foresight work leans on UNDP: the unit was set up under a six-month UNDP initiation plan (NPC news item, 26 November 2020, https://npc.mw/2020/12/npc-undp-sign-agreement-to-initiate-implementation-of-foresight-project/) and the 2022/23 briefs were produced with UNDP, so the record cannot separate national capacity from donor support. The commission is primarily a development planning body that monitors the Malawi 2063 vision and its first ten-year implementation plan, which sits closer to planning than to open-ended scenario foresight; its modelled scenarios are tied to vision targets. No later annual report was read, so status rests on 2022/23, and the report gives no staff count or budget for the foresight function.
Cรดte d'Ivoire's Ministry of Planning and Development completed its fourth national prospective study, Cรดte d'Ivoire 2040, under the technical responsibility of the Bureau National de la Prospective et de la Veille Stratรฉgique, after earlier national studies in 1973 to 1974, 1983 to 1984 and 1993 to 1995, and built eighteen thematic and three global scenarios to the 2040 horizon.
The planning ministry keeps a standing foresight bureau with the rank of a directorate general that runs a national prospective study about once a decade, using an expert group drawn from universities and ministries and national and local commissions to debate the results.
Preconditions
A planning ministry with a mandate for long-term perspective, not only five-year plans
University and ministry staff available for a multi-year expert group
Political sponsors willing to adopt the study's vision, which the 2025 study lacked
Where else it has been used: National prospective studies of the same method ran in several West African states, including Mali, whose 1997 to 1999 Mali 2025 study its government evaluated at national prospective days in January 2014.
Limits: Both figures are in a 12-page PDF (page 5 for the four studies, page 6 for the scenarios); the date comes from the file name (final version, 13 March 2017), not from the document. The ministry reports on itself, and the figures count studies and scenarios, not influence: the same summary says the 2025 study did not lead to a plan and that the 1999, 2002 and 2010 to 2011 crises followed scenarios the government had not used. The four studies span four decades with a gap the summary attributes to structural adjustment, so continuity is episodic, and the bureau's own creation date is not given; started records the first national study. The ministry's own site (plan.gouv.ci/structure/5, read 2 October 2026) still lists the BNPVS among its attached directorates and keeps a prospective and strategic watch section, which supports the operating status, but no dated output after 2017 was read at an official source; a national prospective day in May 2024 and a sectoral foresight manual in September 2025 were seen only in press. The study claims a strategic watch mechanism was to follow; no figure for it was read.
๐ณ๐ดNorwaysince 1999still operatingbears on Government foresight capacity
Norway's Board of Technology, an independent public advisory body created in 1999 on the Storting's initiative that also serves as secretariat for the Storting's technology group, delivered eight publications and held eleven meetings, four of them for the Storting's technology group, in 2025 on operating income of NOK 13.8 million.
The government appoints a 15-member council for four-year terms with a small independent secretariat that chooses its own projects, convenes expert groups and lay panels, and briefs the Storting through a cross-party technology group it services.
Preconditions
A parliament that wants an independent source of technology advice and a cross-party group to receive it
A stable budget line outside any single policy ministry
A council free to choose topics, set in its statutes
Where else it has been used: Denmark's equivalent body, also called Teknologiraadet, was defunded and abolished in 2012 (dnk-teknologiraadet); the Norwegian one survived on a similar design.
Limits: The board reports on itself to its parent ministry; the figures are in a 41-page PDF, publications and meetings on PDF page 5 and income on page 3. It is a technology assessment body more than a strategic foresight unit, though its 2025 work included a project on strategic foresight methods and an annual trend report for the Storting. Publication counts measure output, not influence on law or budgets. It is small: the secretariat had eight permanent staff in 2025. Its funding line moved in 2025 from the trade ministry's research council chapter to the research council's allocation letter from the education ministry, which the report describes as an administrative change; continuity since 1999 is stated on the board's own about page (https://teknologiradet.no/forsiden/om-oss/), not in an external evaluation.
๐ฉ๐ฐDenmarksince 1995dismantledbears on Government foresight capacity
Denmark's parliament abolished the statutory Board of Technology, set up by Act no. 375 of 14 June 1995 to assess technology and report to the Folketing, by Act no. 559 of 18 June 2012, after an all-party budget agreement of 15 November 2011 withdrew its annual grant, which had been DKK 10.5 million in 2011, to fund the research reserve.
A statute gave a self-governing board its own Finance Act grant and a duty to report to parliament each year; an all-party budget deal then reallocated that grant to research funding, and parliament repealed the act unanimously.
Preconditions
A separate, visible grant line that could be reallocated in one budget agreement
No parliamentary committee that owned the board and would defend it
A foundation route that let staff and methods continue privately, which removed resistance to closure
Where else it has been used: Poland's strategic studies centre (pol-rcss) was closed in 2006 by a similar budget-line logic; Norway's board of the same name (nor-teknologiradet) survived.
Limits: The figure was read in the bill's XML (https://www.retsinformation.dk/eli/ft/201112L00079/xml), because the eli page renders with JavaScript; the vote (104 for, none against) and the law number and date come from the Folketing's open data API (oda.ft.dk, case 9940). The board was a technology assessment body with participatory methods, which is adjacent to strategic foresight rather than the same thing. Its work did not stop: Aktstykke 65 of 20 March 2012 let it create a foundation, Fonden Teknologiraadet, which continues as a non-profit that sells projects, so what ended is the statutory public function and its guaranteed grant, not every activity. The bill puts the annual saving at about DKK 10 million and sets aside DKK 4.5 million for winding up in 2012. The 1995 act replaced an earlier Technology Board (Teknologinaevnet) under a 1989 consolidated law, so the public function predates the start year given.
๐ง๐ชBelgiumsince 1959still operatingbears on Government foresight capacity
Belgium's Federal Planning Bureau, which produces the government's economic forecasts and long-term demographic, ageing, energy and transport outlooks, published 34 publications in 2025 and employed 91 full-time equivalents at the end of the year, down from 96 at the end of 2023, after it stopped replacing departing staff in 2024 to meet savings targets.
91 FTE, staff, full-time equivalents at 31 december, 2025-12-31. 96 FTE, staff, full-time equivalents at 31 december (earlier level, not shown to be a peak), 2023-12-31. Bureau federal du Plan, Rapport d'activite 2025, retrieved 2026-10-02.
The move, and what it needed
A statutory, de facto independent planning bureau receives mandates from government, parliament and social partners, and publishes forecasts and long-term outlooks that the federal budget and pension debates rely on; in a savings drive the government cut its grant and it stopped replacing leavers.
Preconditions
A statute that makes its forecasts the official basis for the budget
Social partners and parliament as co-principals, which protects independence
A core grant large enough to keep long-term modelling teams, which savings measures are now eroding
Limits: The bureau reports on itself; the 91 FTE (80 funded by the federal grant, 11 by contracts) and the hiring freeze are on PDF page 12, the 34 publications on page 19. The 96 FTE is from the 2023 activity report (https://www.plan.be/sites/default/files/documents/OTH_ACTIV_2023_FR.pdf) and the 2024 report gives 94 (https://www.plan.be/sites/default/files/documents/OTH_ACTIV_2024_FR.pdf). The record is filed as operating, not eroded: the fall is five FTE, about 5 percent, from a 2023 level that is not shown to be a peak (earlier reports were not found at the same address), and output held, with 37 publications in 2023, 45 in 2024 and 34 in 2025. The staff trend and the bureau's statement that it no longer replaces leavers to meet savings targets are a documented squeeze on capacity, and a later report showing a deeper loss would justify eroded. Most of the bureau's work is medium-term forecasting and policy costing, not scenario-based strategic foresight; its long-term population, ageing and sustainable development outlooks are the foresight part. The report says nearly 40 percent of staff will retire within ten years, which is a risk stated by the bureau, not a measured loss. The 1959 start is the Act of 23 December 1959 named on the bureau's legal basis page (https://www.plan.be/en/about-us/legal-basis).
๐ฌ๐ทGreecesince 2022still operatingbears on Government foresight capacity
Greece's Special Secretariat of Foresight published its Megatrends 2040 report and seven megatrend notes in February and March 2022 and then six further foresight studies to September 2026, including Generative AI Greece 2030 (January 2024), The Future of Work and Skills in Greece (February 2025) and Mediterranean Foresight (September 2026).
A special secretariat at the centre of government runs foresight studies with universities and research centres, trains a network of public servants in foresight tools, and represents Greece in EU foresight networks.
Preconditions
A secretariat placed close to the head of government
Research partners to co-produce studies, since the unit is small
EU foresight networks that supply method and a reporting rhythm
Limits: Both counts are ours: the two studies list pages show no dates, so each item's date is the datePublished stamp in the metadata of its own page (for example Mediterranean Foresight, 29 September 2026; The Future of Work and Skills, 24 February 2025), and the site states no total. Of the six later items, several were made with partners (the National Centre for Social Research on internet use, the UNESCO Chair on Futures Research on innovation to 2035, and the High Level Advisory Committee on AI for the AI blueprint), so the Secretariat's own output is smaller; reposted EU Strategic Foresight Reports are excluded. Output fell from a launch burst in 2022 to one or two studies a year. The site's footer and its July 2026 press release now name the body the Special Secretariat of Long-term Planning, a renaming whose date and legal basis were not read; the 2022 start is the date of the first study, not the founding act, which was not read. A count of studies says nothing about use in policy.
๐ฑ๐นLithuaniasince 2022delivered and closedbears on Government foresight capacity
Lithuania's Government Strategic Analysis Center (STRATA) followed six thematic public discussions on the country's future in May and June 2022 with nine foresight studies on megatrends and scenarios by policy area, which fed the vision and strategic ambitions of the draft state progress strategy Lietuva 2050.
The government's strategic analysis centre ran public thematic discussions, commissioned studies that turned their megatrends into sector scenarios and SWOT analyses, and used them in workshops that drafted the long-term state strategy.
Preconditions
A government-owned analysis centre with a mandate in strategic planning
A strategy cycle that asks for a long-term vision
Experts and citizens willing to take part in discussions
Limits: The live site (https://strata.gov.lt/) returned 403 to scripts, so an archived copy was read. The counts are stated by the centre about its own work and measure output, not influence; the page does not date each study, and 2022 is the year of the discussions. The exercise was tied to one strategy, so it shows a foresight delivery and not, by itself, a standing foresight function; STRATA's wider role (research, impact assessment, training) continues but was not counted. Whether the final Lietuva 2050 strategy adopted the scenarios was not read.
๐ฒ๐ฆMoroccosince 2007still operatingbears on Government foresight capacity
Morocco's Royal Institute for Strategic Studies, announced in the 2003 Throne Speech to strengthen the country's capacities for analysis, adaptation and anticipation and created by Dahir No 1-07-183 of 30 November 2007, published the twelfth edition of its strategic dashboard on Morocco's international positioning on 28 November 2024 and was still issuing general reports in 2025.
The head of state created by decree a strategic studies institute with a small project-based staff and associated researchers, which runs recurring monitoring products, national surveys and studies on questions he refers to it.
Preconditions
A head of state who commissions long-range studies directly
A decree giving the institute a standing mandate and a guidance committee appointed from the top
Associated academics to staff time-limited study projects
Limits: The institute reports on itself, and the figure is an edition number in the title of a listed report, not a count of output the institute publishes. The dashboard tracks Morocco's position in international indicators, which is strategic monitoring (veille) rather than scenario-based foresight, and the institute describes itself as a think tank, a watch function and a forum, working on questions referred by the King; the mandate and the Dahir are on the presentation page (https://www.ires.ma/fr/ires/presentation), where the 2003 Throne Speech that announced it names anticipation as its purpose. It sits outside the elected government, attached to the monarchy, so it is a centre-of-state body rather than a cabinet unit, and nothing read today shows its work entering a government decision. The page gives no budget, staff count or total of reports. Started is 2007, the year of the founding Dahir; the publication list begins in 2009.
๐ณ๐ฟNew Zealandsince 2020still operatingbears on Government foresight capacity
New Zealand's Public Service Act 2020 required every departmental chief executive to publish a Long-term Insights Briefing on future trends, risks and opportunities at least once every three years, and nineteen were published in the first round, which concluded in mid-2023, four of them produced jointly by several agencies.
A public service statute obliges every department head to consult the public on a topic, write a briefing on medium and long-term trends with policy options, and table it at least once every three years, with the centre of government publishing guidance and reviewing the round.
Preconditions
A statutory stewardship duty in public service law
A centre-of-government policy team to issue guidance and review rounds
Departments with analysts able to do futures work alongside daily business
Limits: The live dpmc.govt.nz page returns 403 to scripts, so it was read from a Wayback Machine copy of 18 June 2026; the page is the coordinating department reporting on a duty it guides. It is a statutory duty spread across departments, not a single foresight unit with its own budget and staff, and the count measures briefings published, not their quality or influence: DPMC's own review of the first round assessed the process only and said so. Started is 2020, the year of the Act that created the duty; the page does not give the year the first briefings were published. The briefings are think pieces independent of ministers and are not government policy. Whether the second round has kept the same number of briefings is not stated on the page.
Singapore's Department of Statistics ICT-skills table reports 99.0% in the total row for residents aged 16โ74 in 2025, while 55.2% could use a basic arithmetic formula in a spreadsheet; the government uses the series to track SDG 4.4.1 and its Digital Skills for Life framework.
Singapore pairs a recurring skills measurement with a public baseline framework, Digital Skills for Life, and targeted support such as Seniors Go Digital. The combination makes digital capability a population service obligation rather than a one-off computer course, while the task rows reveal where basic access still masks deeper gaps.
Preconditions
A statistical agency and digital ministry able to repeat a common task taxonomy and publish the results
Universal connectivity and trusted public services that give adults reasons to practise digital tasks in daily life
Accessible coaching, devices and safety guidance for older and lower-income residents who are least likely to learn informally
Where else it has been used: Digital-skills programmes travel when measurement, everyday use and targeted help reinforce one another; connectivity statistics alone can conceal a population that still cannot complete consequential tasks.
Limits: The table records task-specific self-reported or survey-derived skills for Singapore residents, not independently tested proficiency, workplace transfer or equal access. The age base changed from residents aged 15 and above in 2016โ2020 to ages 16โ74 from 2021, and a high total can hide weak performance on harder tasks such as spreadsheet work.
๐ต๐นPortugalsince 2007dismantledbears on Adults doing digital tasks
The European Commission's eSchool case study records more than 1.3 million people requesting laptops and broadband through Portugal's e.escola programme and more than one million laptops delivered in its first two years; the Portuguese government suspended the programme in 2011.
The state used a dedicated fund and mobile-operator partnership to put subsidised laptops and broadband into the hands of students, teachers and adult trainees, using schools and private partners for decentralised delivery.
Preconditions
A mechanism to target subsidies across students, teachers and low-income households
Private network operators able to supply devices and connectivity at national scale
Maintenance, training and replacement funding that lasts beyond the initial hardware handout
Where else it has been used: Device-and-connectivity programmes travel when skills support, maintenance and a durable funding model accompany procurement; a rapid hardware push can leave a capability gap when the subsidy is withdrawn.
Limits: Device distribution measures access, not sustained use, adult or student digital competence, learning outcomes or affordability after the subsidy. The scale figures come from the programme case study, while the suspension is recorded in a Portuguese government archive; the partnership and outstanding state payments also complicate any simple claim of a completed success.
๐ฎ๐ณIndiasince 2017delivered and closedbears on Adults doing digital tasks
India's PMGDISHA rural digital-literacy scheme enrolled 73,571,965 people, trained 63,941,718 and certified 47,836,188 by July 2024 through 20 hours of instruction in five digital-task modules.
MeitY used village-level centres, a standard five-module curriculum and external certifying agencies to move a digital-literacy offer through rural India at household scale, with certificates uploaded to DigiLocker.
Preconditions
A delivery network able to reach Gram Panchayats and rural households
A common curriculum covering devices, internet, citizen services and digital payments
Independent assessment and an identity-linked certificate that makes completion portable
Where else it has been used: Digital-literacy programmes travel when training is tied to tasks people can immediately use and completion is independently checked; enrolment without post-training usage data can become a counting exercise.
Limits: The government reports administrative training and third-party certification totals, not independently tested digital competence or continued use of services. The scheme covered rural areas and one person per targeted household, so it is not a population-wide adult-skills rate; the source also does not report completion quality, attrition or post-training use.
๐ฒ๐ฝMexicosince 2015delivered and closedbears on Adults doing digital tasks
Mexico's communications ministry operated Puntos Mรฉxico Conectado centres in all 32 states, reporting more than 410,000 enrolled members and over 260,000 course enrolments by June 2017.
The communications ministry put free digital classrooms in every state, then used equipment and curriculum partnerships with firms such as Samsung to turn connectivity points into places where people could practise new tools.
Preconditions
A national network of staffed community centres rather than connectivity alone
Public funding for access and instructors across all states
Private partners willing to provide current equipment and course material
Where else it has been used: Digital-learning hubs travel when the state supplies trusted physical access and partners refresh the content; registrations without completion data should not be mistaken for population-wide proficiency.
Limits: The ministry reports registrations and course enrolments, not a comparable adult-skills test, completion rate or retained ability. The article is a programme announcement about a partnership and does not establish how many of the 410,000 members completed training or whether the network continued at that scale after 2017.
Germany funds the Fraunhofer-Gesellschaft's institutes in proportion to what they earn from industry, and in 2025 the society performed 3.2 billion euros of contract research, 966 million euros of it paid by industry.
The state pays each institute a base grant that grows with the industrial income it wins, so every institute director is structurally forced to sell applied research to firms while the public base keeps the science honest.
Preconditions
An industrial base, especially the Mittelstand, that buys research it cannot do in-house
A funding rule stable across decades, since the incentive works only if institutes believe it
Careers competitive enough to keep researchers who could leave for industry
Where else it has been used: Carnot in France and Catapult in the United Kingdom are explicit copies. Most copies struggle to replicate the base-grant-follows-industry-income rule, which is the part that does the work.
Limits: The figures are the society's own. Revenue measures demand for applied research, not whether knowledge moves between universities and firms, which is what the indicator asks about, and Fraunhofer sits beside the university system rather than inside it, so a reader can argue the model routes around university-industry collaboration as much as it evidences it. Roughly two-thirds of the budget is still public money.
๐ฐ๐ชKenyasince 2007still operatingbears on Public-private collaboration
Kenya's central bank let a phone company build a national payments system in 2007 and supervised it into ubiquity, and by June 2026 Kenya had 94.2 million registered mobile money accounts and 572,104 agents, in a population of about 55 million.
The central bank issued a letter of no objection instead of demanding a banking licence, set conditions on trust accounts holding customer float, and wrote the e-money regulations after the system worked, legalising what it had learned rather than licensing what it could imagine.
Preconditions
A regulator senior enough to carry the risk of saying yes
A mobile operator with national agent reach and its own fraud controls
A banking sector too small in reach to block the entrant
Where else it has been used: Mobile money spread across Africa and South Asia, but mostly where regulators copied the permissive sequencing. Where banking law was applied first, including in some of Kenya's neighbours for years, adoption stalled.
Limits: The delivery was Safaricom's, a majority-private company, so the record bears on public-private collaboration only through what the state did: permit, then supervise, then regulate interoperability. Registered accounts overstate people, since many Kenyans hold several, and account counts say nothing about fees, which are materially higher than card rails elsewhere. A single dominant operator is also a concentration the indicator's construct does not reward.
๐จ๐ญSwitzerlandsince 2004still operatingbears on Public-private collaboration
Switzerland runs most upper-secondary education through firms: of 218,259 students enrolled in vocational education and training in 2020, 90.6 percent, 197,782, were apprentices trained inside companies under federal law.
Federal law makes firms, cantons and professional associations joint owners of training: associations write the curricula for their own occupations, firms hire and pay apprentices to do real work, and the state examines and certifies, so the labour market signals what to teach.
Preconditions
Employer associations strong enough to write and update curricula
Apprentice wages low enough that training pays for the firm
Social standing for the vocational route, so strong students choose it
Where else it has been used: Germany and Austria run sibling systems; exports to countries without employer associations mostly produce school-based programmes with the dual name, which drops the part that works.
Limits: Enrolment counts participation in the system, not what the training is worth; the wage and employment evidence sits in separate studies. The system long predates the number: 2004 is the current federal act, not the start of apprenticeship in Switzerland, and the record cannot show how much of the outcome is the law versus a century of employer habit. Firms train because it pays them net of apprentice output, and that arithmetic, not the statute, is the load-bearing part.
๐ง๐ทBrazilsince 1974still operatingbears on Public-private collaboration
Brazil and Paraguay built and operate Itaipu through a treaty-based binational entity, which reached 14,000 megawatts of installed capacity and has supplied both countries since the first unit entered operation in 1984.
Brazil and Paraguay converted a contested shared river into a jointly governed operating institution with its own legal personality, fixed allocation rules and a long-lived technical system.
Preconditions
A legal agreement that specifies rights, obligations and decision authority
A physical project large enough to create mutual dependence
Technical and financial institutions able to keep the partnership operating for decades
Limits: Installed capacity shows the scale of the physical system, not the quality of the bilateral relationship, the project's cost or the distribution of its benefits. A treaty structure can hold a joint asset together while leaving wider coordination between the countries unresolved.
๐ง๐ทBrazilsince 2013still operatingbears on University-industry collaboration
Created in 2013, EMBRAPII funds industrial research that companies contract with accredited public and private research units, and it contracted a record 811 new projects in 2025, 25.35% more than in 2024.
Public funding follows an industry's defined technical problem into a research unit, sharing early risk while leaving the company and researchers responsible for solving a concrete challenge.
Preconditions
Research units with equipment and technical depth ready to work with firms
Companies able to define a problem and contribute a financial counterpart
A funding body with enough flexibility to contract applied work faster than ordinary grants
Limits: Project counts show the throughput of a funding network, not the commercial value, technical success or additionality of the research. EMBRAPII is a social organisation under a management contract with four federal ministries, and the figure is its own report of its activity, drawn from its Power BI panel on 3 February 2026. Its 90 accredited units are research institutions of several kinds, university units among them, and the page does not say how many of the 811 projects ran in a university, so the record evidences firm-research-unit collaboration of which university collaboration is a part, not all of it. Projects vary widely in scale and maturity. The live page answered 'Conteรบdo temporariamente indisponรญvel' (HTTP 503) on 2 October 2026, so the source is the Internet Archive copy of 22 April 2026, which carries the 811 figure.
๐ง๐ทBrazilsince 2007delivered and closedbears on University-industry collaboration
To foster university-firm interaction, the Federal Government set up SIBRATEC, coordinated by the Ministry of Science and Technology with FINEP, BNDES and INMETRO taking part, as a national system of 56 research and development networks by 2010: 14 innovation-centre networks, 20 technological-service networks and 22 state-level extension networks.
The ministry linked research groups, technology centres and state-level extension networks so firms could reach technical support outside universities and turn a production problem into a shared R&D project.
Preconditions
Public research and technology centres able to work on firms' production problems
A coordinating ministry with funding routes through FINEP and partner agencies
Small firms able to identify a technical problem and use an external support network
Limits: The network count records an institutional architecture, not how many firms worked with a university, how much research was done jointly or what it produced. The report states the purpose (university-firm interaction) and the network count, but the networks also include technology institutes and metrology laboratories, and the extension networks support small firms with technical services rather than joint research. It is the ministry's own account of its plan and does not evaluate outcomes. The started year is the first year of the 2007-2010 action plan under which SIBRATEC was implemented; the report gives no creation date. The report says nothing about the networks after 2010, so the concluded status records the end of this delivery period, not the programme's fate. The ministry repository did not answer on 2 October 2026, so the source is the Internet Archive copy of 29 April 2025; the passage is on page 74.
๐ฐ๐ทSouth Koreasince 1961delivered and closedbears on Public-private collaboration
South Korea's Economic Planning Board operated from 1961 to 1994 as the government body responsible for economic development planning and budget planning before its functions moved through later finance ministries.
Planning, budgeting and economic coordination were placed in one central institution, giving government a durable route to align public investment with industrial priorities and private-sector expansion.
Preconditions
A centre of government with authority over development priorities and budgets
Administrative capacity to turn plans into financing and implementation
Firms able to respond to coordinated infrastructure, credit and export policy
Limits: The board's mandate and lifespan show a coordination architecture, not the independent contribution of planning to industrial outcomes or the distribution of its gains. A ministry history is an official account and is not a comparative evaluation.
๐บ๐พUruguaysince 2006still operatingbears on Public-private collaboration
Uruguay required producers to register every calf in a national traceability system from 2006, and by October 2011 the registry covered 100 percent of the herd, 11.75 million live cattle, while identification tags reached more than 45,000 producers each year.
11,750,000 cattle, live cattle registered and individually identified, 2011-10. 45,000 producers, producers receiving identification tags each year, 2011. Presidency of Uruguay, SNIG traceability report, retrieved 2026-08-30.
The move, and what it needed
The agriculture ministry made lifetime identification compulsory, issued paired visual and electronic tags, reused the existing producer register and enlisted the postal service to deliver tags to farms, while producers and processors recorded every movement through one national system.
Preconditions
A pre-existing register of livestock holders and annual herd declarations
Producer organisations and processors able to adopt one identification standard
A ministry with the legal authority and field capacity to enforce universal registration
Where else it has been used: Many exporters trace selected animals or supply chains. Uruguay's distinctive move was to make one public system universal, so small and large producers faced the same standard and any animal could enter a demanding export market.
Limits: A complete administrative register does not prove that every movement is reported correctly, that inspectors can reconcile cattle with the database, or that the commercial benefit exceeds the public and producer cost. The source is a government report from 2011, and the collaboration rests on a legal obligation rather than voluntary agreement, so it evidences coordinated national implementation more strongly than mutual public-private agenda-setting.
๐ณ๐ฑNetherlandssince 1950still operatingbears on Public-private collaboration
The Netherlands established the Social and Economic Council (SER) by law in 1950, giving employers, trade unions and independent Crown-appointed experts a standing forum for social and economic policy; the current council has 36 members and collective agreements cover more than 84 percent of employees.
The state gave organised employers and workers a formal route to negotiate national trade-offs, added independent experts as bridge-builders and required the Cabinet to explain when it rejects a broadly supported recommendation, turning social conflict into a repeated bargaining process.
Preconditions
Representative employer and worker organisations able to make commitments for members
A government willing to share agenda-setting without surrendering final authority
Independent technical expertise and a legal framework that gives consultation continuity
Where else it has been used: The consultation model influenced corporatist bargaining institutions elsewhere, but its results depend on membership, bargaining coverage and trust in representation. Copying a council without those organised constituencies produces a meeting, not a coordination mechanism.
Limits: A tripartite membership structure and collective-agreement coverage show the reach of a consultation system, not whether its compromises are equitable, whether advice changes policy or whether workers outside agreements benefit. The SER describes its own role and the coverage figure; it is not an independent evaluation of the polder model's effects.
๐ณ๐ฑNetherlandssince 2023still operatingbears on Public-private collaboration
In November 2023 the Netherlands renewed its mission-driven innovation compact: businesses, knowledge institutions, governments and other organisations committed an estimated โฌ5.7 billion a year, coordinated through five missions and eight Knowledge and Innovation Agendas.
Government gave industry, research organisations and public agencies a shared mission vocabulary, placed them in sector teams and tied public instruments to jointly written agendas so collaboration could shape research priorities rather than appear only at the procurement stage.
Preconditions
Firms and research institutes organised enough to nominate credible representatives and commit resources
Ministries willing to coordinate instruments and missions across departmental boundaries
A funding system that can support collaborative R&D while keeping room for SMEs and exploratory work
Where else it has been used: Mission-oriented innovation policy has been adapted by other countries and the European Union. The Dutch experience suggests that a mission needs a budget pathway and accountable teams; a slogan without those links becomes a strategy document rather than a collaboration mechanism.
Limits: The โฌ5.7 billion is an estimated annual commitment, not audited additional spending, and the partner count measures signatures rather than durable collaboration or successful products. The ministry's announcement does not provide a counterfactual against which the policy's innovation effects can be judged. Funding agendas can coordinate priorities without changing how projects are selected in practice.
๐ณ๐ฑNetherlandssince 2005still operatingbears on Public-private collaboration
The Netherlands built iDEAL as a shared bank-connected online payment rail; in 2024 it processed 1.47 billion payments worth โฌ141 billion, almost matching the โฌ147 billion spent through debit cards at the point of sale.
Banks and the scheme operator gave merchants one recognisable checkout method while allowing each bank to keep its own customer interface; the shared protocol made account-to-account payment immediate and gave merchants a common confirmation path.
Preconditions
High bank-account penetration and a payment system that can settle account-to-account transfers quickly
Interoperable technical and liability standards accepted by competing banks and payment service providers
A trusted scheme brand with enough merchant coverage to create a self-reinforcing network effect
Where else it has been used: Account-to-account rails have travelled widely, but the Dutch advantage came from early bank coordination and dense merchant coverage; copying the interface without that governance and network does not reproduce the result.
Limits: Transaction volume and turnover show adoption of one payment rail, not fraud rates, consumer welfare, merchant costs or whether iDEAL caused the growth of Dutch e-commerce. The Dutch Payments Association aggregates industry data, and the figures do not separate domestic from cross-border use. A successful network also benefits from incumbency and high bank-account ownership.
๐ณ๐ฑNetherlandssince 2009still operatingbears on Public-private collaboration
The Netherlands developed eHerkenning with commercial identity providers so businesses could use one trusted login across public services; in 2024 the framework recorded 1,140,751 authentication means and 639 connected organisations.
Government set the assurance rules and service-provider obligations while private identity providers competed to issue credentials, giving businesses a choice of provider without asking every agency to build its own business identity system.
Preconditions
A government-wide rulebook that makes credentials portable across participating services
Accredited private providers able to meet identity, security and liability requirements
A business registry and authorised-representative model that links a person to the organisation they act for
Where else it has been used: Federated business identity can travel where governments can certify competing providers and firms need many public services. Without enforceable assurance rules, a multi-provider model becomes a collection of incompatible logins rather than a shared trust layer.
Limits: Credential and organisation counts show the reach of a trust framework, not whether small firms can afford a credential, whether logins reduce administrative burden or how identity providers distribute risk. The service statistics are operator-reported and do not reveal failed attempts, fraud or the cost of compliance at each assurance level.
๐ณ๐ฑNetherlandssince 2004still operatingbears on Public-private collaboration
The Netherlands runs Digipoort as a central gateway that receives, validates and routes secure messages between businesses and public agencies; in 2024 it handled 75,433,353 messages, including a channel with 3,075 connected organisations for financial and social reporting.
75,433,353 messages, messages handled by digipoort, 2024. 3,075 organisations, connected organisations in sbr financial & social, 2024. Logius, 2024 service figures, retrieved 2026-08-31.
The move, and what it needed
The state provided one secure gateway and common message standards so software suppliers, accountants and firms could submit reports to several agencies without negotiating a new transport and validation layer for each one.
Preconditions
Shared data definitions and validation rules that agencies and software providers will keep aligned
A secure national gateway with clear service ownership, certificates and routing responsibilities
Large enough reporting volumes and capable intermediaries to justify integration and maintenance costs
Where else it has been used: The SBR and XBRL ideas behind Digipoort are international, but a national gateway depends on agencies agreeing to one taxonomy and on firms receiving a real reduction in duplicate reporting rather than another mandatory interface.
Limits: Message volume and connected-organisation counts show use of a shared gateway, not whether submissions are correct, cheaper or less burdensome for firms. The 2024 total combines different message streams with different users and service levels, and operator-reported availability does not measure the quality of the underlying tax, payroll or regulatory decisions.
๐จ๐ฑChilesince 2003still operatingbears on Public-private collaboration
Chile runs Mercado Pรบblico as a shared e-procurement platform; in 2024 it transacted US$17.643 billion through 2,031,670 purchase orders involving 110,255 participating suppliers.
ChileCompra set common tender rules, a digital marketplace and open transaction records so public agencies could buy from a broad supplier base through one visible channel rather than each agency maintaining a separate procurement system.
Preconditions
A central procurement authority with the mandate to standardise rules across agencies
An online market, supplier registry and transaction data that can be audited by buyers, firms and civil society
Enough public purchasing volume and digital capability for suppliers to justify participating in a shared platform
Where else it has been used: National e-procurement portals are widely copied. The Chilean case shows that a large transaction count is only a platform capability: integrity, SME access and value for money still need separate outcome measures and active oversight.
Limits: The report's totals describe purchase-order commitments rather than payments, and they are operator-reported. They show the reach of a common market, not whether procurement is cheaper, more competitive, more innovative or less corrupt, nor whether small suppliers receive an equitable share. The supplier classification is based on administrative data and does not establish the quality or delivery of the contracts awarded.
๐จ๐ฑChilesince 2010still operatingbears on Public-private collaboration
Since 2010, Start-Up Chile has supported more than 2,500 startups from Chile and abroad through an equity-free public accelerator that connects founders with investors, mentors and corporates.
CORFO supplied equity-free early funding, a landing platform and a founder network while private investors, mentors and corporates supplied market knowledge, follow-on capital and commercial links. The state used a public accelerator to widen the pipeline rather than trying to pick every winning company itself.
Preconditions
A capable public agency able to run competitive selection and portfolio support
A startup pipeline, visa or landing arrangements and enough market access to attract international founders
Private investors, mentors and corporates willing to provide complements after public entry support
Where else it has been used: Public accelerators can travel to places trying to build an innovation ecosystem, but the model needs a credible pipeline and follow-on private capital. A large alumni count without local links or survival evidence can become a branding programme rather than durable capability.
Limits: The programme's portfolio count is self-reported and records support, not startup survival, additionality or the share of value retained in Chile. It does not show whether public funding displaced private investment, whether benefits reached domestic firms or how outcomes compare with other accelerators. The equity-free model and network describe an intervention, not proof of a causal economic return.
๐จ๐ดColombiasince 2015still operatingbears on Public-private collaboration
Since March 2015, Colombia Compra Eficiente has used SECOP II to move public contracting from publication to online transactions; in 2024 the platform managed COP117.642 trillion, 84.4% of the value managed across the three SECOP platforms.
The procurement agency supplied common transaction rules and a shared digital workflow so public buyers and private suppliers could publish, bid, award and manage contracts in one auditable space rather than exchanging documents through disconnected systems.
Preconditions
A central procurement authority able to mandate common fields, procedures and platform use
Reliable digital identity, supplier registration and data standards for buyers and bidders
Public and private users willing to shift from paper or publication-only processes to a transaction system
Where else it has been used: Transaction platforms can travel across procurement systems, but the Colombian lesson is that digitising the workflow is only the shared infrastructure; competition, integrity and supplier inclusion need their own tests.
Limits: The managed-value totals show the reach of a transaction platform, not whether procurement is cheaper, more competitive, more innovative or less corrupt. Colombia Compra Eficiente reports the figures and combines different contract types and platforms; value managed is not the same as payments or delivered outputs. Open contracting data still require supplier capacity, audit and citizen scrutiny to change outcomes.
๐ต๐ชPerusince 2009still operatingbears on Public-private collaboration
Peru's Obras por Impuestos scheme lets private firms finance public projects against future tax liabilities; the finance ministry recorded S/5.504 billion in certificates issued from 2009 through January 2024 and six projects identified as physically and financially complete in 2024.
Peru let companies propose and finance public works and recover the cost through tax certificates, creating a bridge between private capital and local public priorities while the state retained project approval and service ownership.
Preconditions
A tax-credit and certification framework that makes private investment legally recoverable
Public authorities able to identify, appraise and accept projects without turning the scheme into a private wish list
Transparent project registers and technical supervision so a completed work can enter public service
Where else it has been used: Tax-linked infrastructure finance travels to countries with capable revenue authorities and uneven local capital, but Peru's experience shows the need to track the public service and lifecycle cost, not only certificates issued.
Limits: Certificates and completion counts show the scale of a tax-linked financing mechanism, not whether projects were additional, delivered on time, well maintained or distributed fairly. The ministry compiles project information from public systems and companies; the scheme shifts financing and execution roles but does not remove public responsibility for selecting, regulating and operating the asset.
๐จ๐ทCosta Ricasince 1997still operatingbears on Public-private collaboration
Costa Rica's FONAFIFO Payment for Environmental Services programme paid contracts covering 208,089 hectares in 2024, disbursing โก9.234 billion while using public, donor and market-linked finance to keep private land in forest uses.
208,089 hectares, hectares under payment for environmental services contracts, 2024. 9,234,000,000 Costa Rican colones, payments disbursed through the programme, 2024. FONAFIFO, Board meeting record and 2024 programme accounts, retrieved 2026-08-31.
The move, and what it needed
Costa Rica created a dedicated forest-finance institution that collects earmarked public revenue and combines it with donor and market funds, then pays landowners under monitored contracts for water, carbon and biodiversity services.
Preconditions
A legal mandate and predictable revenue stream that can fund multi-year land contracts
Land records, spatial targeting and field monitoring able to verify who is paid for what
A public institution trusted by landowners and external funders to administer contracts and report results
Where else it has been used: Payments for ecosystem services travel well as a way to make stewardship financially visible, but Costa Rica's case shows the hard part: stable public finance and credible monitoring must survive changes in carbon markets and land prices.
Limits: Contract hectares and payments show the reach of a conservation-finance mechanism, not whether forest cover is additional, permanent or evenly distributed. The board record is an administrative account from the programme operator; it does not independently verify carbon permanence, biodiversity outcomes or the counterfactual land use.
๐ช๐จEcuadorsince 2012still operatingbears on Public-private collaboration
Ecuador's tax authority received approximately 3.299 billion electronic receipts from January through December 2024, using a legally mandated, digitally signed rail that connects private taxpayers to the public tax system.
Ecuador made electronic invoicing a common legal format, supplied a public authorization and validation service, and required private firms to send signed transaction data to the SRI so tax administration and business records shared one machine-readable event.
Preconditions
A tax authority with a unique taxpayer register and power to make the format mandatory
Digital certificates and a validation service that can process high-volume submissions
Free or accessible tools for small firms so the compliance mandate does not become an exclusion mechanism
Where else it has been used: Electronic invoicing travels as a public-private data standard. Ecuador shows the scale benefit of mandatory interoperability, alongside the need to watch who bears the implementation cost and whether the data improves decisions rather than just accumulating.
Limits: Receipt volume measures use of the tax rail, not the accuracy of declarations, additional revenue, compliance costs or service quality. The SRI report notes that totals can change through later cancellations and transmissions, and adoption is partly enforced by regulation rather than voluntary collaboration.
๐ช๐จEcuadorsince 2007still operatingbears on Public-private collaboration
Ecuador's Ministry of Economic and Social Inclusion reported delivering 44,622 Crรฉdito de Desarrollo Humano loans in 2024, including 6,622 to Indigenous people, turning a social-transfer registry into small-enterprise finance and market-access support.
Ecuador let eligible transfer recipients convert part of a social benefit into a productive credit, then paired public finance with group or individual ventures and a government market-access service rather than leaving a cash transfer disconnected from enterprise support.
Preconditions
A reliable social registry that can identify eligible recipients and protect against duplicate claims
Payment and credit channels capable of reaching people outside formal banking
Business-development and market-access support that can follow a loan beyond the disbursement event
Where else it has been used: Transfer-to-enterprise schemes travel where welfare systems already have reach, but Ecuador's experience points to the missing complement: finance needs coaching, markets and repayment feedback if it is to build agency rather than just advance a benefit.
Limits: Credit counts show administrative disbursement, not repayment, business survival, income gains or whether the projects are productive. The ministry's report combines several beneficiary groups and credit modalities, so the total is not a comparable measure of private investment or a causal poverty effect.
๐ต๐พParaguaysince 2017still operatingbears on Public-private collaboration
Paraguay's public-works balance records the Ruta PY02 duplication and expansion as a 149.5-kilometre public-private partnership, awarded under the 2013 APP law with a USD 507 million construction contract and a concession running to 2046.
Paraguay used an APP concession to bundle design, financing, construction, maintenance and operation of a national corridor, moving some delivery and lifecycle obligations to a private special-purpose company while the ministry retained contract oversight.
Preconditions
A procurement and contract-management authority able to specify service levels over a multi-decade concession
Traffic, toll and payment data strong enough to allocate demand and revenue risk transparently
Public capability to monitor construction quality, renegotiate only under clear rules and enforce maintenance after opening
Where else it has been used: Public-private partnerships travel only when the state can remain a demanding contract manager. Ruta PY02 is useful evidence of scale and institutional ambition, but its long concession also makes monitoring and renegotiation part of the capability being tested.
Limits: Contract length and value show the scale and financing model, not whether the road met its promised schedule, lifecycle cost, safety standard or service-level payments. A government project balance can confirm the contract and progress but is not an independent value-for-money review, and construction and operation continue over the concession term.
๐ฆ๐ทArgentinasince 2012still operatingbears on Public-private collaboration
Argentina's energy ministry reported that Vaca Muerta supplied 54.9% of national oil output and 50.1% of national gas output in December 2024, within a year in which national oil production reached its highest level since 2003 and gas its highest since 2006.
Argentina combined state control of the resource and export rules with private drilling, midstream investment and a price framework intended to make unconventional wells financeable, turning a large geological resource into a growing national production base.
Preconditions
A regulatory and export regime stable enough for firms to commit to long-cycle drilling and pipelines
Specialist suppliers, skilled labour and transport infrastructure for high-volume unconventional operations
Public institutions able to collect production data, manage environmental trade-offs and negotiate infrastructure access
Where else it has been used: Resource booms travel only when public rules and private execution reinforce each other. Vaca Muerta is useful evidence of that interface, while its externalities and policy dependence keep the collaboration claim bounded.
Limits: Production shares show geological and industrial output, not the net fiscal return, environmental effects, local infrastructure strain or the durability of investment under changing prices and regulation. The government account attributes the rise to regulation and private investment, but this record does not isolate either contribution or establish the field's full lifecycle economics.
๐ฎ๐ฉIndonesiasince 2015still operatingbears on Public-private collaboration
Indonesia delivered the three-package Palapa Ring fibre network to connect 57 kabupaten/kota; the communications ministry reported average utilisation of 54.35% in 2024, with the western, central and eastern packages carrying different loads.
The Indonesian state split a national backbone into regional packages and used operating partners to extend public connectivity into places where commercial demand alone was too thin, then tracked capacity and utilisation by package.
Preconditions
A public agency able to aggregate demand across remote districts
Long-term contracts that make operation viable after construction
A backbone design with common technical standards and measurable service levels
Where else it has been used: Public-private infrastructure works best when the public side specifies the underserved geography and measures use after the ribbon-cutting. Palapa Ring shows why a completed fibre route can still have a demand problem.
Limits: Network utilisation and the number of connected districts show a delivered platform and the demand placed on it, not the procurement cost, construction schedule, service reliability or whether local users can afford the resulting bandwidth. The utilisation figure is a ministry planning statistic, and its average hides large differences between the western, central and eastern packages.
๐ฒ๐พMalaysiasince 2020still operatingbears on Public-private collaboration
Malaysia's JENDELA Phase 1 connected 7.74 million premises with fibre by 2022, above its 7.5 million target, and raised 4G coverage in populated areas to 96.92% through a regulator-led programme delivered with telecommunications providers.
7,740,000 premises, premises passed by jendela fibre, 2022. 96.92 % of populated areas, 4g coverage in populated areas after jendela phase 1, 2022. JENDELA, Phase 1 concluding report, retrieved 2026-08-31.
The move, and what it needed
Malaysia's communications regulator set common coverage and fibre targets, coordinated multiple providers and retired 3G spectrum to concentrate investment on a shared 4G foundation before expanding 5G.
Preconditions
A regulator able to set measurable targets for competing network operators
Shared access to fibre backhaul, towers and spectrum during the transition
A public investment and reporting framework that makes rural gaps visible
Where else it has been used: Digital infrastructure partnerships travel when the state measures service geography and providers can share the underlying network. JENDELA's target-versus-delivery record is more informative than a generic broadband-penetration claim.
Limits: Premises passed and populated-area coverage measure network availability, not affordability, uptime, actual household subscriptions or rural service quality. The programme report combines regulator and industry-reported delivery and its coverage denominator excludes unpopulated areas, so it is not a national land-area measure.
๐น๐ญThailandsince 2016still operatingbears on Public-private collaboration
Thailand's central bank and commercial-bank network have run PromptPay as a shared instant-payment rail since 2016; by December 2024, 87.5 million proxy numbers were registered and average daily volume reached 66.4 million transactions, with a one-day peak of 89.2 million in September.
87,500,000 registered numbers, promptpay registered proxy numbers, 2024-12. 66,400,000 transactions per day, average daily promptpay transactions, 2024. Bank of Thailand, Annual Report 2024, retrieved 2026-08-31.
The move, and what it needed
The Bank of Thailand set a common proxy-ID and QR standard, while banks connected their existing accounts and customer channels to a central registration and switching layer. That let a small merchant receive an interbank payment with a phone number or citizen ID instead of negotiating a bespoke integration with every bank.
Preconditions
A trusted central bank able to set an interoperable standard across competing banks
A shared identity or account-linking layer that resolves proxy numbers safely
Low-cost mobile and QR channels available to households, firms and government agencies
Where else it has been used: Fast-payment rails have travelled through bilateral links and domestic copies. Thailand's lesson is that a public standard becomes infrastructure only when private banks can reuse it at scale and the operator publishes usage and security performance.
Limits: Registration and transaction volume show a heavily used payment rail, not the distribution of access, reliability during outages, fraud losses or the share of activity that is genuinely additional to cash and cards. The Bank of Thailand reports the aggregate totals, and proxy registrations can include more than one number for a person or business. The public-private mechanism is inferred from the central-bank standard and participating-bank network rather than isolated by the report as a causal estimate.
๐ช๐ธSpainsince 2021still operatingbears on Public-private collaboration
Spain's Red.es Kit Digital programme used public grants, volunteer representatives and private digitalising agents to help small firms adopt software and online services; by 3 August 2026 the programme reported 1,566,846 transformation projects delivered by 10,727 agents, 96 percent of them ICT SMEs.
1,566,846 projects, digital transformation projects launched through kit digital, 2026-08-03. 10,727 agents, adhered digitalising agents, 2026-08-03. Red.es, Kit Digital public-private collaboration results, retrieved 2026-08-31.
The move, and what it needed
Red.es standardised eligible solutions and payment rules, while private agents supplied implementation and volunteer representatives handled applications for firms that lacked administrative capacity. The public side made demand legible and portable; the private side supplied local skills and delivery capacity.
Preconditions
A public grant and verification platform that can process many small awards
A pre-qualified supplier network with enough geographic reach to serve rural firms
Simple application and justification rules plus intermediaries who can assist digitally excluded owners
Where else it has been used: Voucher programmes travel when the state defines an interoperable menu and pays on verified delivery. Spain's agent network shows the reach gained by outsourcing implementation, while its project count keeps the risk of double-counting visible.
Limits: Projects and agents measure programme throughput, not whether the adopted tools raise productivity, remain in use or narrow the digital divide. The results page is a programme dashboard updated in 2026 and does not provide a counterfactual for firms that did not receive a grant. Projects can be delivered to the same beneficiary in multiple agreements, and agent counts include providers rather than unique firms served.
๐ต๐ฑPolandsince 2018still operatingbears on Public-private collaboration
Poland's Gaz-System, Denmark's Energinet and Norway's Gassco coordinated the Baltic Pipe cross-border gas corridor; the interconnection launched on 1 October 2022 and reached its target technical parameters by November, allowing up to 10 billion cubic metres per year toward Poland.
Three transmission-system operators aligned standards, permits, offshore construction and compressor upgrades around one operating date. Poland's public strategy created the demand and political priority, while the network companies supplied the engineering and cross-border interfaces needed to make the corridor usable.
Preconditions
Regulators and transmission operators with authority to coordinate across borders
A shared technical and permitting plan for offshore and onshore assets
An anchor demand and financing model strong enough to justify capacity before completion
Where else it has been used: Cross-border infrastructure travels through interoperable operators and a single commissioning plan. Baltic Pipe shows how a strategic objective can be converted into a working corridor when national and commercial incentives line up.
Limits: Capacity and commissioning dates show infrastructure delivered, not actual annual flows, gas-price effects, environmental costs or the security value of a single corridor under changing demand. The regulator reports technical parameters and the three-operator arrangement, while the project also depended on EU and national permitting outside this metric. The record describes a cross-border build rather than a purely Polish project.
๐ธ๐ชSwedensince 2012still operatingbears on Public-private collaboration
Sweden's six-bank-owned Getswish consortium built Swish into a national mobile payment rail: in 2023 it served just over 8.7 million private users and about 335,000 merchant or business users, processing approximately 977 million payments worth SEK 480 billion.
Competing banks jointly owned a simple brand and API, used BankID for authentication and connected the service to the national settlement infrastructure. The consortium solved the coordination problem of getting rival banks to make one person-to-person rail interoperable before merchants and public services built on it.
Preconditions
A small group of banks willing to share ownership and a common customer experience
A trusted digital identity already used by most account holders
Settlement infrastructure and rules that can move funds between member banks in seconds
Where else it has been used: Swish travels as a bank-led interoperability pattern. The institutional move is joint governance across rivals; the app's reach depends on the identity, handset and settlement layers underneath it.
Limits: The Riksbank's figures are service totals, not a causal evaluation of the consortium or a measure of consumer welfare. Swish depends on mobile BankID, smartphones and participating banks, so people without those prerequisites can be excluded; high payment volume also says nothing about fraud, outage resilience or market competition.
๐ฎ๐ชIrelandsince 2019still operatingbears on Public-private collaboration
Ireland's National Broadband Plan combined commercial investment with a state intervention contract for rural areas covering more than 560,000 premises; by October 2024 the government reported over 102,600 premises connected to the National Broadband Ireland network.
The state defined a universal-service gap, subsidised a network builder in that gap and left commercially viable areas to private operators. A single contract, a published intervention map and connection points for community facilities let public policy steer private construction without requiring the state to own every fibre asset.
Preconditions
A reliable map distinguishing commercial coverage from premises no operator will serve
A procurement and subsidy contract with milestones, open access and clawback protections
An independent regulator able to monitor build quality, wholesale access and consumer take-up
Where else it has been used: The NBP travels as a gap-subsidy model: public money can extend private networks, but the benchmark must separate premises passed, connected and using the service so a rollout headline does not become a coverage claim.
Limits: The connection count is a rollout snapshot, not completion of the national plan or a measure of connection quality, take-up, affordability or economic impact. The state-intervention area is a target geography and the programme's future milestones can move with construction, demand and commercial coverage; premises passed and premises connected are different stages.
The UAE Telecommunications and Digital Government Regulatory Authority reported that UAE Pass had 6.84 million registered users accessing more than 15,000 digital services from government and private-sector providers in 2023.
TDRA coordinated federal and emirate-level governments around one identity, signature and document layer, then allowed banks, telecoms and other private providers to connect to it. A shared authentication rail lets separate organisations reuse verified identity instead of each rebuilding onboarding.
Preconditions
A trusted identity and signature framework with privacy, security and liability rules
Technical standards and an integration team able to connect heterogeneous public and private service providers
A large enough service portfolio that users and providers have reason to adopt the common identity
Where else it has been used: Digital identity travels as an ecosystem bargain: a single login scales only when public and private providers share standards and users can see a practical service benefit.
Limits: Registered accounts and listed services do not show active use, completion rates, accessibility, privacy experience or whether a transaction became faster for residents. The figures are reported by the authority coordinating the platform, and provider totals combine government, semi-government and private services with different quality and accountability arrangements.
๐ฌ๐นGuatemalasince 2003still operatingbears on Public-private collaboration
Guatemala's Ministry of Finance reported 152,676 enabled Guatecompras users in April 2023, including 137,950 registered suppliers alongside public buyers, auditors and oversight bodies using the national procurement system.
Guatemala put public buyers, suppliers, auditors and oversight bodies on a shared procurement record, giving firms a common route into state demand and giving monitors a common transaction identifier. The platform turns a dispersed paper process into a searchable public-private interface.
Preconditions
A procurement rule that requires public entities to publish opportunities and awards in one system
Supplier registration and buyer roles that let firms participate while preserving audit access
A public data interface and complaint or correction process that makes transactions inspectable after award
Where else it has been used: Procurement platforms travel when participation and oversight use the same records: a portal that only publishes winning announcements cannot create the supplier competition or audit trail that Guatecompras attempts.
Limits: Enabled accounts are not the same as completed tenders, competitive prices, delivered works or lower corruption. The ministry's count is an administrative snapshot and does not show how often suppliers bid, whether data are corrected, or whether every public entity follows the system in practice.
๐ฌ๐งUnited Kingdomsince 1975still operatingbears on University-industry collaboration
UKRI reported more than 1,000 Knowledge Transfer Partnerships running at one time and more than 1,100 Associate projects in May 2026, joining businesses with universities or other research organisations through jointly supervised innovation work.
Innovate UK funds a three-way arrangement in which a business names a strategic problem, a university or research organisation supplies expertise and a graduate associate works inside the firm under joint supervision. The structure gives firms absorptive capacity while giving academic knowledge a route into operations.
Preconditions
Research organisations with staff who can translate academic methods into business projects
Firms willing to co-fund a 12โ36 month project and give the associate a real strategic problem
A national intermediary that matches partners, sets quality rules and carries lessons between projects
Where else it has been used: Knowledge-transfer partnerships travel when the intermediary funds the bridge and the firm owns a concrete problem; generic networking events rarely create the same embedded learning or accountability.
Limits: The partnership and project counts show programme scale, not whether collaborations produce durable productivity, patents, spinouts or jobs. UKRI reports the programme's administrative totals and benefit estimates; participation is not a representative sample of all UK firms or universities, and the figures do not provide a counterfactual against which the programme's effects can be judged.
๐ฆ๐บAustraliasince 1991still operatingbears on University-industry collaboration
Australia's Department of Industry reported that its Cooperative Research Centres portfolio facilitated 304 businesses establishing research-focused collaborations in 2023โ24, across 114 CRC and 190 CRC Projects partners, above the year's target of 248.
Australia funds industry-led research centres for three to ten years and shorter CRC Projects for up to three years, giving firms and researchers a shared programme, matching resources and a governance structure for applied work. The long and short formats let a mature collaboration and a focused translation project coexist.
Preconditions
Universities, research organisations and firms able to define a problem together and commit staff or cash
A competitive public grant process with independent assessment and enough duration for trust and technical work to accumulate
A centre-level governance and intellectual-property framework that can turn joint research into adoption or further investment
Where else it has been used: Universityโindustry centres travel when public grants buy durable shared capacity and firms own a consequential problem; short networking schemes without governance, matching effort or a path to adoption produce thinner ties.
Limits: The partner count measures collaborations facilitated through grant agreements, not completed research, commercialisation, additionality or university-only links. It is an administrative portfolio measure whose reporting and retrospective calculations can change, and partners include research organisations and community bodies alongside universities and businesses.
๐ต๐นPortugalsince 2017still operatingbears on University-industry collaboration
Portugal's Agรชncia Nacional de Inovaรงรฃo reported 41 recognized Collaborative Laboratories (CoLABs), and its 2024 innovation report recorded 639 highly qualified staff contracted by CoLABs, 32 percent of them PhDs.
Portugal gives mission-led applied-research organisations a formal recognition and monitoring route, then combines public base or competitive funding with industry-facing agendas so researchers have an institution whose job is to carry work across the university-business boundary.
Preconditions
Universities and firms willing to share researchers, facilities and intellectual-property arrangements
A public agency able to recognize, monitor and renew organisations against research and transfer agendas
Stable mixed funding so an applied laboratory can pursue long projects without becoming a consultancy only
Where else it has been used: The CoLAB form travels when recognition is tied to an accountable applied mission and real employer or industry demand; a list of centres without shared incentives does not create technology transfer.
Limits: The network and staffing totals show institutional reach, not research quality, business adoption, additionality or durable jobs. CoLABs are separate private non-profit entities and report through the programme; the figures do not cover every university-industry link or establish that public recognition caused the reported staffing.
๐ต๐นPortugalsince 2021still operatingbears on Public-private collaboration
Portugal's PRR transparency portal reports 52 Agendas Mobilizadoras and Verdes representing โฌ5,278,338,665.70 of approved investment and โฌ3,051,632,468.10 of approved support as of 17 August 2026.
Portugal used an open call to assemble firms, research organisations, clusters and public bodies into contract-based consortia, then tied public support to innovation, production, skills and technology-transfer work inside each agenda.
Preconditions
A central agency able to run a competitive call and monitor many heterogeneous consortia
Firms and research organisations with enough technical and financial capacity to share a multi-year plan
Milestones, procurement and payment controls that keep public support conditional on joint delivery
Where else it has been used: Consortium programmes travel when public money buys a shared delivery obligation rather than a collection of grants; a call without a lead firm, research partner and milestone gate rarely creates durable collaboration.
Limits: Approved investment and support are programme commitments, not money spent, products delivered, export growth or additional private R&D. The portal aggregates consortia with different sectors and execution rates, and the figures are reported by the programme itself; they show a coordination mechanism at national scale, not that every alliance will complete its milestones by the 2026 deadline.
๐ฎ๐ณIndiasince 2016still operatingbears on Public-private collaboration
NPCI's UPI rail connected 731 live banks and settled 22,716.07 million transactions worth โน28,92,138.67 crore in June 2026 through participating banks and payment apps.
The Reserve Bank-regulated NPCI set a common protocol and settlement layer while banks and private payment apps competed at the customer interface, allowing one public standard to scale through many firms.
Preconditions
A regulated national payments operator trusted to set interoperable rules
Banks connected to a common account and settlement infrastructure
A legal and technical framework that lets private apps join without fragmenting the rail
Where else it has been used: Interoperable rails travel when the public layer is open enough for private distribution and strict enough to preserve settlement and consumer protection; transaction volume alone cannot show whether the partnership is inclusive.
Limits: UPI volume and bank participation show a shared transaction rail, not the quality of public-private coordination across other national objectives. NPCI publishes the figures for the system it operates, the table excludes on-us transactions from August 2018 onward, and volume says nothing about access, fraud, outages or distributional reach.
๐ฟ๐ฆSouth Africasince 2011still operatingbears on Public-private collaboration
South Africa's IPP Office reported 95 independent power producers operational with 7,335 MW online by 31 March 2025, alongside 9,618 MW procured from 116 projects at financial close and R292.2 billion attracted for energy infrastructure.
The state set capacity and development rules, ran repeated competitive tenders and gave private producers bankable power-purchase agreements, while Eskom and government agencies carried the grid and policy interfaces.
Preconditions
A creditworthy public offtaker and a grid connection process private bidders can rely on
Transparent bid windows with standard contracts and independent evaluation
A public procurement office able to monitor construction, commercial operation and economic-development obligations
Where else it has been used: Publicโprivate energy programmes travel when the state makes risk allocation and payment rules credible; selected capacity without financial close or operation is still only a pipeline.
Limits: The IPP Office reports the programme it manages, and its figures combine several procurement windows and technologies. Operational MW and financial close show a functioning publicโprivate delivery channel, not affordability, reliability, local ownership or emissions outcomes; terminated projects are excluded from current totals and the investment figure is attracted capital rather than completed public value.
๐ฒ๐ฝMexicosince 2019still operatingbears on Public-private collaboration
Mexico's labour ministry linked young people to more than half a million registered workplaces for on-the-job training, reporting 2.95 million beneficiaries nationwide after five years of Jรณvenes Construyendo el Futuro in February 2024.
The labour ministry paid a stipend and social insurance while firms, shops and civil-society organisations supplied real work placements, with a shared digital platform matching young people to host workplaces.
Preconditions
A public payment and insurance system able to support people outside formal employment
A large enough network of employers willing to host and report trainees
A matching and inspection platform that gives the state visibility over placements
Where else it has been used: Workplace guarantees travel when the state underwrites the first months and employers provide the learning environment; the transferable component is the matching and verification loop, not the stipend size.
Limits: The ministry's total is a cumulative programme count and the workplace figure is rounded; neither shows unique completions, the quality of mentoring or whether trainees stayed employed. The government funds the stipend and insurance, so the record demonstrates a large public-private delivery interface rather than private financing of the programme.
๐น๐ทTurkeysince 2008still operatingbears on Public-private collaboration
Tรผrkiye's e-Devlet Gateway ended 2024 with 66,753,526 users, 8,309 online services and 1,080 participating institutions, including 127 private providers alongside central government, universities and municipalities.
A central gateway gave public agencies, universities, municipalities and selected private providers one identity and service front door, while the state kept the common authentication and catalogue layer.
Preconditions
A national identity and authentication system that many institutions can trust
A centre-of-government operator able to set integration standards
Institutions with digital case systems mature enough to expose a service end to end
Where else it has been used: Shared portals travel when the common layer is more valuable than each institution's bespoke login; a catalogue of links without transaction completion only moves the queue online.
Limits: User registrations and service listings measure reach and integration, not whether transactions complete, how long they take or whether private providers add value. The gateway's own statistics aggregate institutions with very different roles, and the 127 private providers are included as participants rather than evidence of joint financing or shared accountability.
๐น๐ทTurkeysince 2024still operatingbears on University-industry collaboration
TรBฤฐTAK's 1505 UniversityโIndustry Collaboration Support Programme received 116 project applications and began support for 59 projects in 2024, with 52 fellows working across the collaboration pipeline.
TรBฤฐTAK made firms' needs the entry point for university research, with a public grant and named academic and customer organisations sharing the project rather than leaving technology transfer to informal contacts.
Preconditions
A public research fund able to contract with both universities and firms
Academic teams with usable research and firms willing to specify a problem
A review and reporting process that keeps the two sides accountable to one project plan
Where else it has been used: Universityโindustry programmes travel when the firm is a real customer and the grant funds a jointly owned deliverable; a networking event without a project contract is not collaboration capacity.
Limits: Applications and support starts measure the interface, not the number of products commercialised or the durability of universityโfirm relationships. The activity report is an administrative source and does not identify the firms, project quality or completion rate; 52 fellows are a staffing input, not an outcome.
๐ฎ๐ฑIsraelsince 2007still operatingbears on Public-private collaboration
Israel's Water Authority contracted private operators to build and run large seawater desalination plants under long-term concessions, and by 2022 the five largest held concessions for about 596 million cubic metres a year and supplied 33 percent of the country's fresh water, with the state buying 540 million cubic metres from them that year for about NIS 1.5 billion.
The state kept the planning, the water purchase and the regulation, and let private consortia finance, build and operate each plant on a long-term concession with a guaranteed buyer, so capacity arrived on a tender schedule without the capital sitting on the budget.
Preconditions
A single national water authority and a national carrier able to take every plant's output
A coastline next to the demand and enough grid power for reverse osmosis
Water scarcity severe and lasting enough to sustain the political case across governments
Where else it has been used: Gulf states and Australia built similar plants; where demand fell after a drought ended, as in several Australian cities, plants were mothballed and the fixed concession payments became the controversy.
Limits: The source is an audit, and most of it is criticism: it finds that the Water Authority never completed the national desalination master plan a 2010 government resolution required, that the agreed penalties for poor water quality are too low to enforce anything, that 14 seawater contamination incidents halted production between 2007 and 2023, that three of the largest plants sit near one river mouth, and that no serious preparation had begun for the Palmachim concession expiring in 2029. A third of fresh water is the 2022 national share; desalination's share of drinking water is higher, but the audit does not state it. The started year is the first year the audit reports plants in operation; the first plant opened earlier. Capacity is a contractual figure, not output.
๐ท๐ผRwandasince 2016still operatingbears on Public-private collaboration
From 2016 the Government of Rwanda contracted Zipline to fly blood and later other medical products to health facilities, and by 2024 the Rwanda Biomedical Centre reported deliveries to 624 sites across 27 districts and over 12,000 packages in a single month.
The Ministry of Health and the Rwanda Biomedical Centre bought a delivery service rather than drones: Zipline built and flew the network, the state paid per delivery, centralised the blood supply, and the civil aviation authority opened the airspace.
Preconditions
A centralised national blood service able to stock regional hubs
An aviation regulator willing to authorise beyond-visual-line-of-sight flights at national scale
Health facilities with a reliable ordering channel and a government that pays per delivery
Where else it has been used: Zipline took the contract model to Ghana, Nigeria, Kenya and Cรดte d'Ivoire. Its durability has depended on each health ministry's ability to keep paying fees once donor co-funding ends.
Limits: The source is a slide deck by the public partner, presented at an aviation conference. It is not a statistical release, and its operating figures almost certainly come from Zipline, the private contractor, so the number is the two partners reporting on themselves. 'Today' is undated beyond the 2024 symposium, and 'over 12,000 packages in a single month' is a lower bound for an unnamed month, not an annual series. The site count mixes health facilities with farms and other drop points, and 27 districts is not all 30. The deck's outcome claims (90 percent less wastage, 99 percent less maternal mortality from blood shortage) are unreferenced on the slide and are not used here. The record shows scale, not value for money: contract prices, per-delivery cost to the health budget, and whether road logistics would have done as well are not shown. The collaboration is with one foreign firm under a service contract, which is a narrow form of the joint public-private delivery the indicator means.
๐ญ๐ณHondurassince 2016dismantledbears on Public-private collaboration
Honduras's state utility ENEE awarded Empresa Energรญa Honduras a public-private partnership contract in December 2015, with service starting in 2016, to run national distribution and cut technical and non-technical losses by 17%. Distribution losses fell from 33.38% to 27.11% by February 2018, rose to 35.37% by December 2020 and stood at 30.77% in December 2021. The state intervened the contract and took distribution back when it ended in August 2023.
ENEE, through a Congress-approved PPP trust supervised by SAPP and an external supervisor, handed national distribution and loss recovery to a private operator paid from distribution revenue. When losses rose, the state intervened the contract and then let it lapse, moving loss recovery into its own Programa Nacional para la Reducciรณn de Pรฉrdidas.
Preconditions
A loss baseline and metering both parties accept, so the operator can be paid for points recovered rather than for presence
A regulator and supervisor able to enforce loss targets during the term, not only at expiry
Political tolerance for disconnecting non-paying users and prosecuting theft, which is where non-technical losses are recovered
Where else it has been used: Loss-reduction management contracts are a common utility reform. This case shows the first two years of gains disappearing once enforcement weakened, and a government that ended up reabsorbing the function it had contracted out.
Limits: The loss series is ENEE's own commercial accounting, republished by the Energy Secretariat under the government that intervened the contract and campaigned against it, so the source is not neutral. The table stops in December 2021. The 2022 and 2023 values, including the roughly 37 to 38 percent widely reported when the contract ended, appear only in press coverage, and this record does not carry them. The SEN text does not attribute the 2016 to 2018 fall to EEH by name. The 2020 rise coincides with COVID-19 lockdowns and hurricanes Eta and Iota, so not all of the reversal can be laid on the operator. The contract was intervened and then expired at the end of its term in August 2023. It was not cancelled outright, and 'dismantled' here means the partnership model ended and ENEE took distribution back, not that a court annulled it. The SAPP fact sheet has copy errors: a malformed start date, and a repayment source copied from an airport sheet ('COBROS POR SERVICIOS AEROPORTUARIOS'). The 17% target comes from SAPP's project description, not from the contract text. A loss percentage measures the outcome of the partnership and says nothing directly about how deep the collaboration was. It is evidence that a structured government-firm arrangement on a national goal was tried at national scale and failed, not a measure of collaboration.
๐ธ๐ปEl Salvadorsince 1993dismantledbears on Public-private collaboration
El Salvador ran its national vocational training system from 1993 through INSAFORP, an institute funded by a 1% employer payroll levy and governed by a tripartite council of four government, four employer and four worker representatives; a 2021 reform gave the President the choice of the employer seats, and Decree 893 of 14 November 2023 dissolved INSAFORP and handed its functions to INCAF, whose council holds only the institute's president and four ministries.
El Salvador's Legislative Assembly, on the executive's initiative, first moved the choice of INSAFORP's employer representatives to the President (2021) and then dissolved the tripartite institute, replacing it with a ministry-run council while keeping the employer payroll levy (2023). The state kept the money flow from firms and removed the firms' and workers' seats at the table.
Preconditions
A training system financed by an employer payroll levy, so the state can keep the revenue while changing who governs it
A legislative majority aligned with the executive, able to pass an institutional replacement in a single session
Employer and worker organisations whose seats rested on ordinary statute rather than on constitutional or treaty protection
Where else it has been used: Levy-funded tripartite training bodies of the SENAI and INFOTEP family exist across Latin America; the Salvadoran case shows the arrangement is only as durable as the statute that seats the private sector, since the levy alone does not guarantee a voice.
Limits: The metric is a count of seats read off the text of two laws, not a measured outcome: the value 0 is our reading of Art. 11 of Decree 893, which lists only the INCAF president and four ministries, and the peak of four employer seats comes from a recital in the 2021 reform, not from the 1993 law itself, which we did not open. The 1% levy rate in the claim is read from Art. 19 of Decree 893, which carries the levy over, and was not read from the 1993 text. Neither text shows how much the employer and worker seats shaped training decisions in practice, how many people INSAFORP trained, or whether training volume or quality fell under INCAF; no official participant series was retrieved, so this record documents the end of a governance arrangement, not a fall in delivery. Firms still fund the system: Art. 19 keeps the 1% employer payroll levy, although it reserves only 20% of it for INCAF's training and sends the remaining 80% elsewhere, so the levy-funded collaboration was reshaped rather than abolished in money terms. The collaboration was already eroded before dissolution: the 2021 reform let the President pick employer representatives from ternas built by the Labour Ministry, without requiring membership of employer associations. The Supreme Court documentation host returned intermittent HTTP 522 errors on the day of retrieval before serving both documents.
The Federal Communications Commission ran the Affordable Connectivity Program from 31 December 2021, paying monthly broadband discounts through voluntarily participating internet providers to more than 23 million enrolled households, and ended it on 1 June 2024 when the funding Congress had provided ran out.
Congress funded a fixed pot, the FCC turned it into a monthly discount that private internet providers applied on their own bills, and enrolment climbed until the pot was exhausted; when Congress did not refill it, the FCC froze enrolment, cut the benefit to a partial discount and closed the program.
Preconditions
A national regulator able to certify eligibility against existing benefit rolls such as Lifeline and SNAP
Thousands of private providers willing to apply a federal discount on their own billing systems
A one-time appropriation with no automatic renewal, which made the program's life a matter of each budget fight
Where else it has been used: The design reused the Lifeline telephone subsidy's provider-reimbursement model at much larger scale; its end shows how a program financed by one appropriation, rather than by a standing fund, can be stopped by inaction.
Limits: The 23 million figure is the regulator's own enrolment count, taken at the February 2024 freeze, and it reports sign-ups, not households that would otherwise have gone without broadband: the FCC's claim that 68 percent of ACP households had inconsistent or no connectivity before rests on a self-reported agency survey. The program was a voucher paid through private providers whose participation was voluntary, so a hostile reviewer would call it a subsidy channel rather than joint delivery; the public-private element is that providers carried enrolment, billing and the discount. It grew out of the pandemic-era Emergency Broadband Benefit, so the 2021 start understates how long the capability was built. It ended through a funding lapse rather than a repeal, and the FCC said it could resume if Congress appropriated money, so 'dismantled' records the end of delivery, not the destruction of the administrative machinery. The release is an agency press statement, not an audited statistical series, and the PDF is the script-readable copy; the fcc.gov program pages block automated requests.
๐บ๐ธUnited Statessince 2020delivered and closedbears on Public-private collaboration
Operation Warp Speed, an HHS and DOD partnership, financed private vaccine makers' development, at-risk manufacturing and advance purchases, including a $5.97 billion contract for 300 million Pfizer/BioNTech doses. Three of the vaccines it bought held FDA emergency use authorization by 1 March 2021, the first from 11 December 2020.
HHS and DOD pooled BARDA's countermeasure contracting with defense logistics and paid private manufacturers to build production capacity while trials were still running, carrying the financial risk of failure so that doses existed the day authorization came.
Preconditions
A standing countermeasure agency (BARDA) with contracting authority and existing industry relationships
Domestic firms with mRNA and vaccine platforms already in development
Emergency appropriations large enough to fund parallel bets, some of which failed (Merck/IAVI)
Where else it has been used: Advance market commitments of this kind were proposed for global vaccine supply, but COVAX lacked the purchasing budget and manufacturing ties that made the US version fast.
Limits: The metric counts authorized products, not doses delivered or people vaccinated. It does not show what share of the speed came from federal money rather than private R&D already under way: Pfizer/BioNTech took no federal development funding and joined only through a purchase contract, as the CRS notes. The contract values are awards, not outlays, and CRS cites BARDA and DOD pages without auditing them; the GAO's audit of OWS sits on gao.gov, which blocks automated requests, so it is not the source here. The CRS Insight is a congressional research summary rather than a statistical series, and it does not state the May 2020 launch date, so 'started' is the year vaccine delivery began. Ancillary-supply shortages and distribution problems are flagged in the same document. The Biden administration said it planned to restructure and rename OWS, so 'concluded' marks the end of the vehicle, not of federal vaccine purchasing.
๐ซ๐ทFrancesince 2013still operatingbears on Public-private collaboration
Under the Plan France Trรจs Haut Dรฉbit, defined by the government in February 2013, private operators and local-authority public initiative networks built fibre so that 43.1 million of France's 45.2 million premises, 95.3 percent, were connectable to fibre at the end of June 2026.
The state split the territory: private operators made binding rollout commitments in profitable areas, which Arcep can enforce. In the rest, dรฉpartements and regions tendered public initiative networks, built and run by private firms under concession and co-financed by a state subsidy fund.
Preconditions
A regulator able to turn operators' voluntary rollout commitments into enforceable obligations
Local authorities with the scale and borrowing capacity to tender dรฉpartement-wide networks, backed by a national subsidy window
Wholesale open-access rules so that networks built with public money carry every retail operator
Where else it has been used: Other European countries used gap-funding models under EU state-aid rules. The French design differs in giving the regulator enforcement power over private commitments, which is why the private zones could be left to the market without being abandoned.
Limits: The national coverage figure mixes the parts of the plan. In the very dense zones and the private-initiative zones (AMII and AMEL), private operators invested on their own account, so not every connectable premise is a product of joint delivery. Only the public initiative network zones are the clearly public-private part. Arcep's release gives what remains to be built there, 5 percent, but not the zone's total size. 'Connectable' means a line passes the premises, not that anyone subscribes or that the service is good. The 20 billion euro envelope and the roughly 3 billion euro state subsidy are 2013 announcements, not audited spending. The envelope was announced for the ten years from 2013, and 2.1 million premises were still not connectable in mid-2026. Arcep says more than a third of those are premises where a third party refused the works or the building is still under construction. The deployment figures are reported by operators to the regulator.
๐บ๐พUruguaysince 2009still operatingbears on University-industry collaboration
Uruguay's national research and innovation agency ANII began co-funding joint academia-firm projects through Alianzas para la Innovaciรณn in 2009 and had signed 29 such contracts, committing US$3.85 million and drawing US$1.71 million from firms, by 30 June 2016; in 2023 it approved 23 projects joining the productive sector and academia, matching the 2022 record.
ANII, the state innovation agency, offered non-repayable co-funding only to projects where a firm brought a problem and money and a university or research institute brought the research team, with a firm employee placed inside that team.
Preconditions
A public research base, mainly the Universidad de la Repรบblica, with researchers able to work on applied problems
Firms, often in primary production, with problems worth paying part of the cost to solve
A funding agency outside the ministries with its own evaluation unit and calls that recur every year
Limits: The scale is small: 29 contracts in seven years and US$3.85 million committed is a national instrument, but a thin one, and it says little about how much Uruguayan universities and firms collaborate outside ANII funding. Both figures are the agency reporting on its own instruments, the first from its internal evaluation unit and the second from a year-end press note. The 2023 figure is not limited to Alianzas: it counts every approved project the agency describes as joining firms with academia, under its technology-transfer line, so the two metrics are not one series and no trend can be drawn between them. Signed contracts measure funding decisions, not completed research, patents, products or lasting relationships; the evaluation shows only 14 of the 29 finished by mid-2016. The evaluation report is undated on its cover and the date comes from its own tables.
๐จ๐ฑChilesince 1993still operatingbears on Public-private collaboration
Chile's Ministry of Public Works has delivered public infrastructure through private concession contracts since 1993; in April 2026 its Concessions Directorate reported 81 contracts in force with US$29.615 billion of committed investment across motorways, airports, hospitals, prisons, a reservoir, a desalination plant and a cable car.
The Ministry of Public Works tendered build-operate-transfer contracts under a dedicated concessions law, letting private consortia finance and run motorways and later airports, hospitals and prisons, recovered through tolls and state payments, while a dedicated directorate supervised the contracts.
Preconditions
A concessions law passed with cross-party support and a directorate inside the ministry to tender and supervise contracts
Domestic pension-fund and bank capital able to finance long-tenor infrastructure debt
Projects with user-fee revenue, motorways first, before moving to social infrastructure that needs state payments
Where else it has been used: Chile's concessions law was widely studied in Latin America (Colombia, Peru, Mexico); the retellings tend to leave out the renegotiation record, which is the part that decides whether a copy is cheaper than public procurement.
Limits: The figures are the Directorate's own slide figures, read from a presentation PDF, not audited accounts. Committed investment adds money already spent to money still to be spent, and 35 of the 81 contracts are still in construction, so the headline overstates infrastructure delivered. The number covers contracts in force only, not concessions that have ended, so it is not the system's cumulative output. It says nothing about value for money, renegotiations and supplementary agreements (a known weakness of the Chilean model that the slide does not quantify), contingent fiscal liabilities such as guaranteed minimum revenues, or who pays through tolls. Chile also keeps a large conventional public works budget, so concessions are one channel, not the whole of infrastructure delivery.
๐ฒ๐พMalaysiasince 2018still operatingbears on Public-private collaboration
Bank Negara Malaysia merged its clearing subsidiary with the banks' payment consortium into PayNet, jointly owned by the central bank and the financial industry, which rolled out the Real-time Retail Payments Platform with DuitNow from December 2018; BNM statistics record 5,438.6 million RPP transactions worth RM3,638.1 billion in 2025, up from 109.3 million in 2019.
Bank Negara Malaysia merged its own clearing house with the banks' payment consortium into one jointly owned operator, then used it to build a shared real-time rail with a national proxy (phone and ID number) addressing database that every bank connects to.
Preconditions
A central bank willing to take an equity stake and bring a bank-owned consortium under one shared utility
An incumbent bank consortium with existing ATM and instant-transfer networks to fold in
Regulatory power to require participation and keep interbank fees low
Where else it has been used: Malaysia linked DuitNow to Singapore's PayNow and Indonesia's QRIS through bilateral agreements, and BNM and PayNet joined the BIS Project Nexus proof of concept, so the model is now travelling as cross-border linkage of national fast-payment utilities.
Limits: Transaction volume measures use of a shared rail, not the depth of collaboration: it does not show how costs, governance or pricing are split between BNM and the banks that owned MEPS (11 domestic banks per the 2017 release), nor whether non-bank participants get fair access in practice. Part of the growth substitutes for other electronic channels (Interbank GIRO, Instant Transfer) rather than adding new payments, and the 2019 base is the platform's ramp-up year, so the multiple overstates organic growth. The joint-ownership fact comes from a 2017 announcement of intent ('will be jointly owned'), not a current shareholder register. The central bank is both shareholder and overseer, so this is a central-bank-led utility as much as a partnership.
๐ต๐ฑPolandsince 2017still operatingbears on University-industry collaboration
Poland's science ministry opened the implementation doctorate in the 2017/2018 academic year, funding doctoral candidates to do their research inside a firm or institution. In its first five editions 2,293 projects were approved, 479 people qualified for the sixth edition in 2022/2023, and 69.6 percent of the doctorates ran in the private sector.
The science ministry paid a scholarship to doctoral candidates who were also employed by a firm, with a supervisor on each side. The firm gets a researcher working on its own problem and the university gets a funded doctorate. Each edition was a national competition that universities applied to.
Preconditions
Universities accredited to award doctorates and willing to accept dissertations built on applied work
Firms with enough technical depth to host and co-supervise a doctoral project
A ministry budget line able to fund multi-year scholarships on top of the firm's salary
Where else it has been used: Close in design to France's CIFRE industrial doctorates (since 1981) and Denmark's Industrial PhD. Poland's version runs through annual ministerial editions rather than a standing agency.
Limits: The figures are the ministry reporting on its own programme. They restate a report by the state-owned ลukasiewicz Research Network, not a statistical release. 2,293 is approved projects, not completed doctorates. The page gives no completion, dropout or defence rate, and no figure for patents, products or firm R&D that came out of the work. The latest count read here is from 2022. Figures for the editions since then, including later press totals of about 3,000 participants, were not verified at an official source. A co-funded PhD placement is one channel of university-firm collaboration, and it does not measure the intensity of joint research the indicator asks for, such as co-publications, contract research or co-patenting. 30.4 percent of placements are in the public sector, including state-owned companies and public institutions, so not every project links a university with a private firm. Firms may use the scholarship as a wage subsidy for staff they would have employed anyway.
๐ธ๐ฌSingaporesince 2013dismantledbears on Public-private collaboration
PUB contracted Hyflux's Tuaspring Pte Ltd under a water purchase agreement for Singapore's largest desalination plant, opened in 2013, then terminated the agreement on 17 April 2019 after the operator failed to keep the plant reliably operational, took the plant into public ownership in May 2019 and spent around $80 million on repairs due for completion in mid-2021.
PUB bought desalinated water from a private developer under a long-term purchase agreement and, when the developer's finances failed, used the contract's default and termination remedies to take the plant into public ownership and repair it itself.
Preconditions
A water agency with in-house capacity to run the same plant technology it contracts out, here PUB's own Tuas Desalination Plant
Termination and step-in clauses written into the purchase agreement before anything went wrong
A private partner whose balance sheet depended on an integrated water and power plant that could lose money on the power side
Limits: This is one contract failing, not the end of Singapore's public-private water model: PUB's own page lists Keppel Marina East (2020) and Jurong Island (2022) desalination plants built and run by private firms after the collapse, so the partnership approach survived and the record documents the loss of one delivery inside it. The $80 million is a ministerial answer to a parliamentary question, not an audited figure, and it covers repairs only, not PUB's operating costs since takeover or what the private partner's investors lost. The original 25-year term and contracted capacity of the agreement are reported only by press and secondary sources found today, so the record does not carry them. A hostile reviewer can also read the episode as the state's contractual safeguards working, which is how the minister presented it: the water supply was not interrupted.
๐จ๐ฆCanadasince 2009dismantledbears on Public-private collaboration
Canada's federal Crown corporation PPP Canada, operational from 2009, committed over C$1.3 billion through the P3 Canada Fund to 25 public-private partnership projects across the country before the government dissolved it, with operations ceasing at the end of 2017 and the corporation dissolved effective March 31, 2018.
The federal government created a Crown corporation with a dedicated fund that paid up to a quarter of eligible costs to provinces, municipalities and Indigenous communities that procured infrastructure as public-private partnerships, plus template contracts and screening of federal projects; once it judged the market mature, it folded the corporation into the infrastructure department.
Preconditions
Provincial P3 agencies and a domestic market of builders, financiers and advisers already able to bid long design-build-finance-maintain contracts.
A federal transfer large enough to change a smaller jurisdiction's procurement choice without paying for the whole asset.
A department able to absorb the portfolio and its long-dated payment liabilities when the agency closed.
Limits: This is a dismantling by decision, not a collapse. The government's stated reason was that PPP Canada had 'successfully fulfilled its mandate', and its files were transitioned to Infrastructure Canada. The corporation carried performance-based payment liabilities of approximately C$1.3 billion over five years. A hostile reviewer would say the record shows a federal agency closing, not a national capability lost. The C$1.3 billion is commitments rather than disbursements (the audited statements show far smaller annual payments to recipients) and is a federal top-up of up to 25% of eligible costs, so it does not measure the scale of the joint delivery, its value for money or whether the 25 assets performed. The narrative and the figure are the corporation's own annual report; only the financial statements carry an Auditor General audit opinion.
๐จ๐ณChinasince 2015dismantledbears on Public-private collaboration
China's Ministry of Finance ran a national government and social capital cooperation (PPP) programme under the 2015 State Council guidance, whose management library held 10,332 projects with 16.6 trillion yuan of investment by the end of October 2022; in November 2023 the State Council ended the 2015 regime for every project not yet procured and replaced it with a concession-only, user-fee mechanism that bars availability payments.
The Ministry of Finance built a national PPP framework, a central project library and disclosure platform from 2014-2015 and pushed local governments to procure infrastructure through it; when the library became a channel for hidden local debt, the Ministry purged it in 2017-2018 and the State Council replaced the regime in 2023 with user-fee concessions run by the planning commission.
Preconditions
A central finance ministry able to define eligibility and hold a national register of local projects
Local governments with borrowing limits that make off-balance-sheet payment commitments attractive
A fiscal-risk authority willing to purge and then close a programme it had itself promoted
Where else it has been used: China copied PPP units and project libraries from the United Kingdom and multilateral advisers; the reversal mirrors the UK's own 2018 abolition of PF2, in both cases after availability payments were judged to hide public liabilities.
Limits: Both project figures sit in articles by Ministry of Finance affiliated newspapers (China Finance and Economics News, Economic Daily) republished on mof.gov.cn, not in a Ministry statistical release; the Ministry's own monthly library reports on jrs.mof.gov.cn stop at January 2020, so the 2022 figure cannot be checked against a primary table. The source is a ministry press page rather than a statistics release. A count of library projects measures entries in an administrative register, not collaboration: many entries were government-paid availability-payment schemes that the Ministry itself purged in 2017-2018 as disguised local borrowing, so a large share of the 'partnership' was public debt with a private contractor in front. The library count is cumulative and does not show how many projects reached operation or delivered value for money. Dismantling applies to the 2015 regime for new projects; existing contracts keep running and PPP continues as user-fee concessions under the National Development and Reform Commission, so this is the end of one national model, not of public-private delivery in China. The 2023 notice gives the reasons in general terms (hidden local-government debt, private investment) and no figure for what failed.
๐ฎ๐ชIrelandsince 2013still operatingbears on University-industry collaboration
Science Foundation Ireland funded a network of 16 university-based research centres from 2013, whose cumulative SFI investment of โฌ684 million was matched by โฌ638 million of industry commitment by 2021 (Annual Report 2021), and which by the end of 2022 had signed 1,426 collaborative research agreements with 688 companies, 366 of them in Ireland.
1,426 agreements, cumulative collaborative research agreements signed by sfi research centres with companies, end of 2022. 688 companies, companies party to those agreements, end of 2022. Science Foundation Ireland, Annual Report 2022, retrieved 2026-10-01.
The move, and what it needed
Science Foundation Ireland funded large, multi-university research centres for six-year terms and made part of each award conditional on industry co-funding, so every centre had to sign agreements with firms to keep its public money.
Preconditions
A funding agency with a statutory mandate to tie research money to economic objectives
A large base of multinational R&D-performing firms present in the country
Universities able to pool researchers across institutions under one centre's governance
Limits: These are the funder's own counts of its own programme, in an annual report, inside an 87-page PDF. An agreement count says nothing about the size, duration or research content of each agreement; in 2022 alone 136 agreements carried โฌ32.2 million of industry commitment. The โฌ638 million is industry commitment, not verified cash received, and the programme's design required industry co-funding, so collaboration is partly a condition of funding rather than demand from firms. 317 of the 688 companies are multinationals, so the figure mixes Irish firms' engagement with foreign affiliates attracted by tax and investment policy. Science Foundation Ireland was merged into Research Ireland in 2024, so the series ends with 2023 reporting under the old name and the sfi.ie URLs may move.
๐ฎ๐ณIndiasince 2017operating below its peakbears on Public-private collaboration
India's Ministry of Civil Aviation paid viability-gap funding to private airlines to fly underserved regional routes from April 2017. Of the 774 routes awarded in the first three rounds, 371 commenced, and only 174 were still flying in March 2023; only 54 of the 112 routes that completed their three-year concession kept operating afterwards.
The Ministry of Civil Aviation auctioned regional routes to private airlines, paid the winners a viability-gap subsidy for three years funded by a levy on other flights, and asked states and the Airports Authority to revive unused airstrips; most routes did not survive the end of the subsidy.
Preconditions
A levy on trunk air traffic large enough to fund subsidies to regional routes
State governments and the Airports Authority willing to upgrade small airports at public cost
Private airlines with the aircraft and balance sheets to fly thin routes, which several winning bidders lacked
Limits: The 371 is a cumulative count of routes that ever commenced, not a simultaneous peak, so the fall to 174 overstates how many routes were flying at any one time. Routes from the third round had less time to lapse, which flatters their survival rate. The audit covers UDAN 1 to 3 to March 2021 with route figures updated to March 2023, so later rounds and the scheme's current state are not shown. A lapsed route is a commercial failure for the airline as much as a policy failure, and some routes stopped because airlines collapsed. The figures come from the scheme's own RCS cell, as the audit notes. The 120 page report is a 12 MB PDF and the quoted table is read from text extraction.
๐น๐ทTurkeysince 1986still operatingbears on Public-private collaboration
Turkey's central government signed 277 public-private partnership contracts from 1986 to September 2025, worth US$112.09 billion in September 2025 prices across transport, energy and health, with 262 in operation, including 18 PPP city hospitals holding 29,325 registered beds.
Successive governments, and the Strategy and Budget Office as coordinator, used build-operate-transfer, build-lease-transfer and operating-rights contracts to move large transport, energy and hospital builds off the annual budget. The state then pays through guarantees and availability payments over the contract life.
Preconditions
Sector laws for each contract model, including Law 6428 for health facilities
A treasury willing to issue long-dated demand and availability guarantees
Domestic contractors large enough to bid, often in consortium with foreign firms and lenders
Limits: The registry's own note on this gap says PPP investment databases measure infrastructure finance, not collaboration capacity, and this record rests on exactly that kind of figure. US$112.09 billion is contracted investment size in current prices, not money spent or value delivered. It includes transfers of operating rights (ฤฐHD), which are closer to concessions or privatisation than to joint delivery. The government reports on its own programme. The same document shows what the collaboration costs the public side: in 2026 it programs 57.5 billion TL for goods and services and 78.7 billion TL in capital spending to run the completed PPP city hospitals, and adds 101.8 billion TL in current transfers to finance demand guarantees on completed PPP transport projects. A hostile reviewer would say those guarantees left the risk with the state, so the record shows private construction finance with public risk-bearing more than shared delivery. No cost or schedule performance is reported for individual contracts, and the count spans four decades and very different legal models.
๐ซ๐ฎFinlandsince 2008dismantledbears on University-industry collaboration
Finland ran six Strategic Centres for Science, Technology and Innovation from 2008, company-form consortia of firms, universities and research institutes funded about 60 percent by Tekes, with funding of EUR 88 million in 2014; the Sipilรค government's austerity programme ended the earmarked funding at the end of 2015 and the programme was closed, with the last Tekes-funded projects wound down between 2016 and 2018.
The state made long-term, multi-year research programmes conditional on firms, universities and institutes owning a joint company and co-writing its research strategy, with the innovation agency paying the majority share. The structure held collaboration together only while the earmarked public share kept flowing.
Preconditions
An innovation agency with a large applied-research budget to earmark for a decade
Lead firms in each cluster willing to co-own a company with universities and pay roughly 40 percent
A political consensus on the earmark, since the model had no independent revenue when it was withdrawn
Where else it has been used: The SHOKs borrowed from the competence-centre model used in Sweden and Austria. The Finnish lesson is that a dedicated earmark is quick to cut in an austerity round, while centres that had built their own membership base, like DIMECC, survived the cut.
Limits: The figure is inside a PDF (page 12, Box 1). The case study is hosted by the OECD as a country contribution to its knowledge-transfer project, but it was written by Finnish consultants with a Business Finland co-author, so it is the funder's side of the story rather than an OECD assessment. The EUR 88 million is one year's funding and the source does not say whether it is Tekes money alone or the full public and private budget; the same box says Tekes covered on average 60 percent. Funding is an input: the source carries no count of joint projects, firms or outcomes. The end was a budget decision, not a judgement that the centres failed: the same document says two of the six SHOKs (DIMECC and CLIC) continued as open innovation companies owned by firms, universities and institutes, and that ecosystem funding moved to Business Finland's Growth Engine instrument, so collaboration capacity was partly redeployed rather than wholly lost. The 2013 external evaluation the case study cites (Licence to SHOK?) was openly critical of the programme's governance and its limited commercial and scientific impact. The same PDF dates the programme's start to 2006-2007 in Box 4 and to 2008 in Box 1; started follows Box 1.
๐จ๐ญSwitzerlandsince 2018still operatingbears on University-industry collaboration
Innosuisse, the Swiss federal innovation agency, recorded CHF 113.95 million of contributions in 2025 to innovation projects in which a Swiss research institution works with a company or other implementation partner that bears 40 to 60 percent of the project cost, and held CHF 144.3 million of further commitments to such projects at the end of the year.
The federal agency pays only the research institution's side of a project, and only when a partner in the economy bears a large share of the cost and contributes cash, so public money follows demonstrated private demand for academic expertise.
Preconditions
Universities of applied sciences with a mandate for applied research and staff used to firm-facing projects
A dense base of SMEs able to co-fund 40 to 60 percent of a project
A stable federal budget line and an agency with independent project selection
Where else it has been used: The co-funding rule resembles Fraunhofer's income-linked base grant in effect, but works project by project. Austria's and Finland's agencies use similar rules; the Swiss version depends heavily on the universities of applied sciences, which have no close equivalent in many systems.
Limits: The figures are inside a PDF, in the financial statements (PDF page 32, printed page 29, note 17; and PDF page 36, printed page 33, note 22), in thousands of francs. This is the agency reporting on its own spending. The money is the public share paid to research partners: it does not include the firms' own 40 to 60 percent or their cash contributions, so it understates the joint effort, and the accounts give no count of projects or firms. Research partners include universities, universities of applied sciences and non-university research bodies, and implementation partners can be administrations and non-profits as well as firms, so not every franc is university-firm collaboration. Spending measures input, not whether projects produced products. Started is 2018, the year the current agency took over the instrument from its predecessor; neither cited source states that date, and the instrument is older.
๐ซ๐ทFrancesince 1981still operatingbears on University-industry collaboration
France's research ministry has funded Cifre industrial doctorates since 1981, paying an employer 42,000 euros over three years to hire a doctoral candidate whose thesis research is run in partnership with an academic laboratory; ANRT, which operates the scheme for the ministry, accepted 1,700 new Cifre in 2025 across 1,029 distinct employers.
The state subsidises a firm to employ a doctoral candidate whose thesis is supervised by an academic laboratory, and requires a collaboration contract between employer and lab, so each grant creates a person who works in both and a formal research agreement between them.
Preconditions
Doctoral schools willing to supervise theses whose questions come from employers
A research tax credit that covers much of the employer's remaining cost
A trusted intermediary that vets proposals and contracts quickly
Where else it has been used: Denmark's Industrial PhD, Spain's Doctorados Industriales and Poland's Doktorat wdroลผeniowy follow the same idea. The French scheme's scale comes from four decades of continuity and its interaction with the research tax credit.
Limits: The figures are inside a PDF (PDF pages 17 and 19, printed pages 15 and 17). ANRT is an association that runs the scheme on the ministry's behalf, so this is the programme reporting on itself; the ministry's own page was not reachable to cross-check. Employers include local authorities and associations (13 percent in 2025) as well as firms, so not every contract is university-industry collaboration. The count is of new contracts, not completed theses or research outputs. 2025 was below the scheme's own target: the same report says the ministry expected growth from 1,850 to 1,950 new Cifre and that 1,700 fell 13 percent short, with large firms' demand down 19 percent, so the 2025 figure is a dip against the 2024 level rather than a peak.
๐ธ๐ชSwedensince 1995delivered and closedbears on University-industry collaboration
Sweden's NUTEK, and later Vinnova and the Swedish Energy Agency, ran a ten-year programme from 1995 that built 28 competence centres at eight universities, each founded on a three-party contract between a university, a group of firms and the agency; more than 300 companies joined as contracting parties and industry put in 1.7 billion kronor between 1995 and 2005.
Each centre was a contract among a university, a group of firms and the state agency, with matched funding from all three, boards with a non-academic majority and chair, and staged three-year funding that depended on international evaluation, which kept firms engaged in the whole research agenda instead of single projects.
Preconditions
Research-intensive firms prepared to commit staff and money for up to ten years
Universities willing to cede board control to non-academics
An agency able to run staged evaluations and withdraw funding from weak centres
Where else it has been used: The source notes that Austria, Hungary and Estonia started similar programmes inspired by the Swedish one; Finland's SHOKs and Austria's COMET centres are later relatives.
Limits: The figures are inside a PDF (PDF page 41, printed page 39). The report is published by Vinnova but written by evaluation consultants (Faugert & Co and Technopolis), and Vinnova notes that its analysis series does not necessarily state the agency's position; it in turn cites Vinnova's own 2006 summary. Roughly 70 percent of the industry contribution was in kind, so the 1.7 billion kronor is mostly valued staff time and equipment, not cash. The company count is of contracting parties over ten years, not firms active at any one time, and about 50 firms were counted in more than one centre. The programme ended as planned and was succeeded by VINN Excellence Centres, so its end is not a reversal; this record says nothing about the current centres.
๐ช๐ธSpainsince 2014still operatingbears on University-industry collaboration
Spain's State Research Agency funds four-year contracts for doctoral candidates employed in firms and other entities to write a thesis on an industrial research or experimental development project; in the 2024-2 call, resolved in October 2025, it awarded 65 grants worth 7.6 million euros.
The state co-funds a firm's employment of a doctoral candidate for four years on condition that the thesis is built on the firm's own research or development project, tying a university doctoral programme to a company problem through one person.
Preconditions
Universities that accept company-based thesis topics and co-supervision
Firms with an R&D project long enough to carry a thesis
A national agency running regular competitive calls
Where else it has been used: The design follows France's Cifre and Denmark's Industrial PhD. Catalonia and Galicia run regional versions alongside the national one.
Limits: The figure is one call's awards, not the stock of industrial doctorates in progress; The 2024 round was split into two calls and only the second is counted here, so the figure understates a year's awards. It is small next to France's Cifre. The release does not mention universities at all: the university side is implied by the doctoral thesis, not shown. It does not count the firms, and beneficiaries include 'other entities' as well as companies. Awards are funding decisions: the release does not say how many contracts were signed or theses completed. The start year (2014, the first call under the 2013-2016 State Plan) is not stated on the cited page. The release says funding doubled in 2024 and held in 2025, so the current scale is recent.
๐ต๐ญPhilippinessince 2017still operatingbears on University-industry collaboration
From 2017 the Department of Science and Technology's CRADLE programme funded 82 projects in which a university or research institute did R&D on a problem set by a Filipino company, worth 373.6 million pesos and involving 36 HEIs or RDIs and 83 companies by early 2022.
DOST pays for an R&D project only when a Filipino company names the problem, commits counterpart funding and agrees to adopt the result, and a university or research institute does the research.
Preconditions
University and RDI researchers able to work on applied industrial problems
Companies willing to co-fund and commit to adopting the result
A public grant line, here the Science for Change Program, that can contract with both sides
Where else it has been used: The design is close to Turkey's TรBฤฐTAK 1505 and Malaysia's PPRN: the firm states the problem and shares the cost. DOST later added an industry-led variant, iCRADLE, for larger consortia.
Limits: The figures are DOST's own, stated in a call for proposals rather than an audited annual report, and they are cumulative from 2017 to the date of the call (March 2022), so no single year can be read off them. The number sits inside a PDF, on page 1. A funded project is an input: the document does not say how many were completed, how many technologies the partner companies adopted, or what the companies paid as their required counterpart of at least 20 percent. 82 projects over five years is small against the size of the Philippine economy, and the partner pool includes public research institutes as well as universities. No newer official count was found: the operating status rests on DOST's undated S4CP programme page, which still describes CRADLE as being implemented, so the 2022 figures may understate the programme's current size.
๐ฒ๐พMalaysiasince 2015still operatingbears on University-industry collaboration
Malaysia's Ministry of Higher Education approved 680 Public-Private Research Network projects involving 634 companies by the end of 2022, with a total allocation of RM27.7 million, and 469 of them had closed as successfully completed.
The ministry matches a company's own contribution, on a sliding scale by company size, to pay a university or public research institute researcher to solve a problem the company brings, and pays the researcher's fee only after the project is certified complete.
Preconditions
University researchers allowed and rewarded to do consultancy-style projects for firms
A register of viable SMEs and agencies such as SME Corp that can vouch for them
A small grant line run centrally with a single online application system
Where else it has been used: PPRN 2.0 widened the original scheme after 2021. The same firm-states-the-problem design appears in the Philippines' CRADLE and Turkey's TรBฤฐTAK 1505.
Limits: The figures come from a ministerial speech published on the ministry's site, not from an annual report or audit, and the page is in Malay with no English translation. They are cumulative since the programme began, so they do not show the annual rate. The projects are small: RM27.7 million across 680 projects is roughly RM41,000 of public money each, and the speech describes them as demand-driven projects to raise company productivity rather than joint research of the kind the indicator's survey item imagines. The productivity gain the speech quotes comes from a 2019 study by the Malaysia Productivity Corporation that is not cited here. The page writes the amount as 'RM 27,797,149.25 juta', a typo for RM27.8 million, and the record uses the rounded RM27.7 million the same sentence gives. The start year is not stated in the speech, which places PPRN under the 2015-2025 higher education blueprint.
๐ฟ๐ฆSouth Africasince 2001operating below its peakbears on University-industry collaboration
South Africa's Technology Innovation Agency runs a network of university-based Technology Stations that gave 2,907 small firms science, engineering and technology services in 2024/25, against 3,167 at the five-year high in 2021/22, while the programme's budget fell from R105.9 million in 2020 to R93.98 million in 2025 and only 80 percent of the stations were operational.
2,907 SMMEs, smmes accessing set services through the technology stations network, 2024/25. 3,167 SMMEs, smmes accessing set services through the technology stations network, five-year peak, 2021/22. Technology Innovation Agency (TIA), Annual Report 2024/25, retrieved 2026-10-02.
The move, and what it needed
The state funds equipped technology stations inside universities, mostly universities of technology, whose staff and students sell subsidised testing, prototyping and process improvement to small firms that could not afford the equipment.
Preconditions
Universities of technology with engineering staff and laboratory space
Steady operating and capital budgets, since stations with ageing equipment stop functioning
Small manufacturers close enough to a station to use it
Where else it has been used: The model resembles Canada's college applied research centres and Germany's Steinbeis transfer centres. Here the erosion came from flat budgets and no new capital equipment rather than from a decision to close it.
Limits: The numbers sit inside a 188-page PDF, in the outcome table on PDF page 41 (printed page 39), and the budget figures in the Outcome 3 narrative on PDF page 42. The peak is the highest of the five years TIA reports, not a long-run maximum, and the fall from it is about 8 percent, which is computed here. The yearly count is noisy (1,990, 3,167, 2,671, 3,114, 2,907), and 2024/25 sits above two of the five years, so the erosion rests on what TIA reports about capacity: the budget cut, a fifth of the stations not functioning and no new high-end equipment, rather than on the count alone; the budget cut is in nominal rand and larger in real terms. A count of firms served measures university-to-firm technical service, testing and prototyping, which is a thinner form of research collaboration than joint projects: the same report says 73 percent of station equipment time goes to firms. The agency also reports a separate SMME figure (3,775) across all its programmes; this record uses only the Technology Stations line. A 2022 National Treasury spending review recommended moving the programme out of TIA altogether. The start year is taken from the programme's establishment under the former Department of Science and Technology and is not stated in the cited report.
๐จ๐ฑChilesince 2004still operatingbears on University-industry collaboration
Chile's development agency CORFO had 49 technology programmes and consortia in execution during 2024, 38 technology programmes and 11 technology consortia, the consortia joining firms with universities and technology centres in multi-year R&D portfolios, after awarding six new consortia in 2023 with 1,500 million pesos of co-financing each.
CORFO co-finances long-lived consortia and programmes, for up to six years per award, at about 30 percent of cost for new consortia, inside a stated horizon of more than ten years, in which firms in one sector and universities or technology centres share a portfolio of R&D projects and the intellectual property they produce.
Preconditions
Export sectors such as fruit, wine and aquaculture with firms that face shared technical problems
Universities with applied research groups in those sectors
An agency able to commit funding over a horizon longer than a political term
Where else it has been used: Chile's first-generation consortia of 2004 drew on Australia's Cooperative Research Centres; several, such as Biofrutales, still run.
Limits: The count is CORFO's own, from a slide deck hosted on CORFO's asset store, not an audited report, and it is a stock of initiatives in execution, including six new and ten finished in the year, not a flow of research output. The figure sits inside a PDF (page 13, footnote; the 38 and 11 split on page 9). Not every technology programme is led by a university: several of the 38 programmes are executed by a single company with no university partner named in the deck, so only the 11 consortia are collaborative by design and the 49 overstates university-firm collaboration. The second metric comes from a separate Ministry of Economy press release of 12 December 2023 announcing the 2023 awards, which is an award notice rather than evidence of delivery. The deck also gives figures for participating firms and private co-funding whose layout in the PDF does not make clear which label each number belongs to, so they are not used. The instrument began as the 2004 Consorcios Tecnolรณgicos Empresariales and has been redesigned since; that start year is not stated in either cited source.
๐บ๐ธUnited Statessince 2012still operatingbears on University-industry collaboration
In fiscal 2023 the 16 reporting Manufacturing USA institutes, public-private consortia sponsored by the Departments of Commerce, Defense and Energy, managed 929 applied research and development projects and held 2,921 memberships, 655 of them universities, community colleges and other academic members alongside 1,315 small and 493 large manufacturers.
Federal agencies fund a standing consortium per manufacturing technology for a fixed term; firms, universities and colleges pay membership and cost share to sit on a shared applied research and workforce agenda that no single member would fund alone.
Preconditions
Multi-year federal base funding with an expectation of non-federal cost share
Research universities and manufacturers in the same technology area willing to share pre-competitive IP
A national program office that sets common metrics across institutes run by different agencies
Where else it has been used: The model borrowed explicitly from Germany's Fraunhofer institutes, but institutes are member consortia with term-limited federal money rather than permanent contract-research labs, which leaves each one to find a sustainable funding model after its federal term.
Limits: The figures are the institutes' own reports aggregated by the program office, in Table 2 on PDF page 13 (printed page 11) of the report, and they cover fiscal 2023 although the report was posted in March 2026, so they are about three years old. Sixteen of the 17 institutes report; EPIXC joined mid-year and is excluded. The counts are memberships, not distinct organisations: a university that belongs to three institutes counts three times. The academic count mixes research universities with community colleges, which join mainly for workforce training, and the report does not say how many of the 929 projects actually pair a university with a firm; some are workforce or firm-only projects. Federal base funding was 160.4 million dollars in the year against 379.5 million of cost share and other funding, all as reported by the institutes, with no counterfactual for what members would have done anyway. The start year is the first institute, America Makes, established August 2012 per the report's Appendix B; the program's statute is the Revitalize American Manufacturing and Innovation Act of 2014.
๐จ๐ฆCanadasince 2003still operatingbears on University-industry collaboration
Mitacs, the federally and provincially funded not-for-profit, placed 4,859 students, recent graduates and postdoctoral fellows in Accelerate research internships with 2,314 host organisations under 2,406 professors in 2024-25, out of 20,885 internships across all its programmes that year.
A national intermediary splits the cost of a graduate student's or postdoc's research project inside a firm with the firm, using federal and provincial money, and uses the student's supervisor as the university link.
Preconditions
A large pool of graduate students whose supervisors accept industry-defined projects
Firms willing to pay a fixed partner contribution per internship unit
Stable federal and provincial co-funding, since unit caps follow budget cuts directly
Where else it has been used: Close relatives are France's Cifre and Spain's Doctorados Industriales, which fund a whole doctorate inside a firm; Mitacs uses short, stackable units instead, which scale fast but tie volume closely to each year's budget.
Limits: This is Mitacs reporting on itself to its federal funder: the Accelerate counts are on PDF page 28 and the 20,885 total on PDF page 2. Mitacs is a not-for-profit rather than a government body; ISED funding was 48 percent of its income in 2024-25 and the rest came from provinces and partners. Host organisations include not-for-profits, hospitals and municipalities as well as firms, so 2,314 overstates the firms involved. An internship is a funded unit of four to six months and one intern can hold several, so internships and interns are different counts; the 20,885 total also includes Globalink internships that bring international students to Canadian labs with no firm involved. The next report (2025-26, July 2026) gives 20,819 internships in total, against 21,432 in 2022-23. The capacity is now shrinking by plan: in 2024-25 Mitacs capped Accelerate units per university for lack of federal money, and its 2026-27 corporate plan (PDF page 9) foresees 12,540 internships on a 120 million dollar ISED grant, against more than 200 million of federal support in 2022-23. If that plan is delivered this record becomes an erosion. The 2003 start year, the first Mitacs research internship program, comes from ISED's evaluation of its funding to Mitacs, not from the cited report.
๐ฏ๐ตJapansince 2020still operatingbears on University-industry collaboration
Japan Science and Technology Agency ran 42 COI-NEXT co-creation centres in April 2023, 30 full-scale and 12 in a development stage, most led by universities and some by national research institutes, which in fiscal 2022 counted 1,073 participating organisations, 256 universities and research institutes and 817 companies and others, with 2,566 participants from the company side.
JST funds a university-led centre for up to ten years around a stated vision for society, requires the university's leadership to commit and firms to contribute resources, and steps down its funding in the final years so the centre must find its own.
Preconditions
Universities with industry-liaison offices able to run a multi-partner centre
Firms willing to second staff and contribute in kind over years
A funding agency able to commit to a ten-year horizon with stage-gate reviews
Where else it has been used: COI-NEXT succeeds the 2013 to 2021 Center of Innovation programme (COI STREAM) and carries over its centre format, adding regional centres led by local universities with municipalities and local firms.
Limits: The source is the funding ministry's own briefing, written to defend the programme's budget before a government spending review, and the participant counts are its self-reported totals. The counts are on PDF page 13 (printed page 12) and the centre counts on PDF page 10 (printed page 9). The organisation counts are for fiscal 2022 alone and are now more than three years old. The category is 'companies and others' (ไผๆฅญ็ญ), which can include municipalities and other non-academic bodies, so 817 is an upper bound on firms. The ministry's claim that external income of about 15.4 billion yen already exceeds about 14 billion yen of support counts resources the firms provided at their own valuation. The 2020 start year is inferred from the heading 'activities FY2020 to FY2022' on the same page. MEXT's fiscal 2026 budget sheet plans support for about 37 centres in the main fields, fewer than 42, which partly reflects centres graduating from support.
๐ง๐ทBrazilsince 2018still operatingbears on University-industry collaboration
Brazil's research council CNPq, with CAPES co-funding the 2024 call, has run four calls of its Mestrado e Doutorado Acadรชmico para Inovaรงรฃo programme (2018, 2020, 2022 and 2024), in which about 80 science and technology institutions and more than 200 companies had taken part and at least 5,000 scholarships had been granted by 2024 for undergraduate and graduate student research on projects defined with partner firms.
The national research council pays scholarships for students whose thesis or initiation project is a research project defined with a partner firm, run through a university that must have a technology transfer office and an innovation policy, and since 2024 asks the firm for a financial counterpart.
Preconditions
Graduate programmes willing to accept industry-defined thesis topics
University technology transfer offices able to manage firm agreements and IP
Firms with enough R&D activity to host and co-supervise students
Where else it has been used: The design follows industrial doctorate schemes such as France's Cifre, but pays the scholarship to the student through the university rather than as a salary from the firm, which lowers the firm's commitment.
Limits: (append) The 2018 start year is also stated on the cited page, which lists the earlier calls as 2018, 2020 and 2022.
๐ณ๐ฑNetherlandssince 2006dismantledbears on University-industry collaboration
NWO's Perspectief programme, funded mainly by the Ministry of Economic Affairs, awarded 12 public-private technology consortia in 2025 with 46.4 million euros of subsidy, each joining at least three knowledge institutions with at least four companies or other users paying at least 30 percent of costs; the ministry's own page records that the last Perspectief round started in July 2026.
The research council runs a yearly competition for multi-year consortia in which universities receive the subsidy and companies and civil-society users co-fund in cash and kind, sit on a user committee and get first access to the results.
Preconditions
A ministry willing to route industrial policy money through the research council
Firms willing to co-fund early-stage (TRL 1 to 6) research with no exclusive claim to it
Universities of technology able to assemble multi-group consortia
Where else it has been used: Perspectief was the Dutch counterpart of collaborative programmes such as the competence centres in Sweden; the Netherlands ended it while announcing a new national agency for disruptive innovation, so the bridging function moves to an instrument not yet built.
Limits: The counts are NWO's awards as published by the ministry, and an award is a project starting, not a result: projects run for up to six years, so awarded consortia will continue until the early 2030s although no new ones will be funded after the 2026-27 round. 2025 was an unusually large year: the same table gives 4 projects in 2024 and 5 to 9 in most years since 2016, so 12 is the peak, not the norm. The page's table of participating companies has two columns whose labels do not make clear which is total and which is SMEs (70 and 79 in 2025), so company counts are not used. The termination is stated on the cited page only as 'the last round started in July 2026'; the reason, the end of ministry funding in the first supplementary budget to the 2026 Spring Memorandum, is in NWO's news release of 2 April 2026, which also states NWO's objection. The final round has 26.5 million euros by NWO's account. The start year is approximate: NWO's release says the programme has run for twenty years, and the ministry's series begins in 2016.
HM Treasury's Private Finance Initiative signed on average 55 contracts a year between 1997 and 2010 and left 704 current PFI and PF2 projects with a capital value of ยฃ57 billion at 31 March 2018, after signings fell to an average of nine a year after May 2010 and the Chancellor announced at Budget 2018 that the government would no longer use PF2 for new projects.
HM Treasury let departments and local bodies buy hospitals, schools, roads and prisons as long design-build-finance-operate contracts paid by performance-linked unitary charges, keeping the capital cost off the up-front budget; after value-for-money criticism it reformed the model as PF2 in 2012 and then stopped using it for new central-government projects in 2018.
Preconditions
A deep domestic market of contractors, lenders and advisers able to price 25 to 30 year availability contracts.
Fiscal rules that rewarded keeping capital spending off the public balance sheet, which made the model attractive before it made it suspect.
A central unit (Partnerships UK, later the Infrastructure and Projects Authority) to standardise contracts and collect portfolio data.
Where else it has been used: The PFI model was copied widely, including by Canada and Japan; the UK case is the clearest example of an originating country closing the route for new projects while carrying the legacy portfolio for decades.
Limits: This is a dismantling by decision of the procurement route, not of the assets: the 704 contracts keep running, with ยฃ188.35 billion of forecast unitary charges over the next 30 years, so the record shows a national collaboration model closed to new deals rather than a lost stock of hospitals and schools. The decline started well before abolition (an average of 55 signings a year from 1997 to 2010 against nine a year after May 2010), and PF2, introduced in 2012 'in response to widespread concerns about value for money', was used only six times. A hostile reviewer would say PFI is evidence of joint delivery at scale and of poor value for money at once: the scale figure says nothing about whether the risk transfer paid off, and the Carillion liquidation hit two PF2 projects. Coverage is departments and devolved administrations under standard PFI and PF2 terms only; other PPP forms such as NHS LIFT are excluded, and Scotland and Wales run their own models that the 2018 announcement did not affect. The 1992 start year is not stated in the document, whose earliest year is 1997. The figures are in a PDF: the 55-a-year and nine-a-year averages and the PF2 announcement on PDF page 5 (paragraphs 1.5 and 1.6), the 704 projects and ยฃ57 billion on PDF page 8 (paragraphs 2.2 and 2.3). The publication is HM Treasury reporting on its own programme.
๐ฏ๐ตJapansince 1999still operatingbears on Public-private collaboration
Under the 1999 PFI Act, Japanese national and local governments had published implementation policies for a cumulative 1,219 PFI projects by 31 March 2026, 80 of them public facility operating concessions, with a cumulative initial contract value of ยฅ10.47 trillion (104,730 hundred-million yen).
The national government legislated a standard route for private finance, design, construction and operation of public facilities, and a Cabinet Office promotion office published guidelines, case collections and annual counts, so that municipalities could procure schools, waste plants and civic buildings through PFI and, from 2011, grant operating concessions for airports and water works.
Preconditions
Fiscal pressure on local governments that made spreading capital costs over a service period attractive.
Large domestic contractors and banks willing to form special-purpose companies for small and mid-size local projects.
A central promotion office able to issue standard guidance and keep the national register of projects.
Where else it has been used: Japan adapted the UK PFI model in 1999 and kept using and widening it, adding concessions in 2011, while the UK stopped using it for new central-government projects in 2018.
Limits: The count is of projects whose implementation policy was published, which the Cabinet Office tracks by its own survey; a published policy is not a delivered asset, although the note says projects cancelled during service or abandoned before service began are excluded. Most projects are small and local: the sectoral table in the same annex shows local governments sponsoring the large majority (for example 397 of 454 culture and education projects), so the record evidences a widely used procurement route rather than large national infrastructure. The contract value covers initial contract amounts where a public payment was fixed and excludes concession fees, and the annex itself says it differs from the project-scale measure in the 2026 PPP/PFI Action Plan. Nothing here shows value for money or service quality. The figures are read from charts and a table in a PDF (project counts on page 1, contract values on page 2, concessions on page 3, sector table on page 4); the summary page at https://www8.cao.go.jp/pfi/pfi_jouhou/pfi_joukyou/pfi_joukyou_r7.html states 1,219 and 80 in text. The Cabinet Office is the programme's own promoter.
๐ฉ๐ชGermanysince 1969operating below its peakbears on Public-private collaboration
Germany's dual training system, in which firms employ apprentices under contracts registered by the chambers while state vocational schools teach them, recorded 475,950 new training contracts in the year to 30 September 2025, down from 569,379 in 2011, while 54,372 offered places stayed unfilled.
Federal law sets training regulations agreed with employers and unions, firms recruit and pay apprentices and deliver workplace training, chambers register contracts and run examinations, and the Lรคnder run part-time vocational schools, so qualification is a joint product that no single party can deliver alone.
Preconditions
Chambers with statutory membership able to register contracts and examine to one national standard.
Employers who treat training as recruitment and accept net training costs.
A school system that channels a large share of school leavers into apprenticeships rather than university.
Where else it has been used: The dual model is widely copied (the corpus carries Switzerland's che-dual-vet); German data show that even the reference system loses volume when the supply of school leavers and their preferences shift.
Limits: The fall of about 16 percent from the 2011 high is computed from two published values (column 1). How much of it is the collaboration weakening is contested. On the applicant side, unfilled training places rose from 17,139 in 2009 to a high of 73,431 in 2023, which points to fewer and differently choosing school leavers, demography and the shift to university. On the employer side, the same table shows the places offered (column 6) falling from 598,935 in 2011 to 530,322 in 2025, about 11 percent, including a drop of 25,338 in 2025 alone, while unfilled places fell to 54,372 and applicants left without a place or an alternative (column 3) rose to 38,784, the highest in the series; the latest year reads as firms offering fewer places, not only as missing applicants. A hostile reviewer would also say the system still enrols nearly half a million new trainees a year, the largest joint public-private training delivery in the corpus. The table does not show the share of firms that train, which is the direct measure of employer participation. The 2020 figure (467,484) is a pandemic low and 2025 sits above it. The series starts in 2009, so earlier and higher levels are not covered, and the 1969 start year (the Vocational Training Act) is not stated in the source. Figures are read from a two-page PDF table, Bundesgebiet block. BIBB is a federal institute that runs the survey with the chambers, the bodies that register the contracts.
๐ฆ๐บAustraliasince 2018still operatingbears on Public-private collaboration
The New Payments Platform, whose payments the Reserve Bank of Australia settles one by one in central bank money through the Fast Settlement Service of its RITS system, settled a daily average of 3.6 million transactions in 2024/25, almost 19 percent more than in 2023/24, worth A$4.7 billion a day.
3,600,000 transactions per day, fast settlement service (npp) daily average transactions settled, 2024/25. 4,700,000,000 Australian dollars per day, fast settlement service (npp) daily average value settled, 2024/25. Reserve Bank of Australia, Annual Report 2025, retrieved 2026-10-02.
The move, and what it needed
The Reserve Bank set real-time payment objectives after its 2012 strategic review and built the Fast Settlement Service, while banks jointly financed, built and governed the switching platform and overlay services, so that every participating account could send and receive in seconds around the clock.
Preconditions
A central bank willing to set outcome targets and build settlement infrastructure but leave the customer-facing rail to industry.
A concentrated banking sector able to agree a shared build and its governance.
Exchange settlement accounts for participants so each payment can settle individually in central bank money.
Where else it has been used: Fast-payment rails built by industry under central bank objectives are common; Australia's case shows the slow part is migrating legacy bulk payments, not launching the rail.
Limits: The figure counts settlements in the Reserve Bank's Fast Settlement Service, which exists to settle NPP payments; it is a volume measure, not evidence of reach, outage resilience, fraud losses or how much activity is additional to existing direct entry. The annual report shows the Reserve Bank's half of the arrangement (settlement in RITS, fifteen institutions settling NPP payments directly) but does not say who built, owns or governs the NPP itself; it names NPP Australia Limited only in its glossary, so the industry side of the partnership rests on sources outside this document. The NPP is still the smaller rail for account-to-account payments: the report's Graph 2.3.4 shows bulk electronic clearing (BECS) still carrying more daily volume, and the industry's proposed decommissioning of BECS, which the report discusses, is not yet done. The 2018 launch year is not stated in the annual report. The number is on page 99 of the PDF. The rba.gov.au host refuses scripted requests (Akamai 'Access Denied'); the PDF was read through WebFetch at the same URL. The Reserve Bank reports on a service it owns and operates.
๐ณ๐ฌNigeriasince 2011still operatingbears on Public-private collaboration
NIBSS Instant Payment, run since 2011 by the interbank settlement company jointly owned by the Central Bank of Nigeria and the banks, grew from 2 billion transactions in 2020 to 11 billion in 2024, with values rising from US$457 billion to US$1.1 trillion, and had 57.7 million unique users.
The Bankers' Committee, the forum of the central bank and the deposit-taking banks, created a jointly owned switch in 1992 and used it to launch a 24/7 account-to-account instant payment service in 2011, with the central bank setting rules and fee caps and every bank connected, so a transfer reached any account in seconds.
Preconditions
A standing forum where the central bank and banks already took joint decisions and co-owned infrastructure.
A cashless-policy push from the central bank that gave banks a reason to route retail payments electronically.
A unique bank identity layer (the Bank Verification Number) to reduce fraud on instant transfers.
Where else it has been used: NIP is one of the older African instant payment systems and a reference in AfricaNenda's regional reviews; its joint ownership model differs from rails built by a central bank alone.
Limits: The publisher is a non-profit foundation, not a statistical agency, and the case study does not name the source of its 2020 to 2024 series ('NIP data for 2020 to 2024 shows'); it is presumably NIBSS operator data, but that is an inference. The Central Bank of Nigeria's e-payment statistics page reports channels (internet, mobile app, USSD transfers) without naming NIP, and NIBSS's own pages carry press reprints that mix NIP and all e-payment totals, so no official NIP series was citable. The figures are rounded (2 and 11 billion), the 2024 value is printed as $1.1 trillion in the text and $1,171,642 million in the chart, and the dollar values are converted at unstated rates in a period of steep naira devaluation, so the value growth says little in real terms. Growth was pushed by the 2023 cash scarcity and the pandemic as much as by the rail; the same page reports 69 percent of transaction volumes from men, a large inclusion gap, and the 57.7 million users are not active-user counts. Ownership and governance (jointly owned by the CBN and banks, a board of a CBN deputy governor and bank chief executives) and the 2011 launch are stated in the same document. Figures are in a PDF, on PDF page 8 (printed page 9) and the chart that follows.
๐ต๐ญPhilippinessince 2018still operatingbears on Public-private collaboration
InstaPay, the real-time automated clearing house organized under the Bangko Sentral ng Pilipinas' National Retail Payment System framework, in which banks and e-money issuers take part, processed 4.7 billion transfers worth โฑ11.6 trillion in 2025, up from 232.9 million in 2020, with 95 participating institutions in July 2026.
The central bank set an interoperability framework requiring retail payment providers to join industry-run automated clearing houses, while banks and e-money issuers governed the clearing rules through a self-regulatory payment management body, so any account or wallet could pay any other in real time.
Preconditions
A central bank willing to mandate interoperability but delegate rule-making to an industry body.
Mobile e-money issuers large enough to bring unbanked users onto the rail.
A national QR standard (QR Ph) to carry the rail to merchants.
Where else it has been used: The arrangement resembles Thailand's PromptPay and India's UPI in the corpus, with a stronger role for industry self-governance.
Limits: Transaction counts show a heavily used rail, not who uses it: a large part of the 2025 jump (230.9 percent in volume against 57.3 percent in value) is consistent with small e-wallet cash-ins and transfers, so value per transfer fell, and the bulletin does not split person-to-person use from wallet top-ups. The participant count (87 sender-receivers and eight receive-only) is computed from the bulletin's table. The bulletin lists InstaPay among the payment systems organized under the National Retail Payment System and lists its participants by type, but does not say who operates or governs the clearing house, so the industry side of the arrangement rests on sources outside this bulletin. The 2018 start year is not stated in the bulletin, whose annual table begins in 2020. The central bank's 2024 digital payments report puts digital at 57.4 percent of retail payment volume, so cash still carries a large share. The bulletin is a slide deck PDF: the annual table is on page 6 and the participant table on page 5. bsp.gov.ph refuses scripted requests (Akamai 'Access Denied'); the PDF was read through WebFetch at the same URL. The central bank reports on a framework it regulates.
๐ช๐ชEstoniasince 2002still operatingbears on Public-private collaboration
Estonia's state eID ecosystem, in which the state issues ID cards with certificates from a private trust-service partner and accepts the privately issued Smart-ID in its authentication service TARA, handled 27,094,512 Smart-ID authentications through TARA in 2023, 57 percent of the total, beside 22,351,282 by state ID card and Mobile-ID.
The state made a chip ID card with authentication and signing certificates compulsory and built common authentication and signing services, while a private certification provider issued the certificates, mobile operators issued Mobile-ID SIMs and banks relied on the same credentials, so one identity layer served public and private services and later absorbed a bank-led smartphone credential.
Preconditions
A compulsory national identity document with a single personal code.
A domestic trust-service provider and mobile operators willing to run certification as a business on state rules.
Banks and other large private relying parties that adopted the state credential early, giving citizens a reason to use it.
Where else it has been used: Estonia's eID model is widely cited; Smart-ID has spread to Latvia and Lithuania through the same provider, which makes the private side of the partnership the part that travelled.
Limits: The counts are authentications through one state service, TARA, not all eID use (banks authenticate customers directly, outside TARA). The split shows a shift from state-issued means to a private credential rather than growth of joint delivery: the white paper reports that ID card and Mobile-ID signatures fell from 94.8 percent of digital signatures in 2020 to 76 percent in 2022 as Smart-ID was enabled, and state-means signatures through DigiDoc fell from 8.6 million in 2021 to 6.9 million in 2023. A hostile reviewer could read that as the state losing control of its identity layer to a private provider. The document is a strategy white paper, not a statistical release, and draws on the Police and Border Guard Board and SK ID Solutions for issuance data. The 2002 start year (first ID cards) is not stated in the source. Figures are in a PDF on PDF pages 22 and 23 (printed pages 20 and 21), text and Tables 1 to 4.
๐ซ๐ฎFinlandsince 2017still operatingbears on Public-private collaboration
From May 2017 Finland's Trust Network obliged banks and mobile operators issuing strong electronic identification to let brokers resell their authentications at a regulated maximum price, cut to 3 cents in 2019, and mediated authentication events rose from an estimated 20 million in the second half of 2019 to about 60 million in the first half of 2020.
Parliament and the communications regulator turned bank and mobile-operator credentials into shared national infrastructure by law: providers must sell authentications to licensed brokers at a capped wholesale price, so any public or private e-service can accept every strong credential through one contract, and the state's own Suomi.fi service buys them the same way.
Preconditions
Near-universal online banking credentials already held by the population.
A regulator with the power to set and lower a wholesale price cap and enforce interconnection terms.
A state authentication service large enough to be the anchor customer for brokered identification.
Limits: The volume figures are the regulators' own estimates ('arviolta') built from provider returns that some providers could not separate, and the report says it could not trace a multi-year series; for 2019 it gives only a range of 40 to 80 million. The figures are from 2020 and published in March 2021, so they are dated, and the first half of 2020 overlaps the first months of the pandemic, which also pushed public and private services online. The collaboration was regulatory compulsion as much as partnership: the report says identification providers held at least a weak monopoly over their own credentials before price caps, and the network only worked from 2019 after repeated legislative fixes, with 2020 the first full year of comprehensive brokerage. Bank credentials made up an estimated 85 to 90 percent of mediated events in late 2019, so the system depends on a few banks. The government is building a state digital identity alongside it. Figures are on PDF pages 79 and 80 (printed pages 78 and 79), section 6.5.
๐ง๐ดBoliviasince 2021still operatingbears on Public-private collaboration
Bolivia's banks first offered QR payments under their own SIMPLE standard, and the Banco Central de Bolivia then made QR transfers interoperable across banks, cooperatives, housing lenders and development finance institutions with a universal standard compatible with the banks' codes, mandatory electronic channels and free transfers; immediate QR payments rose from 6 million operations in 2021 to 892 million in 2025, 86 percent of electronic fund transfer orders that year.
The banking system launched its own QR standard, and the Banco Central de Bolivia then built a universal QR standard compatible with it into its settlement module, required every financial institution to offer electronic channels and free transfers, and settled QR payments through its module and the private clearing houses, so each institution's app could pay any account in the system.
Preconditions
Prior interconnection of more than 90 percent of financial institutions for electronic transfers, which the BCB had already mandated
A central bank with regulatory power over every deposit-taking institution, including cooperatives and development lenders
Mobile banking apps already offered by most institutions, so the QR standard only had to be added
Where else it has been used: The same move as Thailand's PromptPay and Malaysia's DuitNow, made by a central bank in a lower-income, cash-heavy economy and accelerated by a currency shortage rather than by a planned cash-reduction policy.
Limits: Both numbers are data labels on Grรกfico 17 of a 72-page PDF (PDF page 35, printed page 29), with the 86 percent share on Grรกfico 18 on the next page; there is no machine-readable series. The central bank is the regulator, the standard-setter and the reporter, so this is the system's owner grading its own system. The collaboration mixes a private standard with a public mandate: the banking system built SIMPLE, the BCB built a compatible universal standard and imposed interoperability and free transfers, and private institutions and the clearing houses ACCL and UNILINK carried it out. The 2025 jump coincided with Bolivia's dollar shortage, fuel queues and rising inflation, which push people off cash; the value series is in nominal bolivianos and overstates real growth in a year of high inflation. A payment count does not show inclusion: the report gives no figure for first-time account holders. The series counts every immediate QR payment processed through the clearing houses and the BCB's settlement module, not only codes generated under QR BCB Bolivia: the banks' own SIMPLE QR standard came first, the BCB's 2021 vigilance report describes the BCB's compatible QR standard in its settlement module as due in the last quarter of 2022, and the BCB's QR page (https://www.bcb.gob.bo/?q=pagos_qr_bcb_bolivia) calls QR BCB 'un solo estรกndar universal compatible con todos los QR implementados por el sistema financiero'. The 2021 figure therefore predates the BCB standard, and the start year is the first year of the series, not the launch of either standard.
By May 2024 the Dominican government's Familia Feliz plan, in which private developers build low-cost homes and the state pays down-payment, tax and mortgage-rate subsidies, had built 3,979 homes and assigned RD$1,108 million in subsidies.
The Presidency, and since Decreto 114-25 the housing ministry, ran a national trust that certifies private housing projects and pays each buyer's down payment, sales tax and seven years of mortgage-rate relief, so private developers and banks build and finance low-cost homes that the state does not build itself.
Preconditions
Law 189-11 on mortgage development and trusts, which lets each project sit in a certified autonomous trust that protects buyers
A banking sector willing to lend long-term mortgages to lower-income buyers once the state covers the down payment and part of the rate
Domestic developers able to build several thousand units across provinces
Where else it has been used: A version of Mexico's and Brazil's developer-led subsidised housing models (Minha Casa Minha Vida), run at much smaller scale and paused after a change of ministry.
Limits: The numbers come from a ministry press note quoting the minister, not from a statistical series, and the government is reporting on its own flagship. 'Built' is not 'handed over': the same note says 704 apartments were delivered in January to May 2024 and 934 were waiting to be inaugurated. It also promised 9,600 deliveries in 2024, and no official page opened today gives a cumulative delivered total to check that against; Diario Libre (12 May 2026) reports the housing ministry putting Familia Feliz deliveries at 5,922 units since 2020, which, if right, means the 2024 promise was badly missed. That figure is press only and is not used. The programme's intake has stopped: FONVIVIENDA's news item of 26 April 2026 (https://fonvivienda.gob.do/2026/04/26/ministro-ito-bisono-reafirma-continuidad-del-programa-familia-feliz-entrega-144-apartamentos-en-laderas-del-este-vf/) says projects approved up to 30 December 2025 continue unchanged, that applications made up to that date are being re-evaluated, and that the focus is on finishing and handing over projects approved up to 2025, and it names no intake after that date; the status is eroded for that reason, although deliveries from the pipeline continue (890 apartments across Familia Feliz and Mi Vivienda in the first half of 2026, per FONVIVIENDA, not split by programme). A few thousand homes is small against the national housing deficit. The private side is paid by subsidy and mortgage, so this is a subsidy-for-supply model; banks still screen buyers through pre-approval, which excludes the informal poor. The start year is taken from Decreto 191-21, which created the FONVIVIENDA trust.
๐ฎ๐ฉIndonesiasince 2021still operatingbears on University-industry collaboration
Indonesia's Directorate General of Higher Education, Research and Technology funded 427 university-industry collaboration proposals through the Kedaireka Matching Fund in 2021, its first year, and 1,093 in 2022.
The higher-education directorate built an online matchmaking platform (Kedaireka) where university staff and firms register a partnership, then paid public money to match what the partner commits, so the state funds only collaborations a partner has already put resources into.
Preconditions
A national register of lecturers and higher-education institutions to verify applicants
A ministry budget line for competitive grants that can be paid to universities within the fiscal year
Firms willing to commit cash or in-kind resources before public money is released
Where else it has been used: Matching grants for university-industry projects are common (Uruguay's ANII Alianzas para la Innovaciรณn, ury-anii-alianzas, is one in the corpus); Kedaireka's variation is the platform that requires the partnership to be registered before a proposal can be filed, which let it scale to over a thousand funded projects in its second year.
Limits: The source is the ministry's own news release, and the cited URL is an Internet Archive copy: the original host (dikti.kemdikbud.go.id) no longer answers, and the successor ministry's copy on kemdiktisaintek.go.id sits behind a Cloudflare human-verification check that this protocol does not get past. The release is self-reporting, not an annual report or audit. The counts are funded proposals, not distinct firms or universities, and the release gives no count of partner firms. Partners can be businesses, industry or government bodies ('pihak bisnis, industri, maupun pemerintahan'), so not every funded proposal is a university-firm pairing. The figures show money granted, not research or commercial results. A separate ministry release of 15 September 2022 (same archive) says 5,407 proposals were submitted in 2022 against 1,231 in 2021, so roughly one in five 2022 proposals was funded; it also reports Rp 11.2 trillion of combined ministry and partner 'collaboration funds' for 2022, which mixes partner and possibly in-kind contributions and is not used here. No funded counts for 2023 to 2025 were found at an open official source. The programme was renamed Dana Padanan for 2024 and passed to the new Ministry of Higher Education, Science and Technology after the October 2024 restructuring; the programme site (backoffice.kedaireka.id) lists a 2025 call and a July 2025 award announcement, which is the basis for 'operating', but not the scale of those rounds, and the 2025 budget cuts to the ministry may have reduced it.
๐ป๐ณVietnamsince 2020operating below its peakbears on Public-private collaboration
Vietnam built three component projects of the 2017-2020 North-South eastern expressway phase, 178 km at about VND 25,343 billion, as BOT public-private partnerships, then financed all ten 2021-2025 component projects on the Hanoi to Ho Chi Minh City stretch as public investment, and in June 2025 the Ministry of Construction reported that it had asked the Prime Minister to keep widening the corridor with public investment money, leaving PPP as a later option once tolls were collected.
The transport ministry split the corridor into component projects and offered some of them to BOT investors, with state capital covering more than half of each project's cost and tolls repaying the private share. For the next phase the National Assembly financed every section from the budget, with the state to recover its outlay through tolls, which took private investors out of the corridor.
Preconditions
A ministry able to package a long corridor into sections a domestic investor can finance
State capital support large enough to make toll revenue cover the private share
A budget and a legislature willing to finance the corridor directly when investors do not close, which is also what ended the partnership route
Where else it has been used: Viability-gap BOT on a national trunk road shows how far a state will go before it simply pays for the road itself: here, once public money was available and the schedule was political, the partnership share went to zero.
Limits: The figures are the Ministry of Construction's own account, relayed by the Government's e-newspaper; no audit body has published them. The zero is read off the page, not printed as a number: the ministry lists 11 public-investment component projects, 10 of them from the 2021-2025 phase, against 3 BOT projects from 2017-2020. The full 2021-2025 phase was 12 component projects of about 729 km at VND 146,990 billion, all assigned to public investment by National Assembly Resolution 44/2022/QH15 (scanned PDF at datafiles.chinhphu.vn/cpp/files/vbpq/2022/01/44-qh.signed.pdf: 729 km in 12 component projects as public investment on page 1, Article 2.2; VND 146,990 billion on page 2, Article 2.5). The peak was itself a retreat from a plan. The Ministry of Transport listed eight sections for PPP in December 2017 (Decision 3535/QD-BGTVT, baochinhphu.vn, 26 December 2017), and only three reached contract, out of 11 component projects over 654 km. Most of the money in those three came from the state. The Government portal of 28 April 2024 puts Cam Lam-Vinh Hao at VND 8,925 billion with about 5,139 billion of state capital, and Dien Chau-Bai Vot at 11,157 billion with about 6,067 billion. The same page records the opening of Cam Lam-Vinh Hao and of Dien Chau-Bai Vot as far as National Highway 46B, with Nha Trang-Cam Lam already in service; the Ministry of Transport announced the full Dien Chau-Bai Vot opening for 30 June 2024 (baochinhphu.vn, 28 June 2024). Construction started in 2020-2021, per the Ministry of Transport on the 2024 page. The retreat is a choice as much as a market failure: the same 2025 page says the ministry's own study found BOT widening highly feasible, and it chose public investment, citing toll levels and the push to disburse public funds. This is one corridor. The record does not show whether PPP procurement fell on other expressways. The question is still open: a pre-feasibility report of June 2026 (baochinhphu.vn, 6 June 2026) proposes public investment for 15 component projects and PPP with state support only for the three existing BOT sections. That is a proposal, not a decision.
๐บ๐ฆUkrainesince 2023still operatingbears on Public-private collaboration
Ukraine's government set up the Brave1 cluster in 2023 under the Ministry of Digital Transformation. By September 2026 it had awarded more than 1,000 non-repayable grants worth UAH 6.2 billion to private defence-technology developers, whose products it tests with the armed forces and moves toward serial production.
The Ministry of Digital Transformation, with the Ministry of Defence and the General Staff, set up a state cluster. It registers private developers, funds them through tiered grants, arranges battlefield testing and military feedback, and steers money toward priority areas the military defines.
Preconditions
An acute, shared national objective that private firms and the state both accept
A large pool of small engineering and software firms able to iterate quickly
A military willing to test immature products and return field data to developers
Limits: The source is the cluster reporting on itself in a news release, with a minister's quote. Neither the 6.2 billion nor the grant count is broken down by year or audited. The figures measure money and grants going out, not what was delivered: they do not show how many funded products reached serial production or the front. The impact claims on the page, a 1.7-point gain in technology readiness and 79 percent of firms saying they had no alternative, come from a survey of defence-tech manufacturers in a study by the BRDO Office for Effective Regulation and are self-reported. Grants are one-way subsidies, so the record evidences public funding of private developers plus joint testing, which is a thin reading of joint delivery. The standard grant programme accounts for UAH 2.3 billion of the total; most of the rest, UAH 3.7 billion, went to 34 competition projects for missiles, air defence and autonomy, so the money is concentrated in a few large awards. Wartime urgency and secrecy limit what can be verified.
๐ท๐บRussiasince 2010still operatingbears on University-industry collaboration
Under Government Resolution 218 of 9 April 2010, Russia's Ministry of Science and Higher Education subsidised 41 joint firm-university complex projects in their R&D stage in 2024, from the 14th and 15th selection rounds. Federal funding was 2.708 billion rubles, and firms put in 4.14 billion rubles of their own (156 percent of the budget funds). The ministry reports 161 technologies developed and handed over to production by January 2025.
The Russian government subsidises the manufacturer rather than the university. A firm wins a competitive subsidy to commission R&D from a Russian university or state research institute, co-funds it at least one for one, and must build the production line and sell the product from its own resources.
Preconditions
Manufacturers with enough capital to match the subsidy and build production
Universities and state institutes with applied R&D capacity that firms are willing to contract
A federal budget line sustained across 15 selection rounds since 2010
Limits: The ministry is reporting on its own programme in its own annual report, and none of it is audited. The figures are counts of subsidised projects and money, not a measure of how much research firms and universities do together. The 156 percent co-funding largely reflects the rule itself, which requires each firm to match the subsidy at least one for one. The report's cumulative total of 133 projects cannot be reconciled, from this document, with 15 selection rounds since 2010, so the programme's scale over time is unclear. Whether output or new projects per year have grown or shrunk is not shown. The '161 technologies transferred' figure is self-defined and does not show sales. The terms of the scheme and the 2010 to 2020 funding total come from the operator's site, which a private contractor runs. That site says that from 2025 the projects run inside a departmental project of the state programme for scientific and technological development. That reorganisation could change the scheme, and this record does not track it. The figures are read from a 3 MB Russian-language PDF. The figures say nothing about the effect of post-2022 sanctions on the firms involved.
๐ช๐ฌEgyptsince 2015delivered and closedbears on Public-private collaboration
Egypt's Ministry of Electricity and Renewable Energy took private sector-owned renewable generation from 0 MW in October 2015 to 1,500 MW at financial closure by the end of FY2017/18 under its feed-in tariff programme, with about US$2 billion of private investment, much of it in the Benban solar park.
The Ministry of Electricity and Renewable Energy legislated a renewable energy law in 2014, set a feed-in tariff by cabinet decree, licensed private developers through the regulator and allocated land in one desert park, Benban, so many small IPPs could share grid connection and permitting.
Preconditions
A power deficit (5,540 MW in FY15) that made new private capacity politically urgent
A state offtaker whose payment risk international guarantors were willing to cover
Desert land under state control with high irradiation, allocated as one park
Limits: The indicator was softened mid-programme: DPF1 asked for 1,500 MW commissioned, and DPF2 redefined it as 1,500 MW at financial closure, so the number shows contracts financed, not plants generating, and the ICR's statement that Benban was then the world's fourth-largest solar plant sits in a footnote that says it 'promises' capacity. Most developers and lenders were foreign, and the deal flow leaned on IFC debt (US$653 million) and MIGA risk insurance (US$210 million), so how much of the collaboration was the Egyptian state's own capacity and how much was World Bank Group de-risking cannot be separated. The figure does not show the tariff cost carried by the state offtaker, or that the feed-in tariff was later replaced by other procurement. The source is the World Bank's own evaluation of operations whose results framework included this indicator, and the number is read from a PDF results table.
๐บ๐ฟUzbekistansince 2019still operatingbears on Public-private collaboration
Uzbekistan's government signed its first competitively tendered solar PPP in November 2019 at USยข2.679 per kWh, the 100 MW Navoi plant began commercial operation in December 2021, and by January 2024 operational solar PV capacity in the country stood at 1,497 MW, with agreements for a further 5,677 MW of solar and 5,100 MW of wind by 2030.
The government, through the energy ministry and a new PPP agency, ran a World Bank Group Scaling Solar tender with standard contracts and a sovereign support agreement, and used the record low tariff it produced to raise its renewable target and repeat the tender model.
Preconditions
A PPP law and a PPP agency created in 2019 to sign and manage the contracts
Multilateral payment guarantees covering a state offtaker with below-cost-recovery tariffs
An electricity tariff reform commitment that made long-term offtake credible to bidders
Limits: The 1,497 MW is all operational solar, and the ICR's own table of operating projects includes 400 MW from Gezhouba procured non-competitively, so not every megawatt came through the competitive PPP route. One foreign developer, Masdar, holds most of the competitive capacity listed (Navoi, Scaling Solar 2, Sherobod), so the collaboration is concentrated. Most deals carry World Bank Group, ADB or EBRD payment guarantees, so the state and multilateral lenders carry offtake risk and the number cannot separate Uzbek institutional capacity from multilateral de-risking. Capacity is nameplate, not generation, and solar remains a small share of a system of roughly 14 GW of mostly gas-fired plant. The 5,677 MW and 5,100 MW are agreements, not delivery. The source is the World Bank evaluating its own guarantee operation.
๐ฆ๐ดAngolasince 2022still operatingbears on Public-private collaboration
Angola's Ministry of Transport signed a 30-year concession with Lobito Atlantic Railway, in which Mota-Engil holds equity, on 4 November 2022 to rehabilitate, operate and maintain about 1,300 km of the Benguela line from Lobito to Luau and the Lobito mineral terminal. By September 2026 the concessionaire was operating the corridor with about 945 workers, 529 of them transferred from state rail and port operators.
Angola's Ministry of Transport tendered the freight operation of a state railway to a private consortium for 30 years, keeping passenger service public and moving state rail and port staff into the concession company.
Preconditions
Mineral traffic from the DRC and Zambia large enough to attract a commodity trader as operator
Political risk cover and development finance (MIGA, DFC, AfDB) to make a 30-year private commitment bankable
Limits: Neither figure measures delivery. 1,300 km is the length of an asset handed over, and the press release does not say what was rehabilitated, what freight moved or what the concessionaire invested. The publisher is a guarantor promoting its own transaction, and the April 2024 summary that dates the contract was written by the concessionaire. 2022 is the contract date, and the source does not say when operations passed to the concessionaire. The concession is four years into 30, so it shows that the state handed over operations, not that the partnership delivered. Passenger service stays with the state operator. The corridor is driven by DRC and Zambian copper exports and by US and EU strategic finance, so the national objective it serves is as much foreign as Angolan. It is one contract, and the construct asks for the frequency and scale of public-private delivery across the country.
๐ฒ๐ฟMozambiquesince 2004dismantledbears on Public-private collaboration
Mozambique's government concessioned the Sena and Machipanda railways for 25 years in 2004 to a joint company in which the state rail enterprise CFM held 49 percent; the Sena line reopened to Moatize more than two years late, carried 266,000 tonnes against a target of one million, and the government terminated the concession on 7 December 2011.
The government of Mozambique, through CFM, handed rehabilitation and operation of two state railways to a private concessionaire in a 51 to 49 joint company, financed largely by an IDA credit, and took the lines back when rehabilitation ran late and traffic did not materialise.
Preconditions
A state rail enterprise able to supervise a concessionaire, which a 49 percent shareholding in the same company undercut
A traffic forecast tied to coal mines whose own concessions were being negotiated in parallel by another ministry
A dispute mechanism or independent arbiter, which the concession lacked
Limits: The source is the World Bank's own completion report on a project it financed, so the lender is grading a failure it shares, and the independent review of that report found its cost tables mutually inconsistent: the text claims a cost overrun of more than US$50 million while the annex shows actual cost below appraisal, which is why no cost figure is used here. The ICR also dates the termination both to September and to December 2011. The 266,000 tonnes is coal moved in the months after the line opened in August 2011, against a target set for 2009, so it measures lateness as much as capacity. The concession covered two lines in the centre of the country, not the national network. The termination ended the partnership but not the asset: CFM kept operating the rehabilitated Sena line, and later coal traffic on it is not in this record. Blame is contested: the report faults the concession design, the state's 49 percent stake and ineffectual Bank supervision as well as the private operator.
๐ฌ๐ญGhanasince 2018still operatingbears on Public-private collaboration
Ghana structured the Sankofa offshore gas development as a public-private deal in which Eni and Vitol invested and the state oil company GNPC bought the gas under take-or-pay terms backed by an IDA payment guarantee, mobilising US$4,653 million of private capital and bringing first gas to power plants in 2018.
The Government of Ghana and GNPC contracted to buy domestic gas from a privately financed offshore field on take-or-pay terms, and the World Bank guaranteed those payments so that private sponsors and banks would commit capital to a sub-Saharan gas project.
Preconditions
A proven offshore gas discovery and an international operator willing to develop it
A multilateral guarantee to cover the sovereign's payment risk
A power sector able to pay for the gas, which Ghana's state utilities could not do on time
Limits: The capital was private and the build was done by the Eni-Vitol consortium; Ghana's institutional move was the deal structure and the offtake, and that side performed worst. GNPC could not pay: arrears reached US$192 million by March 2020 and the sponsors drew US$191.6 million on the letter of credit backed by the IDA guarantee, so the public half of the partnership needed a World Bank backstop within two years of first gas. Off-take in 2018 and 2019 was a fraction of the 170 mmscfd the state contracted to take or pay for, because the pipeline interconnector to the eastern market was late and demand in the west was too small, which means Ghana paid for gas it did not use. The source is the World Bank's own completion report on its own guarantee, written in 2020, and its figures are not published as a machine-readable series; the source stops in mid-2020. One deal, however large, says little about the frequency of public-private delivery the indicator asks for.
The Ivorian government let a third Abidjan lagoon crossing as a toll-bridge concession to SOCOPRIM, the project company formed in 1997, whose investors include Bouygues Travaux Publics, after a 1996 tender. It funded initial works from October 2011 after the conflict. Private equity and loans, insured by MIGA guarantees of $145 million, built the 1.5 km bridge and 6.7 km road at a cost of about $282 million, and the bridge opened on 16 December 2014.
The government kept a 1990s concession alive through a decade of conflict and then restarted it in 2011, funding the first works itself. A political-risk guarantee from MIGA let a French contractor and development lenders put equity and debt into a toll bridge in a post-conflict country.
Preconditions
A concessionaire and sponsor willing to hold a dormant contract through 15 years of instability
Multilateral political-risk cover that lenders accept as a substitute for sovereign creditworthiness
Urban traffic dense enough to carry tolls, which only Abidjan offers in the country
Limits: The source is a two-page promotional brief from the World Bank Group, whose agency MIGA guaranteed the deal. It is an interested party describing its own transaction. The cost is given as approximate, with no baseline budget, so the record shows no cost or schedule performance against plan. The 100,000 vehicles a day is a forecast ('will use'), not observed traffic. No toll revenue, traffic or state guarantee and payment figures are published here, so the fiscal risk the state kept (minimum revenue guarantees, its own up-front works) is not visible. The brief is from 2015. This is one PPP. It does not show how often or at what scale Cรดte d'Ivoire delivers national objectives jointly with the private sector, which is the construct, and the 18 years from tender to opening were spent mostly in conflict, not in delivery.
๐ฟ๐ผZimbabwesince 2019still operatingbears on Public-private collaboration
Zimbabwe's government contracted the Zimborders consortium under a build-operate-transfer concession to rebuild the Beitbridge Border Post with a US$300 million investment; works began in March 2019 and the President commissioned the modernised post in September 2022.
The Ministry of Finance concessioned the rebuild of the country's busiest port of entry to a private consortium under build-operate-transfer, so the consortium financed construction and recovers it from user fees before handing the facility back.
Preconditions
A border post whose traffic volume can repay private capital through user fees
A government willing to cede fee revenue for a fixed concession term
Financial closure achievable despite sovereign arrears that cut the state off from most concessional lenders
Limits: The US$300 million is a headline project value stated in a revenue authority news release during construction, not an audited outturn, and the release itself credits 'the government' with the investment, which blurs how much the concessionaire financed. The concession's term, tariff and risk allocation are not given in the official sources behind this record. One concession, however large, says little about the frequency of public-private delivery, which is half of what the indicator asks; the Treasury's own text shows the model only 'being considered' for other border posts. The record has no official series on clearance times or traffic before and after, so it shows that the facility was built and commissioned, not that it performs better.
๐ง๐ฎBurundisince 2005dismantledbears on Public-private collaboration
Burundi had run its coffee value chain through a public regulator, ARFIC, beside private-led bodies: InterCafe, whose members were producers, millers, roasters and exporters, and the growers' consortium CNAC with about 109,000 members. The government named all three as joint implementing agencies of the Coffee Sector Competitiveness Project. In January 2020 it replaced them by decree with ODECA, an office under state control. The project closed early in 2021, and 44 percent of the grant was cancelled.
Over two decades the Government of Burundi handed coffee sector governance to an independent public regulator alongside a private interprofessional association and a growers' consortium, and wrote them into a donor project as co-implementers. In January 2020 it took the value chain back by presidential decree under a new state office, without consulting the partners.
Preconditions
Private and cooperative operators organised enough to run washing stations, sector data and an interprofessional body
A state willing to share control of the main export crop, a commitment that here rested on donor conditionality more than on domestic consensus
Legal footing that can survive a change of government, which a decree-based arrangement did not have
Limits: The only source is the lender's own completion report, and it is written from the side that lost the argument: the World Bank calls the 2020 decision unilateral and made without advance warning. The 44 percent is the share of one donor project cancelled. It measures what the dismantling cost that project, not how much public-private delivery Burundi lost. The ICR publishes no series on the scale of the private-led arrangement, such as the private share of washing stations, InterCafe's budget or the volumes it governed. The 109,000 CNAC members is a membership count at appraisal, not a delivery figure. The liberalisation began under structural adjustment in the 1980s and was pushed by donors (IFC advised on the third phase), so how far the arrangement was Burundi's own institutional choice is contestable. The 2005 start year rests on a decree the ICR names as the one that reformed the sector. The ICR does not say in which year InterCafe itself was created. Nothing here shows whether coffee output, quality or grower prices fell after ODECA took over. The ICR reports the share of AA/A/FW+ grade coffee rising from 59 to 66 percent, 70 percent of the incremental target. No data exists for the second quality indicator, the specialty share of exports, because ARFIC was removed and ODECA provided none.
๐ฒ๐บMauritiussince 2014still operatingbears on University-industry collaboration
The Mauritius Research and Innovation Council's Collaborative Research and Innovation Grant Scheme, which co-funds projects led by a local company with a local academic or research institution, had completed 17 projects and had 14 more under way as at 30 June 2020, with 61.5 million rupees of private contribution to the completed ones.
The national research council pays roughly half the cost of an R&D project only when a registered company applies with a local university or research institute as partner, so public money follows a firm's own commitment.
Preconditions
A national research and innovation council with a competitive grant budget
A university and research institute base able to take on contract work for local firms
Firms large enough to co-fund half of a multi-million rupee project
Limits: The council reports on its own scheme. The figures sit in a PDF on page 23 of 66 (printed page 22); the portfolio table on the same page gives a different split for the reporting year alone (10 ongoing and 3 completed CRIGS initiatives), so the cumulative and annual counts must not be mixed. 2019-20 is the latest annual report the council has published: its publications page lists the 2020-21 to 2023-24 reports as awaiting audit or in process, so nothing after June 2020 can be checked against an annual report. The scheme page (mric.mu/collaborative-research-and-innovation-grant-scheme) still describes the scheme and lists 18 funded projects, all marked completed, with no ongoing project and no call dates, so the scheme may have funded little or nothing since 2020; 'operating' rests on the page still offering it and is the weakest part of the record. Thirty-one projects over six years is small even for a country of 1.3 million people, the counts are projects rather than distinct firms, and project value includes the firm's in-kind contribution. The start year is the first call for proposals, June 2014, stated in the council's Annual Report 2014, not at the cited URL. Nothing here shows the research results or commercial outcomes.
๐ฒ๐บMauritiussince 2019still operatingbears on Public-private collaboration
The Bank of Mauritius launched the Mauritius Central Automated Switch in August 2019 as a central hub through which banks and non-bank payment providers route instant payments, and its Instant Payment System grew from 126,000 transactions in FY2019-20 to 19.4 million transactions worth Rs63.0 billion in FY2024-25, with 11 banks and 4 non-banks participating.
19.4 million transactions, maucas instant payment system transactions in the financial year, FY2024-25. 126,000 transactions, maucas instant payment system transactions in its first financial year, FY2019-20. Bank of Mauritius, Annual Report for the year ended 30 June 2025, retrieved 2026-10-02.
The move, and what it needed
The central bank built and ran a national switch that every bank and licensed non-bank could connect to, making transfers between any two accounts settle in seconds, then cut fees and raised limits to push use.
Preconditions
A central bank with licensing power over banks and non-bank payment providers
A Real Time Gross Settlement system the switch could settle into
A banking population already using mobile apps
Where else it has been used: The same model as Bolivia's QR interoperability (bol-qr-bcb) and India's UPI, run on a small island economy where the central bank, not a bank consortium, owns the switch.
Limits: The central bank owns and operates the switch, sets the fees and reports the volumes, so this is the system's owner grading its own system. The FY2024-25 figures are in the text on PDF page 88 of the 2025 annual report; the FY2019-20 baseline and the August 2019 launch are in the Annual Report for the year ended 30 June 2020 (https://www.bom.mu/AnnualReport/2020/BOM_June_2020.pdf, PDF pages 129 to 130), where only seven banks and one non-bank were connected. The 95.4 percent growth in FY2024-25 followed the Bank's own 2024 rule changes (interchange revisions, removal of person-to-person charges, a higher limit), so it measures a policy push as much as adoption. Nineteen million transactions a year in a country of about 1.3 million people is still modest beside card and cash use; the report gives no per-capita or merchant figure. The collaboration is a public switch with private participants, not a contractual PPP.
๐บ๐ฌUgandasince 2005dismantledbears on Public-private collaboration
Uganda ran most of its electricity distribution through private concessions, with Umeme selling about 96 percent of grid energy to end users in 2024, but a Cabinet directive of October 2022 approved the non-renewal of all expiring concessions, and in 2024 the state utility UEDCL took over the assets of two private distributors whose licences were not renewed.
The government concessioned the national distribution network to a private operator for 20 years under ERA regulation, then the Cabinet decided not to renew any expiring concession and folded each private distributor back into the state company as its licence ran out.
Preconditions
A regulator able to set tariffs and verify investment over a long concession
A state utility able to absorb staff, customers and assets at handover
A settled buyout value for the operator's unrecovered investment, which was disputed
Where else it has been used: Uganda also took back Eskom Uganda's 20-year generation concession at its end in 2023 (per an ERA annual report, era.go.ug download wpdmdl=7485); Lesotho's hospital PPP and other African utility concessions show the same pattern of a long PPP ended at expiry rather than renewed.
Limits: The figures are in a PDF (the 96 percent and Table 6 on PDF page 21; the Cabinet directive and the KIL and KRECS takeover on PDF page 24). The report covers 2024, when Umeme still operated; it does not record the end of Umeme's own concession, which ran to 2025. That the state utility took over Umeme's area is documented by UEDCL itself: its '100 Days in the Driving Seat' report (https://www.uedcl.co.ug/wp-content/uploads/2025/07/100DaysInTheDriving-Seat-.pdf, page 1) says that on 1 April 2025 UEDCL began its national mandate of electricity distribution; ERA's tariff schedules for Umeme on its site also stop after January to March 2025. The started year is the start of the Umeme concession as commonly dated, not stated in this report. The concession's record is contested: ERA reports Umeme's losses at 16.0 percent in 2024 against a 13.24 percent target, and the decision to end private distribution was a policy choice by the Cabinet, so 'dismantled' records a decision, not a collapse. The report gives no buyout amount; an ERA annual report on the same site (wpdmdl=7485) lists 'no clear asset registers to determine buyout amount for UMEME' as a risk of the concession's expiry.
๐ฒ๐ฑMalisince 2003dismantledbears on Public-private collaboration
Mali and Senegal concessioned the 1,288 km Dakar to Bamako railway to Transrail in 2003 for 25 years, terminated the concession on 5 December 2015 after it failed, and rail traffic, which at its peak had carried about 1.5 million tons of freight a year, stopped completely in 2018.
Two governments handed a run-down binational railway to a private consortium in a single full concession covering track and operations, keeping minority stakes; the operator could not fund maintenance from traffic revenue, service degraded, the states terminated the contract and the line stopped.
Preconditions
A track and rolling stock base good enough for operating revenue to fund maintenance, which was missing
A concession design that separated infrastructure funding from train operations, which was missing
Two governments able to coordinate oversight of one operator across a border
Where else it has been used: The PID places it in a wave of full rail concessions in Sub-Saharan Africa and draws the lesson that infrastructure funding should be separated from train operations, which is the model the follow-on corridor project proposes.
Limits: The facts are in paragraphs 7, 9 and 16 of a concept-stage document (pages 4 to 6 of 14 in the plain-text version); the zero is the document's statement that traffic 'stopped completely in 2018', not a published tonnage, and the 1.5 million ton peak has no year, so it may predate the concession. The World Bank supported the wave of African rail concessions this one belonged to and is preparing the follow-on corridor project, so the publisher is not neutral about why it failed; the PID also says both governments ignored the Bank's advice to restructure the concession as a lease. The concession was binational: Senegal is an equal party and the record is filed under Mali, where Transrail was headquartered, but it describes both states' collaboration. The PID's diagnosis is shared blame: a 'full concession' design that made the operator pay for track and trains from revenue that could not cover both, decades of underinvestment by both governments before 2003, and road competition after the 2002 Cote d'Ivoire crisis sent Mali's trade through Dakar. After termination both governments paid staff salaries through an interim company (Dakar Bamako Ferroviaire) with a heavy fiscal cost.
๐ฆ๐นAustriasince 2007still operatingbears on University-industry collaboration
Austria's COMET programme, run by the FFG for the federal ministries and co-funded by the Lรคnder, had approved 81 competence centres and K-projects with eligible costs of 1,669.80 million euros by 30 March 2017, in which about 450 scientific partners and about 1,000 company partners cooperated, according to the Court of Audit.
Two federal ministries fund, through their research promotion agency, multi-year competence centres in which universities and firms run a joint research programme, with the Lรคnder adding about half the federal share and companies paying roughly 40 to 60 percent of costs.
Preconditions
A federal research promotion agency able to run multi-year centre funding and evaluations
Regional governments willing to co-fund centres located in their territory
Firms willing to commit cash to a shared pre-competitive programme for up to eight years
Where else it has been used: COMET succeeded Austria's 1990s Kplus and Kind/Knet competence centre programmes, and sits in the same family as Sweden's competence centres (swe-kompetenscentrum) and Norway's SFI scheme.
Limits: The figures sit in an 86-page PDF: Table 2 on printed page 33 (PDF page 35) gives the 81 approved centres and projects and 1,669.80 million euros of eligible costs (federal funding 522.06 million, Lรคnder 262.79 million), and the partner counts are on printed page 24 (PDF page 26), both rounded ('rd. 450' and 'rd. 1.000'). The counts are approvals since 2007, not centres running in a given year, and the partner figures count organisations once without saying over what period; the court notes elsewhere about 970 company participations including multiple participations, so 1,000 is not a count of distinct firms. Scientific partners include non-university research bodies, not only universities. The audit period ended in 2016 and the numbers are nine years old; the FFG programme page said on 2026-10-02 that the network currently comprises 23 COMET centres, 17 COMET modules and 15 COMET projects, with no reference date, which supports 'operating' but is not the metric. The court audits spending and compliance, not research output, and it criticised conflicts of interest in centre governance.
๐จ๐ฟCzechiasince 2012delivered and closedbears on University-industry collaboration
The Technology Agency of the Czech Republic funded 34 competence centres under its Centra kompetence programme from 2012, joining 334 participants, 115 research organisations, 118 SMEs and 101 large firms, against a planned 31 centres and 248 participants.
The national technology agency funded long-term consortia of research organisations and firms, each at least one research organisation and several companies, to run shared applied research programmes with a minimum private share of costs.
Preconditions
A national agency for applied research with its own programme budget
Universities and Academy of Sciences institutes able to lead multi-partner consortia
Industrial firms, including large manufacturers, willing to co-fund for up to eight years
Where else it has been used: The programme copied the competence-centre model used in Austria (aut-comet) and Sweden; its successor, the National Competence Centres, kept the consortium rule.
Limits: The agency evaluated its own programme, and the figures sit in a 101-page PDF: Tables 1 and 2 on printed pages 23 to 24 (PDF page 25), with parameters updated to 30 June and 15 September 2015 from the agency's IS Patriot and the national R&D information system. The counts are of supported centres and participants at contract signature, not of active collaboration in a given year, and a participant is an organisation, so one university can count several times across centres. The evaluation itself says the main objective of developing cooperation between research organisations and firms proved 'nepลรญliลก dostateฤnรฉ' (not sufficiently met), with results showing it only in part, and that the reported 25.9 billion koruna of product sales is dominated by one centre and was not verified. Funding received by the centres in 2012 to 2014 was 1.53 billion koruna. The programme ran to 2019 and was followed by Nรกrodnรญ centra kompetence; this record does not cover the successor.
๐ฉ๐ฐDenmarksince 1970still operatingbears on University-industry collaboration
Denmark's innovation fund (Danmarks Innovationsfond, formerly Innovationsfonden) awarded 138 Industrial Researcher grants in 2025, from 252 applications and for 147.4 million kroner, each funding a researcher employed by a company and enrolled at or supervised by a university.
The state pays part of a company's salary cost for a researcher it employs and the university's supervision cost, so that a doctorate or postdoc is carried out on a company problem with joint supervision.
Preconditions
Universities willing to supervise doctoral candidates who spend most of their time in a firm
Firms with research capacity able to co-fund a three-year salary
A national fund with an open, recurrent call and a mixed academic and business assessment panel
Where else it has been used: Industrial doctorate schemes elsewhere in the corpus include France's CIFRE (fra-cifre), Spain's Doctorados Industriales (esp-doctorados-industriales) and Poland's Doktorat wdroลผeniowy (pol-doktorat-wdrozeniowy); the Danish one is the oldest.
Limits: The fund reports on its own programme, and the page states only 'All figures are from 2025' with no earlier years, so no trend can be read from it. The page was read through a fetch service; the fund's web firewall refused direct script requests and the annual report PDFs, so the multi-year series could not be checked. A grant is an award at the start of a three-year project, not a completed doctorate, and the count mixes Industrial PhDs with Industrial Postdocs and does not say how many host organisations are public bodies rather than firms. The start year is the commonly cited 1970 founding of the Danish industrial PhD scheme and is not on the cited page. Applications fell to 252 in 2025, and a search snippet gave 140 grants and 149.6 million kroner for 2024, which we could not open at source.
๐ฎ๐นItalysince 2022delivered and closedbears on University-industry collaboration
Italy's Ministry of University and Research funded 1,708 company co-financed innovative doctoral scholarships in the 38th doctoral cycle (academic year 2022/2023), against the 5,000 that decree 352/2022 had allocated for that year, and cut the measure's target from 15,000 scholarships and 450 million euros to 6,000 scholarships and 360 million euros.
The ministry offered universities EU recovery money for doctoral scholarships on condition that a company co-financed each one and hosted the candidate, then cut the target and raised the public share when firms took up far fewer places than planned.
Preconditions
Universities accredited to run doctoral programmes with industrial partners
Firms with research functions willing to pay half a three-year scholarship
A temporary EU recovery budget large enough to fund thousands of scholarships at once
Where else it has been used: Industrial doctorate schemes funded from regular budgets (dnk-erhvervsforsker, fra-cifre) grew over decades; the Italian attempt to buy thousands in one year ran into the limit of firm demand.
Limits: The 1,708 is our sum of the 'Borse assegnate' column over 533 rows of Tabella A (PDF pages 3 to 21), which prints no total; it agrees with the decree's Article 2 total of 85,400,000 euros at 50,000 euros a scholarship. mur.gov.it refused script requests, so the cited copy is the Internet Archive's. The 15,000 target, the 450 million euro original allocation and its redetermination to 360 million euros are in the decree's preamble; the 6,000 figure is on the ministry's measure page (archived 13 June 2026, https://web.archive.org/web/20260613095831/https://www.mur.gov.it/it/pnrr/misure-e-componenti/m4c2/investimento-33-introduzione-di-dott-innovativi-che), not in this decree. The 5,000 allocated by decree 352/2022 covered the 2022/2023 year across courses accredited under both the XXXVII and XXXVIII cycle rules; we read all of them as XXXVIII cycle scholarships, which the decree does not state in so many words. A scholarship counts as assigned when funded, not when the doctorate is completed, and firm co-financing is a condition of the measure rather than a separately counted payment. The shortfall against the 5,000 and the cut in the target are a plan that was not met, not a loss from a delivered peak, so the record is not classed as a reversal: it shows that far fewer firms co-funded than the ministry budgeted for, and the decree's response was to raise the ministry's unit contribution from 30,000 to up to 60,000 euros per scholarship (capped at 20,000 euros extra for this cycle), which shifts cost from firms to the state. Later cycles (decree 117/2023) are not counted here, and the measure ended with the recovery plan in 2026.
๐ฑ๐นLithuaniasince 2012delivered and closedbears on University-industry collaboration
Lithuania's Agency for Science, Innovation and Technology (MITA) funded 815 innovation-voucher projects in three calls from 2012 to 2014, in which small firms bought applied research, experimental development or feasibility services from universities and research institutes, and 776 of them were completed successfully.
The innovation agency gave small firms a fast, small grant that could only be spent on research services from accredited universities and research institutes, to make a first contact between the two cheap and quick.
Preconditions
A list of research services that universities and institutes are able and allowed to sell
A simplified application procedure so very small firms can apply
Structural fund money willing to finance many small grants despite the administrative cost
Where else it has been used: Innovation vouchers began in the Netherlands (Limburg, 1998) and spread across Europe; the Lithuanian evaluation found the EU-funded version carried higher administrative costs than the earlier nationally funded pilot of 2010 to 2012.
Limits: The report was commissioned and paid for by the ministry and written by a consultancy, and the copy that opens is held by SIPER, an academic repository of innovation policy evaluations, not by the ministry; the counts themselves come from MITA's activity reports for 2012 to 2015, which the evaluation cites. The figures are on printed page 19 (PDF page 19) of a 112-page PDF, with the call-by-call table (Table 6: 185, 174 and 456 funded applications). Each voucher was small, up to 5,792.4 euros at up to 80 percent public funding, so a project is a short service purchase rather than joint research, and 104 firms were funded in two calls, so 815 projects are fewer distinct firms. The third call relaxed the private contribution and favoured firms with no prior research contact, which raised volume. The vouchers were EU structural fund money and the measure ended with the 2007 to 2013 period; a successor, 'Inovaciniai ฤekiai', was being prepared for 2014 to 2020 and is not covered here.
๐ฑ๐ปLatviasince 2011delivered and closedbears on University-industry collaboration
Latvia's six industry competence centres, funded through the Investment and Development Agency with 53.2 million euros of ERDF money, had run 231 research projects involving 164 companies and 33 scientific partners by 31 December 2014, of which 78 were completed and 44 percent involved cooperation with scientific institutions.
The economics ministry funded industry-led competence centres, companies grouped by sector, which commissioned or ran research projects for their members, partly with universities and institutes, from EU structural funds.
Preconditions
EU structural funds large enough to finance a sector-by-sector research programme
Industry associations able to form a centre company and choose projects
A few large public universities able to supply contract research
Where else it has been used: Latvia borrowed the competence-centre label from Austria (aut-comet), which the evaluation cites, but put firms rather than universities at the centre, so most projects ended up as firm research with optional academic input.
Limits: The ministry evaluated a programme its own agency ran, and the figures are in a 43-page PDF: the 164 companies, 231 projects and 78 completed are on printed page 10 (PDF page 10) and Table 3 with 33 scientific partners on printed page 10 to 11 (PDF page 11). Most projects were firms' own research: the same evaluation says only 44 percent of projects involved cooperation with private or public scientific institutions, with 24 registered institutions, so the 231 overstates university-firm collaboration, and the 33 scientific partners count partners of the centres, not distinct universities. The programme was suspended from December 2011 to May 2012 and the centres only started full operation in 2013; by the end of 2014 they had drawn 56 percent of their funding and completed 34 percent of projects. Patents (3) and publications (25 against 107 planned) were few. The data are eleven years old. The evaluation states the 2007 to 2013 centre projects were to run until 31 December 2015, so this delivery is concluded; a successor competence centre programme for 2014 to 2020 under the same agency is what this ex-ante evaluation prepares, and this record does not show its scale.
๐ณ๐ดNorwaysince 2007still operatingbears on University-industry collaboration
The Research Council of Norway allocated 354 million kroner in 2025 to its Centres for Research-based Innovation, eight-year consortia of research institutions, companies and public bodies whose business and public-sector partners must co-finance at least half the cost, and in December 2025 named eight new centres with 98 business partners and 22 public-sector partners.
The research council funds eight-year centres in which a university or institute and research-active firms run a shared programme, with the council paying up to half and the partners the rest, and renews the portfolio in generations through open competitions.
Preconditions
Research-active firms in sectors such as energy, maritime and seafood that can co-fund long programmes
A research council with a stable budget line for long centre grants
Host institutions with the administration to manage multi-partner consortia
Where else it has been used: The SFI model mirrors competence centre schemes in Sweden and Austria (aut-comet); Norway pairs it with Centres of Excellence (SFF) for basic research, keeping the two instruments separate.
Limits: The council reports on its own scheme. The 354 million kroner is a funding figure shown as a number block on the programme page, not a count of collaborations, and the page gives no partner count for the centres running in 2025; the same page lists 22 centres under SFI-IV (2020-2028) and 17 under SFI-III (2015-2023), counted by us from the lists. The partner figures come from a separate council news item of 15 December 2025 (https://www.forskningsradet.no/nyheter/2025/nye-forskningssentre-innovasjon/), describe centres that were only being contracted, and the centres also include 16 research institutions, 17 international research partners and 13 foreign companies. Host institutions include research institutes such as SINTEF, not only universities. The 2007 start is from the history of the scheme (first centres appointed 2006, operating 2007) and is not stated on the cited page. Nothing here measures the innovation the centres produced.
๐ฉ๐ฐDenmarksince 2021still operatingbears on Public-private collaboration
Denmark's Agency for Digital Government, representing the state, regions and municipalities, and the banks, through a Finance Denmark subsidiary, jointly own MitID, launched in May 2021 to replace NemID and migrated in 2021 and 2022; by July 2026, 97.2 percent of residents over 15 had an active MitID, and it averaged 94.4 million transactions a month in 2025.
The state and the banks tendered and co-own one identity solution that works for both online banking and public services, contracted one supplier to build and run it, and moved about five million users off the previous shared solution, NemID.
Preconditions
A previous shared public and bank identity (NemID) that had already made one login normal
A banking sector organised to act through one industry body
Mandatory digital communication with the state, which drives near-universal take-up
Where else it has been used: Sweden's BankID (swe-bankid) is bank-owned with state acceptance, the Netherlands' DigiD is state-only; MitID is the case where the two sides share ownership of the same credential.
Limits: The page is run by the partnership the record describes, so this is the owner grading its own system; it says the figures were updated in July 2026 and the 94.4 million is labelled an average for 2025, while the 97.2 percent and the 5.3 million registered users carry no separate date. The page refused WebFetch with HTTP 403 and was read with a browser user agent. The partnership terms (established in 2016 after a tender, the state represented by the agency, the banks by an FR I subsidiary of Finance Denmark, the partnership owning the rights to the solution, name and logo) and the timeline (go-live May 2021, migration through 2021 and 2022) are on https://digst.dk/it-loesninger/mitid/om-mitid/. The build and operation were contracted to one private supplier, Nets DANID, for DKK 939 million plus ten years of operation, so the private side is both the banks as co-owners and a vendor. Near-universal coverage follows from mandatory digital post and bank login in Denmark, not only from the partnership's design, and the share says nothing about people who cannot use the app; the minister sent parliament a project evaluation in September 2024 (ft.dk, DIU Alm.del bilag 3), which this record has not read.
๐ฎ๐นItalysince 2016still operatingbears on Public-private collaboration
Italy's Agency for Digital Italy runs SPID, the public digital identity system in which accredited identity providers, private firms such as Aruba, InfoCert, Namirial and Register alongside state-controlled Poste Italiane and public companies such as Lepida and InfoCamere, issue the credentials under agreements with the agency, and by 8 October 2025 41 million SPID identities had been issued, when the agency extended the providers' agreements to 2027.
The digital agency set technical rules and signed agreements with accredited identity providers, who verify citizens and issue credentials that every public administration must accept, so private firms carried the cost of enrolment while the state set the standard.
Preconditions
Firms and public companies with existing identity verification channels (post offices, trust service providers, chambers of commerce)
A legal mandate obliging public administrations to accept the credential
A fee model or public payment that keeps providers in the scheme, which is what the 2025 renewal negotiated
Where else it has been used: Sweden's BankID (swe-bankid) and Denmark's MitID (dnk-mitid) put banks at the centre; SPID spread issuance across many accredited providers, and Italy is now weighing a return to the state card.
Limits: The 41 million is a rounded figure in a news item, not a statistics series, and counts identities issued, not active users; one person can hold SPID from more than one provider, so the count overstates people. The mix of providers is read from the agency's list of accredited identity providers (https://www.agid.gov.it/it/piattaforme/spid/identity-provider-accreditati), which shows the first accreditations in December 2015, the first agreements in February 2016, an additional act signed with each active provider on 8 October 2025, and one provider (In.Te.S.A.) that ceased in April 2023; the agency publishes no split of identities by provider, so the private share of issuance is not known, and the largest provider, Poste Italiane, is state-controlled. The future of the partnership is uncertain: the news item frames the extension as accompanying a transition toward a single national identity built on the state-issued CIE and the IT Wallet. If the government moves logins to CIE the partnership will shrink; this record does not count that as erosion because the item reports no loss yet. The item gives no transaction figure, so the record shows reach, not use. The 55 million CIE in the second metric are cards issued; the same item says only 9 million CIE digital credentials had been activated, so the two counts in the item's 96 million total are not like for like.
๐จ๐ฟCzechiasince 2021still operatingbears on Public-private collaboration
The Czech Ministry of Transport procured the completion of the D4 motorway between Pลรญbram and Pรญsek as the country's first transport public-private partnership, and the concessionaire Via Salis opened 32 km of new motorway and 16 km of modernised sections on 17 December 2024 after 43 months of construction, as the ministry had announced in June 2021, with the state paying an availability payment until 2049.
The ministry tendered a design, build, finance, operate and maintain concession for one motorway corridor, the concessionaire took over the existing sections in 2021 and built the gap, and the state pays monthly only once the road is open, with deductions when availability or quality falls short.
Preconditions
Parliamentary approval of a multi-decade availability payment as a quasi-mandatory expenditure
Advisers able to structure the first deal in a country with no PPP track record
A corridor already permitted by the roads directorate, which spent CZK 364 million preparing it
Where else it has been used: Latvia's ฤถekava bypass (lva-kekava-ppp) and Slovakia's D4/R7 (svk-d4r7-ppp) used the same availability-payment model in the region; the Czech audit office's warning about missing domestic cost data is the lesson most likely to recur.
Limits: The opening, the kilometres and the 43 months are the ministry's own press release; the planned December 2024 opening is in the ministry's release of 9 June 2021 as listed on its D4 tag page, so on-time delivery is the ministry's claim checked against its own earlier announcement. The cost figures come from the Supreme Audit Office (NKร, audit 22/29, press release of 11 December 2023, https://www.nku.cz/cz/pro-media/tiskove-zpravy/nku-upozornuje:-vystavba-dalnice-d4-formou-projektu-ppp-mozna-pro-stat-nebude-tak-vyhodna--jak-se-predpoklada-id13640/), which warned that the value-for-money case rested mainly on the adviser's foreign experience because the ministry could not state its own maintenance costs for the 1,276 km of motorway it ran, and that with inflation the state's availability payment would be about CZK 4.2 billion above the CZK 29.5 billion approved. One project is not a programme: the record shows that the Czech state can close and deliver a PPP, not that it does so often. The concessionaire is French-owned (VINCI Highways and Meridiam), and the construction was done by VINCI's Czech subsidiary.
๐ฑ๐ปLatviasince 2021still operatingbears on Public-private collaboration
Latvia's Ministry of Transport and the state roads company signed a design, build, finance and maintain contract for the ฤถekava bypass on the A7 on 16 July 2021, the private partner Kekava ABT built the roughly 18 km road from 2021 to 2023 and it opened on 13 October 2023, with the state paying EUR 250.13 million in availability payments over 20 years.
The transport ministry and the state roads company procured a bypass as one design, build, finance and maintain contract, the private partner built it in about two years, and the state pays quarterly for 20 years once the road is available.
Preconditions
A PPP law and a supervising agency (CFLA) to run procurement
Eurostat clearance that the project sits off the public balance sheet
Lenders willing to finance a first-of-kind deal in a small market
Where else it has been used: The Czech D4 (cze-d4-ppp) followed the same availability model at larger scale; Lithuania's PPP portfolio (ltu-vpsp) shows the municipal route the Baltic states use more often.
Limits: CFLA supervises PPP projects in Latvia and is part of the delivery it reports, so this is the supervisor's announcement on opening day, not an audit. The 27 months are our arithmetic from the two dates on the page; the page gives no planned opening date, so it does not show whether the road opened on schedule, and it gives no construction cost separate from the availability payments. The same page says the state roads company had to obtain several Eurostat opinions to keep the project off the government balance sheet, and that CFLA is working to remove that step: the procedure was slow before construction began. One project is not a programme; the page calls it the first large road PPP in the Baltic states, which also means Latvia has no record of repeating it. Representatives of the Nordic Investment Bank and the European Investment Bank took part in the opening; the page does not describe the financing.
๐ฑ๐นLithuaniasince 2009still operatingbears on Public-private collaboration
Lithuania's state and municipal institutions had concluded 76 public-private partnership contracts by 31 December 2025, 49 of them in force at the end of 2025 across culture and sport, waste, energy, law enforcement, education, transport and defence, with only one new contract signed in 2025.
A central agency advises and monitors, and individual ministries and municipalities procure PPP contracts for buildings, lighting, waste and defence facilities, reporting each year so the agency can publish a national portfolio.
Preconditions
A PPP law with a central agency that collects data from every contracting authority
Municipal borrowing limits that make off-budget delivery attractive
Eurostat treatment of the contracts, which decides whether they count against the deficit
Where else it has been used: Latvia's single large road deal (lva-kekava-ppp) is the alternative model; Lithuania now plans larger transport PPPs (a regional road and bypasses worth EUR 20 to 250 million) on top of its many small municipal ones.
Limits: CPVA compiled the figures from data submitted by state and municipal institutions; it is the national PPP centre, not an auditor, and the underlying annual report is linked from the page but was not opened. Most contracts are municipal (24 municipalities in 2025), so the count mixes national objectives such as defence and law enforcement, where the public sector made its largest payments in 2025 (EUR 32.5 million and EUR 10.3 million), with local sports halls and street lighting. A contract count says nothing about value for money; the National Audit Office has published a critical audit of PPP project management (valstybeskontrole.lt), which this record has not read. The gap between 76 concluded and 49 in force is not explained on the page (expiry, termination or both). The pace is slow: one contract in 2025, and 18 projects planned for 2026 that are still in preparation. The start year 2009 is the year of the finance minister's reporting rules (order 1K-489 of 24 December 2009) cited on CPVA's statistics page, not a documented first contract.
๐ธ๐ฐSlovakiasince 2016still operatingbears on Public-private collaboration
Slovakia built the D4 motorway and R7 expressway around Bratislava, 59.0 km at EUR 1,670 million, as a public-private partnership between 2016 and 2021, the only PPP among the major road projects the Supreme Audit Office listed in its 2025 audit of road investment.
The transport ministry procured the Bratislava bypass as one availability-payment concession covering design, construction, financing and decades of operation, so a single private partner delivered 59 km at once rather than in budget-limited sections.
Preconditions
A ministry able to procure and supervise a EUR 1.7 billion concession
International lenders and sponsors willing to finance it
Budget room for decades of availability payments
Where else it has been used: The Czech D4 (cze-d4-ppp) and Latvia's ฤถekava bypass (lva-kekava-ppp) used the same model later and at smaller scale.
Limits: The figures are one row of a table on printed page 12 of the audit office's PDF, which lists the period 2016 to 2021, 59.0 km, EUR 1,670.00 million and 'PPP' as the funding source; the audit is about road investment generally and does not evaluate the PPP itself, its cost against a public alternative or its schedule against plan. The table gives no definition of the cost (construction cost, capital value or total availability payments), so the EUR 1,670 million cannot be compared with the other rows without that caveat. The opening dates of the sections are not on this page; 2021 is the end of the period the table gives. The same table shows every other major road project financed from EU or state funds, so the record documents one delivered PPP, not a programme, and this source says nothing about Slovakia's earlier PPP attempts.
๐ณ๐ดNorwaysince 2005still operatingbears on Public-private collaboration
The Norwegian Public Roads Administration has delivered four road sections as public-private partnerships (OPS) now in operation, opened in 2005, 2006, 2009 and July 2020, each run by a private company for 20 to 25 years, and had two more under planning or construction in 2026, Rv. 555 Sotrasambandet and E10/Rv. 85 Tjeldsund to Langvassbukt.
4 projects, ops road projects completed and in operation, 2026-10. 2 projects, ops road projects in planning or under construction, 2026-10. Statens vegvesen, Vรฅre OPS-prosjekter, retrieved 2026-10-02.
The move, and what it needed
Parliament designates a road section for OPS, the roads administration tenders it, and a private company plans, builds, finances, operates and maintains the section for 20 to 25 years against state payments, after which it returns to the state.
Preconditions
A parliament willing to commit payments decades ahead project by project
Domestic contractors and financiers able to take on whole-life road risk
A roads agency that can write performance requirements instead of designs
Where else it has been used: Sweden's Trafikverket is studying OPS projects for its 2026 to 2037 plan; Norway's experience shows a model that survives but stays a small share of road building.
Limits: The page is the roads administration's own list with no date of update, so the asOf is the retrieval date. It gives no cost, schedule or value-for-money figure, so the record shows frequency and duration, not performance. Six projects in two decades is a small programme for a country that builds many roads, and it moved in fits: three sections opened between 2005 and 2009, then nothing opened until 2020, after the transport ministry asked the agency in 2015 to use the model on three new projects (stated on the agency's OPS page, https://www.vegvesen.no/vegprosjekter/prosess/ops-ppp/); the list does not say what became of each of those three. The page says parliament decides which projects use OPS, so the count reflects political choices project by project. Projects by Nye Veier, the state road company, are not on this list.
๐ญ๐ทCroatiasince 2006operating below its peakbears on Public-private collaboration
Croatia's register of public-private partnership contracts lists 17 contracts with an estimated capital value of EUR 338.3 million, 13 of them signed between 2006 and 2008, mostly Varaลพdin county schools, one in 2013 for Zagreb Airport, one in 2016 and two street lighting contracts in 2018, and none since October 2018.
Counties and cities signed design, build, finance and maintain contracts for schools, halls and a bus station in a burst before the 2008 crisis, the state concessioned the Zagreb Airport terminal in 2013, and then PPP contracting all but stopped while responsibility for PPP passed to the economy ministry.
Preconditions
A PPP law and a ministry able to approve projects quickly
Local governments with the capacity to prepare PPP tenders, which only a few had
Eurostat treatment that keeps contracts off the deficit, whose tightening after 2008 made the model less attractive
Where else it has been used: Lithuania (ltu-vpsp) kept a steady flow of small municipal PPPs with a central agency's support; Croatia's register shows what happens without that pipeline.
Limits: The counts are ours: the register page lists 17 contracts and a total of EUR 338,260,027, and the signature dates are on each contract's own page under /jpp-project/ (for example Zagreb Airport, 2 December 2013; Novi Vinodolski street lighting, 25 October 2018, the latest), which we opened one by one and tallied. The register holds only contracts entered under the PPP framework; concessions under the separate concessions law, such as motorways, are not in it, so the record does not show that private participation in Croatian infrastructure stopped, only that the PPP model did. The ministry's project list (https://investcroatia.gov.hr/jpp/projekti/) also shows approved projects not yet contracted, such as Trg pravde in Zagreb (EUR 179.2 million) and county schools, so the pipeline is not empty. The peak itself was narrow: ten of the 13 early contracts are in Varazdin (eight school extensions, the county building and a sports hall). Zagreb Airport alone is EUR 188.6 million, more than half the register's value. The register page has no date of update, so the zero is as of retrieval.
๐ฌ๐ทGreecesince 2006still operatingbears on Public-private collaboration
Greece's Inter-Ministerial PPP Committee and the finance ministry's PPP Unit have taken 33 public-private partnerships to signed contracts since 2006, from Attica school buildings and fire stations to waste plants, the Kalamata to Rizomylos and Crete northern road sections and ultrafast broadband, with another 34 projects listed in procurement or preparation in 2026.
A 2005 law set up an inter-ministerial committee that approves which projects may be procured as PPPs and a central unit in the finance ministry that prepares and tenders them for ministries, regions and municipalities, mostly as availability-payment contracts.
Preconditions
A central PPP unit with its own procurement expertise, co-financed by EU funds
Committee approval at ministerial level, which gives projects political cover
EU structural and recovery funds that can co-finance availability payments
Where else it has been used: The same centre-led model as Lithuania's CPVA (ltu-vpsp) at larger scale; Croatia (hrv-jpp-registar) had a similar law but no comparable pipeline after 2008.
Limits: The home page states 33 signed PPP contracts and 34 approved PPP projects since 2006, without a date of update. We read the 67 project pages in the unit's list (https://pppunit.minfin.gov.gr/en/sdit/erga, seven pages) and tallied their status fields ourselves: 33 'Contract signed', 13 in competitive dialogue, 13 at request for proposals, 6 at expression of interest and 2 approved and in preparation, so the 34 in the second metric is our count. The list counts the ultrafast broadband programme as seven lots, each showing the programme's EUR 868 million cost, so a sum of project costs would be wrong and the record gives none. 'Contract signed' is the unit's last populated status; no project is marked completed, so the list does not say how many assets are in operation or whether they came in on time and budget, and several roads and waste plants took three to ten years from committee approval to signature (for example the Peloponnese waste system, approved December 2011, signed June 2018, and the Nafplio marina, approved November 2013, signed April 2023). The unit is part of the ministry that runs the programme, and its home page cites a World Bank ranking as an endorsement; the ranking is not the evidence here. 2006 is the first committee approval in the list (Attica schools, April 2006).
๐ง๐ชBelgiumsince 2023still operatingbears on Public-private collaboration
Belgium's federal Buildings Agency had the new 440-place prison de Kade in Antwerp, part of the federal prisons masterplan, designed, built and financed by the consortium Hortus Conclusus under a DBFM contract, with works from 2023 to 2026 at a construction cost of EUR 195.5 million and an annual availability fee of EUR 23.1 million for 25 years of maintenance, after which the prison returns to the federal state free of charge.
The federal Buildings Agency tendered a design, build, finance and maintain contract for a new prison, a consortium of an infrastructure investor and contractors built and financed it, and the state pays an annual availability fee for 25 years while the Justice ministry runs the prison.
Preconditions
A federal prisons masterplan with a pipeline of projects to justify the procurement effort
Legal and financial advisers for the agency (here a consortium led by Stibbe)
Domestic contractors and investors able to carry construction and 25-year maintenance risk
Where else it has been used: The same availability-payment structure as the Czech D4 (cze-d4-ppp), applied to social infrastructure instead of a road.
Limits: The page is the contracting agency's own project sheet. It says the prison 'a รฉtรฉ rรฉalisรฉe' under a DBFM contract and that Antwerp 'now has two prisons', but it states the works period as 2023 to 2026 and gives no opening date or planned date, so the record cannot show whether the prison opened on time or on budget; status operating rests on the page's own past tense, updated September 2025. The construction cost and the availability fee are the agency's figures, not an audit; 25 years of fees at EUR 23.1 million in 2023 prices would be about EUR 578 million before indexation (our arithmetic), well above the construction cost, which is the normal shape of a DBFM but means the cost to the state is the fee stream, not the build cost. One project is not a programme: the page places the prison in the federal masterplan for detention and internment, but this record documents only this contract and says nothing about the masterplan's other projects. The consortium's members are named on the page (TINC, Jan De Nul, EEG, with ISS for services).
๐ฒ๐ฉMoldovasince 2024still operatingbears on University-industry collaboration
Moldova's National Agency for Research and Development (ANCD) financed 13 innovation voucher projects in 2025, in which small and medium firms buy research and development services from research organisations, after 28 applications to the 2025-2026 call yielded 10 funded projects and four funded projects were stopped when the firms declined to sign contracts.
The national research agency gives small firms a voucher that pays a public research organisation to do a defined R&D or testing job for them, with the firm co-financing a share.
Preconditions
A research funding agency able to contract firms as well as research institutes
Research organisations with services firms are willing to buy
Firms willing to sign a state contract and co-finance, which four 2025 winners were not
Limits: The agency reports on itself, on pages 28 and 29 of a 137-page PDF. The arithmetic in the report does not close on its face: 10 new projects plus 7 continuing ones is 17, and the report also says four projects were stopped when three firms refused to sign financing contracts, and states the total of 13 without showing the subtraction. The volume is very small: 1,073.0 thousand lei of state money and 153.4 thousand lei of firm co-financing in 2025, roughly 60,000 euros, and the same report counts five technology transfer projects approved in 2025 under a separate call, not added here. The 2024 report gives seven voucher projects funded in 2024, so the instrument is two years old and its scale is pilot-like even if national in reach. Counts are projects financed, not completed or commercialised. The started year is the first funded call read in the 2024 report, not a programme launch date.
๐ท๐ดRomaniasince 2014still operatingbears on University-industry collaboration
Romania's research funding agency UEFISCDI accepted 35 firm-led projects with research organisation partners for funding in the 2024 PTE (transfer to the economic operator) competition, plus one more on a second list, after 47 in the 2021 competition.
The national research funding agency runs a periodic competition in which a company leads a project to take a research organisation's result to market, with the state grant paying both sides under state aid rules.
Preconditions
A national R&D plan with a partnerships programme and a budget line for it
Public universities and national R&D institutes with results ready for transfer
An agency with an online submission and peer evaluation platform
Limits: The count is ours: the four-page PDF numbers its rows 1 to 35 and is a list of projects accepted for funding, not contracts signed or completed; a separate 'Lista nr. 2' adds one more project (PN-IV-P7-7.1-PTE-2024-0336), and a reserve list exists, so the contracted number may differ. The 2021 figure is the row count of a different file on the same agency site (resource-862977, 'Lista proiectelor PTE 2021 acceptate la finantare'), not at the cited URL. Each project is coordinated by a company with one or more universities or national R&D institutes as partners, and the maximum grant was about 1.5 million lei, so this is a few dozen projects a competition in a country of 19 million. PTE competitions are not annual (2019, 2021, 2024, and a 2026 call that closed on 16 September 2026 with 378 proposals submitted and none funded yet), so the gaps between rounds are part of the record. The started year is approximate: PTE ran under PNCDI III from about 2014 to 2016, which is not stated in the cited file. The figures say nothing about technology actually transferred or sold.
๐ท๐ธSerbiasince 2019still operatingbears on University-industry collaboration
Serbia's Innovation Fund financed 79 joint research and development projects of private small and medium firms with public research organisations across eight calls of its Collaborative Grant Scheme, the eighth call, decided in 2024, funding eight.
The national innovation fund runs competitive calls in which a private SME leads a consortium with a public research organisation, pays part of the budget, and must route at least 15 percent of the grant to the research partner.
Preconditions
An arm's length innovation agency with grant management capacity built under donor projects
Accredited public research institutes and faculties able to contract with firms
SMEs able to co-finance a third or more of an R&D budget
Where else it has been used: Montenegro's Innovation Fund runs a Collaborative Grants programme on the same template.
Limits: The Fund reports on its own programme and publishes no total: 79 is our count of the project entries on its per-cycle pages (14, 9, 13, 12, 11, 6, 6 and 8 projects in cycles one to eight; each page renders every project twice, and the count is of distinct titles). The pages carry no dates, so the year of each cycle is not read there; that the eighth call funded eight projects in 2024 is from the Ministry of Finance CFCU release of 15 August 2024 ('8 under the Collaborative Grant Scheme Program'), and a Fund release of 20 March 2024 reports six joint projects in the call before. The counts are grants awarded, not completed projects or commercial results. The scheme is co-financed by the European Union through IPA and began under World Bank financing, so the delivery is the Fund's but the money is partly external. Annual volume has fallen from 11 to 14 projects a call in the early cycles to six to eight, and the programme page shows the call as closed with no ninth cycle listed on 2026-10-02, so 'operating' rests on the absence of any announced end. The started year is approximate: the first cycle is not dated on the cited page. Each consortium needs only one firm and one public research organisation, and requires firms to co-finance 30 to 40 percent, so 79 projects is a small national volume.
๐ท๐ดRomaniasince 2009operating below its peakbears on Public-private collaboration
Romania's Prima Casa programme, renamed Noua Casa in 2020, had the National Credit Guarantee Fund for SMEs (FNGCIMM) issue 334,048 state guarantees and guarantee promises worth RON 31.68 billion on mortgages made by participating banks between 2009 and December 2024, but only 811 guarantees in 2024, and the ratio of guaranteed lending to banks' unguaranteed housing loans fell from about 70 percent in 2017 to about 10 percent in 2023.
The finance ministry, through a state guarantee fund acting as its agent, guaranteed half or more of each first-home mortgage that participating banks underwrote, set a lower interest cap and annual guarantee ceilings, and then cut the ceilings year by year so private lending took over.
Preconditions
A guarantee fund with a mandate from the state and contracts with every major bank
Annual guarantee ceilings set in the budget law, which is the lever that also wound the programme down
A housing market where banks were willing to lend once the state took part of the risk
Where else it has been used: First-home guarantee schemes run through banks exist across Europe; Romania's is one where a crisis measure from 2009 grew to dominate the mortgage market and was then deliberately tapered.
Limits: The live file at mfinante.gov.ro did not answer a script on the day of retrieval, so the record cites the Internet Archive copy; the figures are on pages 1 and 2 of a nine-page PDF, prepared with FNGCIMM, which runs the programme, so the ministry is grading its own scheme. The decline was largely chosen: the note says the 2017-2021 medium-term strategy set falling guarantee ceilings to decelerate state intervention and that its objectives were met, while private mortgage lending without a state guarantee grew, so the erosion is a planned withdrawal as much as a loss of capacity. The 70 and 10 percent are approximate ('aprox.', 'circa') and describe a ratio of guaranteed to unguaranteed lending as the note words it, which is not the same as a market share. 2017 is the first year of the strategy, not a demonstrated peak: the note gives no earlier year, and the programme's share was probably at least as high in its first years after 2009. 811 guarantees in 2024 is a calendar-year count against a cumulative total, and the note does not give the annual peak; the 1,449 guarantees the Ministry of Finance paid out to banks since launch show the state carried real credit risk. Whether the programme raised home ownership or mainly house prices is not shown.
๐ญ๐บHungarysince 2013delivered and closedbears on Public-private collaboration
Hungary's central bank refinanced commercial banks at zero interest on condition that they lent to small and medium firms at a capped margin under the Funding for Growth Scheme, and across its phases from June 2013 to 2022 some 75 thousand enterprises received about HUF 6,400 billion of loans, more than half of them investment loans.
The central bank offered banks zero-cost refinancing for loans to small and medium firms on condition that they cap the client rate, so banks did the origination, credit assessment and servicing while the central bank bore the funding cost.
Preconditions
A central bank willing to run a targeted quasi-fiscal programme on its own balance sheet
A banking sector with branch networks reaching small firms nationwide
Low inflation at launch, so zero-rate refinancing did not fight monetary policy, which ceased to hold by 2022
Where else it has been used: The design follows the Bank of England's Funding for Lending Scheme (2012); Hungary ran it longer and steered it to fixed-rate investment loans, and the end of low inflation closed it.
Limits: The central bank designed, financed and here evaluates its own scheme; both headline numbers are rounded ('some', 'approximately') and appear in a PDF on page 7 and again on pages 9 and 20; the MNB also publishes a Power BI dashboard of the loans that a script cannot read. The MNB carried the funding cost and banks the credit risk, so the partnership was a public subsidy delivered by private lenders, with the policy cost sitting off budget on the central bank's balance sheet. The book's estimate that the scheme added more than 5 percentage points to GDP between 2013 and 2021 is the MNB's own model result and is not used here. Enterprises are counted across phases (FGS 1 to 3, FGS fix, FGS Go!, Green Home), so the 75 thousand cannot be compared with the stock of firms in any one year, and the volumes include refinanced foreign currency loans in the first phase. Some lending would have happened without the subsidy; the record does not measure additionality.
๐ท๐ธSerbiasince 2020operating below its peakbears on Public-private collaboration
Serbia's Government paid a monthly allowance to young people without work experience while employers trained them for nine months under the Moja prva plata programme run by the National Employment Service, which placed 9,237 young people with 5,471 employers in the 2021 cycle and 7,214 with 4,520 employers from the 2023 call in January 2024, but ran no call for 2024 and started no activity under the measure in 2025 against a plan of 10,000 places.
The Government let firms post training positions on a national portal, let young people apply, had employers pick candidates, and paid the trainees directly through a tripartite contract with the National Employment Service, so firms trained at no wage cost.
Preconditions
A national employment service able to sign and monitor thousands of tripartite contracts
Employers short of staff and willing to train, which the programme was designed around
A budget line of about RSD 2 billion per cycle that the Ministry of Finance would renew
Where else it has been used: Close to the youth traineeships run under the EU Youth Guarantee; Serbia's version selects through a single online portal and leaves the choice of candidate to the employer.
Limits: The agency reports on its own programme. The zero is in Table 6 (PDF page 23), measure 5.8, planned 10,000 and realised 0, and section 5.8 (PDF page 41) states that no activities to implement the measure were started in 2025; the report gives no reason. The 2021 figure is from a Government portal release (srbija.gov.rs/vest/595588, 3 December 2021), not at the cited URL, and counts young people employers selected at the start of a cycle, not completions; the 2022 cycle was not read, so 2021 is the highest dated count found, not a demonstrated peak. The 2024 report (Izvestaj o radu NSZ za 2024. godinu, Table 6 and page 45, nsz.gov.rs) gives 7,214 young people at 4,520 employers from the 2023 call, contracted between 1 and 19 January 2024, of whom 4,950 were with 3,892 private employers and 2,264 with 628 public employers, and says no public call for 2024 was carried out; the first cycle in 2020 also reserved 2,648 places for public-sector employers, so the programme was never wholly public-private. The 35 percent of first-cycle participants who signed an employment contract is stated in the 2021 release without a denominator or follow-up period. In 2025 municipalities co-financed 36 placements under a local variant (Table 18 of the 2025 report), which this record does not count. The programme portal (mojaprvaplata.gov.rs) still announces a new cycle of up to 10,000 places with no date, so the halt may be a gap between calls rather than an end; the record documents two years without a national call, not abolition. The state paid the allowance and the employer supplied the training and the workplace, so it is a wage subsidy for trainees as much as a training programme, and some employers may have used it in place of hires they would have made anyway.
๐ฒ๐ฉMoldovasince 2013dismantledbears on Public-private collaboration
Moldova's Public Property Agency concessioned the assets of Chisinau International Airport and its land to Avia Invest under contract 4/03 of 30 August 2013, with investment obligations of EUR 244 million over the life of the concession, and in 2020 notified the concessionaire that the contract was rescinded for failing to post the performance guarantee, while Avia Invest claimed EUR 900 million from the state at arbitration in Stockholm.
The state property agency handed the airport's assets and land to a private concessionaire for investment and operation in exchange for fees and investment obligations, then used the contract's performance-guarantee clause to terminate it once a new government contested the deal.
Preconditions
A concession law and a state property agency with authority to award and terminate concessions
A contract with enforceable performance-guarantee clauses, which is what made the rescission possible
Political continuity, which failed here: the concession did not survive a change of government
Where else it has been used: An airport is often a small state's first large concession; Moldova's shows how a concession awarded without broad political cover can be unwound by the next government and end in arbitration.
Limits: The source is one party to a dispute stating its case: the agency wrote the communique to answer press coverage, and the EUR 900 million is the concessionaire's claim as the agency reports it, not an award. The live page now returns 404 on app.gov.md and the record cites the Internet Archive copy. The EUR 244 million is a contractual obligation over the whole concession, not an amount invested; the communique gives no figure for what Avia Invest actually invested in 2013 to 2020, so the record shows the partnership's scale on paper and its end, not its delivery. The communique shows a rescission notice that the concessionaire contested, not the end of its operation: no official page opened shows when the airport returned to the state, and secondary sources (not used as evidence) date the final annulment to a Chisinau Court of Appeal ruling of November 2022. The state enterprise I.S. Aeroportul International Chisinau that appears in the agency's procurement notices held the conceded assets throughout the concession, so its presence does not by itself show a return of operations. The outcome of the arbitration is not covered. The concession was the country's flagship infrastructure PPP: the agency's own list of PPP contracts it signed (archived April 2021) showed a single other contract, for bus stations, signed on 19 December 2018, so Moldova's national PPP portfolio was thin before and after this rescission.
๐ฐ๐ฟKazakhstansince 2019delivered and closedbears on University-industry collaboration
Kazakhstan's innovation ministry (MDDIAI), through the Innovation Consortia Matching Grant Program of the World Bank financed Fostering Productive Innovation Project, had supported 33 technology consortia of research teams and firms, including 12 competence centres, by September 2024, whose first sales and signed contracts totalled about 17 billion tenge.
The ministry ran a competitive matching-grant window that paid consortia of research teams and firms to take a technology to first sales, with firms co-financing and a project unit monitoring sales and contracts.
Preconditions
Loan financing from a development bank with a project management unit able to run calls
Research teams with near-market technologies and firms willing to co-finance
A follow-on budget to keep calls running once the loan closes, which was not yet in place
Where else it has been used: The consortia window scaled up the earlier Technology Commercialization Project; its continuation depends on a second loan rather than the state budget.
Limits: The figure is the World Bank's mission account of a programme it financed, not a ministry statistic, and the number is in a PDF (paragraph 6 of the aide memoire). The sentence 'supported 33 technology consortia, including 12 competence centers, and 9 IICs' does not make clear whether the nine inclusive innovation consortia are inside or beside the 33. The aide memoire does not define a consortium's membership; the 2020 restructuring paper describes the window as a tool 'for fostering industry-science linkages' with an average 40 percent private co-financing in the first six grants, and the annex shows competence centres hosted by universities (Astana IT University, the ENU-Lab centre), but not every consortium is shown to include a university. The 17 billion tenge is first sales and contracts as reported by beneficiaries, rounded ('around'), and is not audited. The programme ran on loan money: the 2020 restructuring cut the consortia component from US$35 million to US$25.1 million, and in 2024 the ministry asked to cancel a further US$5.4 million of unused consortia grants because no new round could be run before closing (extended to June 30, 2025), so the status is concluded and a follow-on (FPIP-2) was only in preparation. The 2019 start is the year the first consortia grant agreements were signed according to the 2020 restructuring paper.
๐ฆ๐ฒArmeniasince 2000still operatingbears on Public-private collaboration
Armenia ran a sequence of water public-private partnerships from 2000, a Yerevan management contract, a 2006 to 2016 Yerevan lease and management contracts for the national and three regional utilities, which raised daily supply from about 6 hours in 1998 to 24 hours in most of Yerevan by 2015, and from January 2017 consolidated them into a single 15-year national lease covering about 2.2 million people.
The government contracted private operators in stages, first to manage the Yerevan utility for a fee with performance incentives, then to lease it, then to manage the national and regional utilities, and finally to lease all of them as one national concession paying a lease fee to the state, while the state kept financing capital works.
Preconditions
Donor and state finance for network rehabilitation, which the operators did not provide
Government willingness to raise tariffs in steps matched to visible service gains
Successive contracts long enough to build a local record that justified the next, larger one
Limits: All figures are in the executive summary of a PDF (printed pages ix to xi). The review was written by World Bank staff with support from PPIAF, a facility that promotes infrastructure PPPs, and credits the operators generously, and the 24-hour figure is 'most of Yerevan', not the country. The same review says non-revenue water barely improved, that tariffs outside Yerevan stayed below cost recovery when the management contracts ended in 2016, and that capital investment was financed by the government, so the private operators ran networks without financing them. The record's newest verified fact is the start of the national lease in January 2017; we did not find an official account of its performance since then at an open URL, so 'operating' rests on its 15-year term, and any later dispute, renegotiation or early termination is not covered. The national lease introduced a single national tariff, a cross-subsidy from Yerevan to the rest of the country, whose effect on supply hours outside Yerevan is not shown here.
๐ฆ๐ฟAzerbaijansince 2023still operatingbears on Public-private collaboration
Azerbaijan's government brought its first utility-scale private solar plant, Masdar's 230 MW Garadagh plant built with US$262 million of foreign investment, into operation on 26 October 2023, followed on 16 September 2026 by the 100 MW Gobustan plant won at a 2024 auction, with further private wind and solar plants under investment agreements and power purchase contracts.
The government signed investment agreements, land leases, power purchase, grid connection and balancing agreements with foreign developers, first bilaterally with Masdar and then through an auction, for private plants connected to the national grid.
Preconditions
A state counterparty able to sign long-term power purchase and grid agreements
Gas exports valuable enough that each kilowatt-hour from renewables frees gas for sale, which is how the ministry frames each plant
Gulf developers willing to invest on bilaterally negotiated terms before an auction was held
Where else it has been used: Uzbekistan's tendered solar IPPs (uzb-renewable-ipp) used the same Gulf developers in the same years.
Limits: The ministry describes its own programme on an undated web page that has been updated at least to September 2026; the investment figure and capacities are the ministry's, with no audited cost and no published tariff. Two private plants are operating, 330 MW together by our sum, against 2,170 MW of renewable capacity the same page reports in operation, most of it state hydropower; the rest of the private pipeline (ACWA Power's 240 MW Khizi-Absheron wind farm, bp's 240 MW Shafag plant, Masdar's 1 GW 'mega project', two 50 MW plants in Jabrayil) is under construction or planned, and the page gives no completion dates. 'Implemented jointly with local and international partners' covers different structures: Garadagh and Gobustan are built and owned by foreign developers, though the page does not name their buyer or contract terms, Shafag serves bp's own terminal through a 'virtual' transmission swap, and the page does not say which plants were competitively tendered beyond Gobustan. The collaboration is recent, so the record shows a start, not a sustained programme. The ministry's page describes Garadagh as 'put into operation as a pilot project'; it is a 230 MW grid-connected plant, so we read the word as first-of-kind rather than as a trial, but the source uses it.
๐ฏ๐ดJordansince 2015operating below its peakbears on Public-private collaboration
Jordan's Ministry of Energy and Mineral Resources added utility-scale solar and wind plants selling to the grid every year from 2015 to 2021, mostly private projects under direct proposals and competitive rounds, peaking at 345 MWac commissioned in 2020, and then added none: its annual reports put grid-selling renewable capacity at 1,498 MW in 2022 and 2023 and 1,478 MW in 2024.
0 MW (stock 1,498 MW in 2022 and 2023, 1,478 MW in 2024), utility-scale renewable capacity selling to the grid added from 2022 to 2024 (computed from the reported stock), 2024. 345 MWac, utility-scale renewable capacity added in 2020, the peak year, 2020. Ministry of Energy and Mineral Resources of Jordan, Annual Report 2024, retrieved 2026-10-02.
The move, and what it needed
The ministry offered sites and long-term power purchase agreements with the state-owned single buyer NEPCO to private developers through direct proposals and competitive rounds, alongside donor-financed public plants, and added no utility-scale plant after 2021.
Preconditions
A single buyer, NEPCO, able to sign 20-year take-or-pay contracts with sovereign backing
A renewable energy law allowing direct proposals and a regulator able to license IPPs
Grid capacity and demand growth to absorb more contracted capacity, which the report does not discuss and which is the most likely limit
Where else it has been used: Egypt's feed-in tariff (egy-renewable-fit) and Uzbekistan's solar tenders (uzb-renewable-ipp) are the same move; Jordan is the case where the additions stopped.
Limits: The 2024 annual report's year-by-year table, 'Renewable Energy Power Generation Projects: Direct Proposals and Grants', runs over PDF pages 13 to 15 with annual totals on page 15 (134.5, 270, 14, 339.1, 235.1, 345 and 135.5 MWac for 2015 to 2021, under a note that the 2015 to 2021 list 'remain[s] as stated'); the 2021 total is a misprint for 153.5, the sum of its three projects and the figure in the 2022 report. The zero is ours, not printed: it rests on the grid-selling capacity the ministry reports, 1,498 MW in the 2022 report (https://www.memr.gov.jo/ebv4.0/root_storage/en/eb_list_page/annual_report_2022.pdf) and the 2023 report (same path, annual_report_2023.pdf) and 1,478 MW on PDF page 12 of the 2024 report, and on the 2022 report's project table also stopping at 2021; the 20 MW fall in 2024 looks like reclassification (a new 10 MW biogas and hydro line) rather than a closure. The 345 MWac peak is a commissioning year for plants contracted in earlier rounds, so it overstates the pace at which new partnerships were being signed in 2020. The table mixes private independent power producers (Masdar's Baynounah 200 MW, ACWA's Mafraq plant, the Tafila wind farms) with plants financed by Gulf, EU and Spanish grants and loans, including an Air Force plant, so not every megawatt is a partnership; we did not separate them. The reports do not say why additions stopped. What kept growing was subscriber-owned capacity under net metering and wheeling, 1,079 MW in 2022 and 1,320.8 MW in 2024, which is private investment without a public counterpart contract. The erosion is in the frequency of new joint deliveries, not in the stock, which still operates. The ministry reports on its own programme, and the stop may be a pause rather than an end.
๐ฐ๐ฟKazakhstansince 2015still operatingbears on Public-private collaboration
By the end of 2025 Kazakhstan's central and local governments had signed 1,257 PPP and concession contracts worth 3.6 trillion tenge, of which 697 were in implementation, 1,240 were local projects mostly in social services, health, education, energy and utilities, and 17 national projects accounted for more than 1.8 trillion tenge.
The state set up a national PPP operator and let regional akimats sign PPP contracts for local social infrastructure, with the operator appraising projects, ranking regions each year and standardising contract documents.
Preconditions
A PPP law that lets local governments commit future budget payments to private partners
A national operator to appraise projects and publish the portfolio
Budget room for long availability-payment obligations, which is what later pushed policy towards fewer, larger projects
Limits: The figures come from the state-owned PPP Center reporting on the portfolio it promotes, at a briefing of the government's Central Communications Service; the page carries no date beyond 'results for 2025' and no table, and the Center's own homepage gives a different, later count (1,438 contracts and 3.8 trillion tenge as of 1 September 2026) that we could not reconcile with the briefing's 1,257. The gap between 1,257 signed and 697 in implementation is not explained: it mixes completed, not yet started and terminated contracts, so the share of contracts that failed is unknown. Most contracts are small local projects in social sectors, and the Center's PPP White Book describes availability payments from the budget as a standard repayment route, so many contracts may be deferred public procurement rather than shared risk; the briefing gives no split by payment mechanism. The briefing itself says the number of new contracts is falling while average project size rose from about 6 billion tenge before 2022 to 18 billion in 2022 to 2025, a policy shift away from the small local contracts that built the count. The start year is that of the Law on Public-Private Partnership of 31 October 2015, as cited in the Center's PPP White Book; the cumulative count also includes concessions signed under earlier law.
๐ท๐บRussiasince 2019still operatingbears on Public-private collaboration
The Bank of Russia's Faster Payments System, which commercial banks connect to so that customers can pay any account by phone number or QR code, processed more than 3 billion operations worth 14.4 trillion rubles in 2022 with 213 banks connected, and 18.3 billion operations worth 103 trillion rubles in 2025.
18.3 billion operations, operations made through the faster payments system in the year, 2025. 3 billion operations (more than), operations made through the faster payments system in the year, with 213 banks connected, 2022. Bank of Russia, Faster Payments System key indicators, Q4 2025, retrieved 2026-10-02.
The move, and what it needed
The central bank built a national instant-payment switch addressed by phone number, capped what banks may charge individuals, and had commercial banks connect it to their own apps, so that a transfer or QR payment works across all connected banks.
Preconditions
A central bank with its own retail payment operator (NSPK) able to build and run the switch
Regulatory power over every bank's fees and app offerings
Mobile banking already in near-universal use, so the system only had to be added to existing apps
Where else it has been used: The same move as Brazil's Pix and Bolivia's interoperable QR (bol-qr-bcb); in Russia, sanctions removed the foreign card schemes that compete with it elsewhere.
Limits: The 2025 figures (18.3 billion operations, 103 trillion rubles, 1.4 and 1.5 times 2024) are in the text of the Bank of Russia's quarterly indicator page; the 2022 figures, the 213 connected banks and the comparison with 2019 to 2021, which dates the start, are on the Q4 2022 page (https://www.cbr.ru/analytics/nps/sbp/4_2022/). The pages are HTML text over charts, with no downloadable series. The central bank owns the system, sets its tariffs and reports on it, so this is the operator grading itself. The collaboration is asymmetric: the Bank of Russia and its payment-card subsidiary NSPK built and run the infrastructure, and banks connect and offer it in their apps; the indicator pages do not say how far connection was voluntary, and the record does not claim it was. Growth after 2022 coincided with the exit of Visa and Mastercard and with sanctions on Russian banks, which pushed users towards domestic rails, so the count reflects a shock as much as the design. The values are nominal rubles in a period of high inflation. Russia's infrastructure PPP and concession programme is not covered by this record.
๐น๐ณTunisiasince 2024still operatingbears on Public-private collaboration
Tunisia's national sanitation utility ONAS launched procurement of its first two ten-year regional concessions in September 2017; the south-east concession started in June 2024 and served 797,000 people by July 2026, and the North Tunis concession entered into force only on 9 July 2026 after the original consortium had to be reorganised.
The national sanitation utility bundled the operation, maintenance and rehabilitation of whole regional networks and treatment plants into ten-year performance-based concessions, procured with IFC transaction support and paid from a World Bank loan, replacing short task-based management contracts.
Preconditions
An inter-ministerial decision (August 2016) approving a performance-based concession structure for a public utility
External finance to guarantee payments to operators while the utility's own tariffs covered only part of its costs
A concession unit inside the utility able to supervise operators, which the project had to build
Limits: The 797,000 and the North Tunis entry into force are on pages 2 and 3 of the September 2026 status report, which says the count derives from the south-east lot only and is 92 percent of that lot's end target; the end target for both lots is 2,028,000 people by July 2030. The south-east concession with SCAST was 'scheduled to start on Jun 10, 2024' in the June 2024 status report, and the 783,000 beneficiaries the September 2026 report gives for February 2026 confirm it did. The zero signed agreements against a December 2020 target are in the June 2024 status report (https://documents.worldbank.org/curated/en/099061224041597450/pdf/P162957-61a112cc-f5f3-4ef2-ac1e-0e3c14a54b06.pdf, which also attributes the North Tunis delay to the reorganisation of the ADP-led consortium). The 2017 launch is in the project appraisal document of November 2022, paragraph 13. The World Bank finances the operators' payments through this loan (the status report lists the operation and maintenance component at a cost of 415.30, in US$ millions), so the concessions are public money paying private operators under performance contracts, not private finance, and the lender is reporting on its own project. 'People receiving services' counts the population in the concession area, not households newly connected: only 3,400 people had gained access to improved sanitation through the project by July 2026. The record shows a partnership delivered at national-utility scale but nine years after procurement began, which is as much evidence of slow delivery as of collaboration.
๐ณ๐ฟNew Zealandsince 2010still operatingbears on Public-private collaboration
The Crown's infrastructure investment company, now National Infrastructure Funding and Financing, co-funded a fibre network built and owned by Chorus, Northpower, Enable and Tuatahi First Fibre, which was completed in December 2022 covering 87 percent of the population in 412 towns and cities, with NZ$1.75 billion of government funding in a programme costing more than NZ$5.5 billion.
A Crown-owned company invested in regional fibre companies on the condition that they build an open-access, wholesale-only network to a fixed rollout plan, and recovered its investment as the companies sold connections, recycling the repayments into further extensions.
Preconditions
Structural separation of the incumbent telecoms firm, so that the main partner sold wholesale only
Local lines companies able to bid for regions alongside the incumbent
A Crown investment vehicle with commercial terms, not a grant ministry
Where else it has been used: Australia chose a fully public National Broadband Network for the same goal at the same time; Singapore's Next Gen NBN used a similar regulated open-access model.
Limits: The Crown company that ran the programme reports on itself, and the page is promotional ('world leading', 'on time and under budget'); no audit figure for cost against the original budget was checked. The 87 percent is coverage (premises passed), not use: the same page gives uptake as 73 percent of UFB premises passed at 30 June 2025. The total cost above NZ$5.5 billion includes partner co-funding, so the private share is inferred, not stated, and NZ$390 million of the Crown's investment was repaid and recycled into extensions, which the page treats as reinvested rather than spent twice. The page does not describe how partners were chosen or on what terms; the corporate history (Crown Fibre Holdings, then Crown Infrastructure Partners, then National Infrastructure Funding and Financing) and the role of the incumbent's structural separation are background not shown on this page. The page has no date stamp; the uptake figure dates it to after June 2025.
๐ง๐ฉBangladeshsince 2010still operatingbears on Public-private collaboration
Bangladesh's Public Private Partnership Authority listed 81 projects in its portfolio as last updated on 19 August 2026, of which four were in operation (including Patenga Container Terminal, the NIKDU and Chittagong haemodialysis centres and two Info-Sarkar 3 rural broadband components), three were partly operating while still under construction (among them the Dhaka Elevated Expressway), and 10 were under construction.
7 of 81 portfolio projects (4 operational, 3 partial operation and construction), ppp projects in operation, fully or partly, 2026-08-19. 10 projects, ppp projects under construction, 2026-08-19. Public Private Partnership Authority Bangladesh, PPP projects list, retrieved 2026-10-02.
The move, and what it needed
A dedicated authority under the head of government screens line-ministry projects, runs procurement with technical assistance and viability gap funds, and takes them to the economic cabinet committee for approval and award.
Preconditions
A PPP law (2015) and rules for viability gap funding and unsolicited proposals
Line ministries willing to give up projects they could fund from the development budget
Investors willing to take Bangladeshi currency and land risk, which is the precondition the low operating count suggests is scarce
Limits: The counts are ours, tallied from the stage column of the authority's project list, which the page says was last updated on 19 August 2026 and which was read on 2026-10-02; the page gives no totals, and it reports stages, not dates of financial close or of operation. Two of the four operating entries are components of one rural broadband project (Info-Sarkar 3), so the operating count overstates distinct partnerships by one. The authority promotes its own portfolio. The headline is how little has reached operation: 7 of 81 projects after some 16 years, with 24 still at feasibility study and 6 at conditions-precedent stage, so the record documents scale of activity more than delivery. Several listed projects are land leases for commercial development (shopping malls, textile mills on state land), not public services. The Dhaka Elevated Expressway opened in parts against long delays. A 2026 Invest Bangladesh Act appears on the same site and may reorganise the authority; its Bangla text could not be read in this session, so the record does not say whether the authority continues in its current form.
๐ฑ๐ฆLaossince 2005still operatingbears on Public-private collaboration
The Government of Lao PDR granted a build-operate-transfer concession for the 1,070 MW Nam Theun 2 plant to the Nam Theun Power Company, in which it held 25 percent through Lao Holding State Enterprise; the plant was completed in 2010 in 58.5 months, 4.5 months over plan, at US$1,308 million against a US$1,250 million base estimate (US$1,450 million with contingencies), and paid the government US$180 million in net revenues from 2010 to 2017.
The government granted a concession running to 2034 to a special-purpose company, took a minority stake through a new state holding company financed by multilateral loans, and relied on a long-term export power purchase agreement with a Thai buyer that made the debt bankable, with royalties, taxes and dividends earmarked to poverty programmes under a revenue management agreement.
Preconditions
A creditworthy foreign buyer for most of the power under a long-term contract
Multilateral guarantees covering political and regulatory risk that commercial lenders would not take
A government willing to accept an external revenue management and audit regime
Limits: The World Bank financed and guaranteed the project (an IDA grant and a US$50 million partial risk guarantee), so the review is the lender's evaluation unit grading a flagship; it rates the outcome Satisfactory. The figures are in a PDF (cost and financing in the project data section, completion time and capacity and the revenue figures in the efficacy section). The US$180 million is net of nothing the review can confirm: it notes that the US$30 million a year target should be reduced for the government's repayments of the ADB and EIB loans that bought its 25 percent stake, information the Bank team could not provide, so the true net figure is uncertain. Revenue was about US$25 million a year, below the appraisal's US$30 million for the early years, and the much larger US$110 million a year expected after 2020 is a projection, not a result. Equity came from EDF, Italian-Thai Development and EGCO, debt from 14 commercial banks and multilaterals; the delivery is a foreign-led consortium that the state contracted and part-owns, and most of the output is exported to Thailand rather than serving Lao consumers. The resettlement and environmental programme, which the project was also judged on, is not covered by these numbers. The start year, 2005, is our inference from completion in 2010 after 58.5 months. The review also records that by 2014 a significant portion of NT2 revenues were being deposited directly into the Treasury rather than through the agreed special account, which weakened the earmarking of revenue to poverty programmes.
๐น๐ฟTanzaniasince 2007dismantledbears on Public-private collaboration
Tanzania concessioned the central railway network for 25 years to Tanzania Railways Limited, 51 percent owned by a consortium led by Rail India Technical and Economic Services, which took over operations in October 2007; freight fell from 470,000 tons in 2008 to 237,000 tons in 2010, and after a dispute the RITES shares were transferred to the government in 2011, leaving the operator fully state owned four years into the term.
The government unbundled the railway into an asset-holding company and an operating company, sold 51 percent of the operator to a foreign rail consortium on a 25-year concession, and kept the track; when traffic, investment and maintenance fell short, the state bought back the private shares and ran the operator itself.
Preconditions
A base load of bulk freight large enough to make a concession pay, which the central line had lost before the handover
Contract terms that oblige and finance track maintenance, which here sat between an asset holder and an operator
A short bidding and handover period, so traffic does not leak to road while the deal is negotiated
Where else it has been used: The Beira railway concession in Mozambique, signed in 2004 and terminated on 7 December 2011 according to its World Bank completion report, ended the same way in the same year (see moz-beira-railway).
Limits: The figures are in a PDF, on pages 16 and 17 of the appraisal document (printed pages 4 and 5), written by the World Bank as it prepared a new loan to the now state-owned operator, so the publisher had an interest in describing the concession as a failure. The PAD gives no date in 2011 for the share transfer and calls it amicable, and it is not consistent on the year: paragraph 16 (page 17) puts the transfer in 2011 and an annex (PDF page 116) says TRL took over in 2011, but another annex paragraph (PDF page 117, Annex paragraph 28) says the concession was terminated in 2010; the concession was not formally terminated by a court or an arbitral award in this source, so what ended was private participation, not the railway. The earlier completion report for the Central Transport Corridor Project (ICR00001299, June 2010, page 9) records the concession as performing poorly and the parties discussing termination, and adds (paragraph 94) that most of the traffic had already moved to road during the two years of bidding delay, from 1.2 to 0.5 million tons a year, before the concessionaire took over: the decline therefore began under public operation. The 2008 to 2010 fall also includes the 2009 to 2010 floods in the Kilosa to Gulwe section, which the PAD says halted services between Dar es Salaam and Dodoma, so not all of it is the concessionaire's doing. 2008 is the first full year of concession operation, not a demonstrated peak. The 1.5 million tons TRC carried in 2002, which the PAD also gives, is the public railway before the concession and is not used as the peak here.
๐ฟ๐ฒZambiasince 2003dismantledbears on Public-private collaboration
Zambia concessioned its national railway, after a decade in which traffic halved from 1,072 million net ton kilometres in 1993 to 540 million in 1999, to the private operator Railway Systems of Zambia, which took over on 3 December 2003 and raised traffic by about 30 percent in net ton kilometres in 2004 against 2003; the World Bank later reported that the network and service kept deteriorating and that the concession was cancelled in September 2012.
After two decades of falling traffic under the state railway, the government retrenched staff with World Bank money and let one private consortium run all freight and passenger services and the track under a vertically integrated concession; the operator lengthened hauls and paid fees for a few years, then a later government judged the network was still deteriorating and took the railway back.
Preconditions
A concession agreement with enforceable investment and maintenance obligations and a monitoring body able to apply them
Bulk mining freight that the railway can win back from road
Political agreement across governments that private operation is the policy, which did not survive here
Limits: The cited completion report was written two years into the concession and supports the start and the early gains only; it says the 30 percent rise came with a fall in tonnage as the operator dropped short hauls, and that the borrower disputed the concessionaire's fee compliance and Parliament set up a review in December 2005. The cancellation rests on a second World Bank document, the Country Partnership Strategy for FY13 to FY16 (Report No. 75089-ZM, February 2013, http://documents.worldbank.org/curated/en/182591468178437607/pdf/750890CORRIGEN0IC00IDA0R20130002703.pdf, PDF page 22, paragraph 49), which says the rail network condition and service continued to deteriorate and the concession was cancelled in September 2012 with a commitment to recapitalise the railway, and elsewhere (page 11) calls it a reversal of the railway concession that stakeholders saw as policy incoherence. Neither document gives traffic for the later concession years, so the record shows a delivered early gain and a cancellation, not a measured decline from a peak. Both documents come from the lender that financed the concession process and argued for privatisation. The figures are in PDFs (the 1993 and 1999 traffic on page 5, the fees on page 10, the 2004 traffic gain on page 19, the 20-year term on page 43). The concession was originally 20 years with possible extensions (Annex 10), so it ran about nine years of a 20-year term, a figure computed here.
๐ฌ๐ณGuineasince 1989dismantledbears on Public-private collaboration
Guinea ran its urban water service from 1989 under a ten-year lease in which the public asset company SONEG owned and financed the network and the privately led operator SEEG ran it, but after renegotiation failed the private operator withdrew in December 2000, the government decided in July 2003 to keep the service public, and revenue collected from private customers fell from US$5.8 million in 1998 to US$1.7 million in 2005.
The state split the urban water utility into a public asset-holding company that owned and financed the network and a private-led operating company that ran it under a lease and was paid from tariffs; when the lease came up for renewal, tariffs were frozen, the private shareholder saw too much risk and left, and the public company took over operations it was not prepared to run.
Preconditions
Tariffs set high enough and indexed so the operator's revenue does not erode in real terms
Renegotiation started well before the lease expires, with a deadline
A public asset company able to run operations itself if the private partner leaves
Limits: The figures sit in a PDF (page 9, printed page 4; the ten-year lease ending in 1999 is on page 6). The World Bank financed the lease from its first project in 1989 and rated this follow-on project and the borrower unsatisfactory, so it has a view on who was to blame. The revenue figures are in US dollars, and the report says tariffs were frozen until 2005 and fell from US$0.86 to US$0.21 per cubic metre in real dollar terms, so much of the fall in dollar collections is a currency and tariff effect, not only worse collection; the same paragraph says SONEG's net income went from US$2.9 million in 1998 to a loss of US$0.5 million in 2005. 1998 is a year chosen by the report, not a demonstrated peak. The report also says 24 hour service and compliance with water quality standards, achieved under the earlier operation, were lost and service became intermittent, with no figure for when. SEEG's private shareholder is named only as SAUR, in passing, and the share structure is not given. The 1989 start is inferred from the ten-year lease ending in 1999 and the first supporting credit of May 1989; the report gives no signing date. The lease ended at the end of its term after two years of failed renegotiation and contract extensions, so this is a partnership that was not renewed rather than one terminated for breach.
๐ฒ๐ฟMozambiquesince 2000dismantledbears on Public-private collaboration
Mozambique's water asset fund FIPAG leased Maputo's water service from 2000 to the private operator Aguas de Mocambique, whose lead shareholder SAUR pulled out in December 2001 and offered its stake to Aguas de Portugal for one euro; under the lease Maputo connections rose from 77,793 at the end of 2000 to 109,932 in March 2009, and in 2010 FIPAG bought out the Portuguese shareholder, leaving the operator primarily publicly owned.
The state kept the water assets in a public fund and leased operation of the capital's system to a private consortium paid from tariffs, with a regulator able to order tariff reviews; when the lead foreign operator left and its successor underperformed, the asset fund bought the operating company itself and kept the lease structure with a public owner.
Preconditions
Bids priced on reliable information about network condition, which the security situation prevented here
A lead operator whose corporate strategy commits it to the country for the contract term
A regulator with power to order tariff reviews, which existed (CRA) but could not save the original consortium
Limits: The connection counts are in a PDF (page 7, the results table; the SAUR withdrawal on page 18, paragraph 22), reported by the lender that designed and financed the lease. The end of private participation rests on a second World Bank document, the completion report for the Greater Maputo Water Supply Expansion Project (Report No. ICR00005527, February 2022, http://documents.worldbank.org/curated/en/306811653587372748/pdf/Mozambique-Greater-Maputo-Water-Supply-Expansion-Project.pdf, page 22, paragraph 37), which says the operator's performance fell well below expectations, relations with FIPAG deteriorated, FIPAG bought out the Portuguese partner's roughly 85 percent stake in 2010, and the renamed company, Aguas da Regiao de Maputo, has run the lease since then as a primarily public company, with the lease extended to 2019 and private partners for the regional utilities only scheduled for 2022. The two reports do not agree on the lease length (15 years in 2009, 14 years in 2022). The record shows growth in connections under the lease and the end of private ownership, not a fall in service: connections were not shown to decline after the buyout, so what eroded is the partnership, not access. The 2009 report rates the private operator's performance poorly through much of the period: the 2000 floods, SAUR's underbidding, and two years of renegotiation delayed the programme by about four years, and supply in 2009 averaged 14 hours a day against a target of continuous supply. Lease revenue to FIPAG stayed far below expectation (2022 report, paragraph 52).
Since 2001, Cadastro รnico has identified and characterized low-income families across Brazil and has become the entry point for more than 40 federal, state and municipal social programmes.
One shared identity and needs record lets multiple programmes see the same household, reducing the need for each programme to build its own eligibility system and making the state more legible to people it serves.
Preconditions
A national identity key that can link people across programmes
Municipal service points able to collect and update household information
Rules that let several programmes use the same record while preserving eligibility distinctions
Limits: A registry and its programme count show administrative reach, not whether citizens trust the state or whether benefits arrive accurately and on time. The source is the ministry responsible for the registry, and coverage depends on families keeping their records current.
๐ง๐ทBrazilsince 2004still operatingbears on Trust in public institutions
Launched in 2004, Portal da Transparรชncia has spent 20 years publishing federal revenue, expenditure, personnel and transfer information so citizens can inspect how public money is used.
The government turns its own administrative records into a public interface, so oversight can begin with the same transaction detail that the bureaucracy uses to execute and audit spending.
Preconditions
Administrative records detailed enough to publish at transaction level
A legal duty to disclose public spending
An oversight body able to maintain the interface and respond to findings
Limits: Publication creates the possibility of scrutiny, not evidence that people use the data or trust the institutions being scrutinized. The portal mainly covers federal government execution and excludes some state-owned and non-federal activity.
๐ฐ๐ทSouth Koreasince 2023still operatingbears on Trust in public institutions
South Korea's MyData policy gives people a route to request and direct the transfer of personal data between services, with the Personal Information Protection Commission describing the scheme as a way to give people control over data portability after the March 2023 legal amendment.
The state gives people a legal route to move their own data between providers, shifting data access from a private platform choice toward a user-directed right.
Preconditions
A privacy authority able to define consent and portability rules
Interoperable data formats and secure transfer channels
Providers willing and able to respond to a person's authorised request
Limits: A portability right shows a legal design for user control, not whether people understand it, use it safely or receive better services. The policy notice describes the legal framework and is not an independent assessment of outcomes.
๐ธ๐ฌSingaporesince 2006still operatingbears on Court case clearance rate
Singapore's State Courts disposed of 21,816 civil cases against 20,358 filings in 2023, a 107% civil clearance rate; its Community Justice and Tribunals courts recorded a 99% clearance rate in the same year.
107 % of incoming civil cases, state courts civil-case clearance rate, 2023. 99 % of incoming cases, community justice and tribunals clearance rate, 2023. Singapore Judiciary, Caseload statistics 2023, retrieved 2026-08-31.
The move, and what it needed
Singapore Courts publish workload by court and case type, set clearance as a management measure and combine procedural reform with specialised tribunals and dispute-resolution pathways. The public series makes backlogs visible while the court system can shift suitable disputes into faster channels without pretending that every case is the same.
Preconditions
A judiciary with reliable case-management data and authority to redesign procedure across court tiers
Registries, judges, mediators and tribunals that can route cases by complexity rather than use one queue for all disputes
A public accountability norm that treats unresolved caseload as an operational problem while protecting judicial independence
Where else it has been used: Court-clearance management travels when administrative data, triage and dispute resolution reduce stock without turning speed into the only definition of justice.
Limits: Clearance is disposals divided by filings in the same year, so it can exceed 100% when courts work through earlier stock; it does not show timeliness, fairness, legal-aid access, enforcement or the quality of decisions. The Singapore figures cover court workloads and are not a comparable cross-country civil-justice index without harmonised case definitions.
๐ต๐นPortugalsince 2013still operatingbears on Court case clearance rate
The CEPEJ Portugal country fiche records a 96.1% clearance rate for civil and commercial litigious cases at first instance in 2023, alongside 101.1% at second instance and 105.5% at the Supreme Court; first-instance disposition time was 267 days.
Portugal's courts feed case-flow data into a recurring national and European monitoring framework, making incoming, resolved and pending civil cases visible by instance and allowing backlog pressure to be compared over time.
Preconditions
Court case registers with consistent definitions of incoming, resolved and pending matters
A justice statistics authority that can aggregate courts without changing the underlying case counts
A management response that uses the series to allocate judges, staff or procedural capacity
Where else it has been used: Clearance monitoring travels when case definitions stay stable and managers act on the series; a published percentage without backlog, duration and instance detail can reward fast closure rather than justice.
Limits: The clearance rate is an administrative throughput measure for one year and one case class, not a judgment-quality, access-to-justice or public-trust measure. Disposition time is a theoretical estimate derived from pending and resolved cases, and rates differ sharply by court instance; CEPEJ republishes national data and the fiche cannot identify which reforms or court-management practices caused the result.
Singapore gave every citizen aged 25 and over an individual training account in 2016 and keeps topping it up, and in 2024 260,000 Singaporeans spent credit on courses, up 35 percent on the year before.
The state deposits money in a named account each citizen owns, publishes a course registry with subsidies, and lets individual demand rather than employers decide what gets learned, with top-ups aimed at mid-career workers.
Preconditions
A single national identity and payments rail to run the accounts on
A subsidised training market large enough to spend the credit in
Fiscal surpluses to fund top-ups a decade running
Where else it has been used: France's Compte Personnel de Formation is the closest sibling. Both show the same failure mode of low-value courses chasing the credit, which each has had to prune from its registry.
Limits: Usage is the programme reporting on itself, and a usage count says nothing about whether the training raised skills or wages. About 260,000 users in a year is under a tenth of the adult citizen population, and separate reporting notes most citizens had not touched their top-up. The record evidences a durable mechanism for adult learning, not a transformation of participation.
๐ซ๐ฎFinlandsince 1928still operatingbears on Adult learning participation
Finland runs the world's most used public library system under a statutory duty dating to 1928, and in 2024 Finns borrowed 85.4 million items, 15.3 per inhabitant, and visited libraries 50.7 million times.
The state made libraries a statutory municipal duty with national subsidy, kept them free by law including reservations and inter-library loans, and municipalities compete on quality, with the flagship in Helsinki placed opposite parliament as a statement.
Preconditions
Near-universal literacy and a reading culture that predates the statute
Municipal finance strong enough to fund a non-mandatory-feeling service
A language community small enough that the state subsidises publishing too
Where else it has been used: The Nordic neighbours run the same model at slightly lower intensity. Elsewhere the statutory free-of-charge rule is the part that erodes first, and with it the usage.
Limits: Borrowing is reading, not structured adult learning, so the record sits one step from the indicator's construct, which asks about participation in education and training; it is filed because the library statute explicitly assigns libraries a lifelong-learning duty and no better-measured Finnish delivery covers the gap. Loans per head are also falling slowly here as everywhere, and the world-leading claim depends on which countries publish comparable statistics.
๐ง๐ทBrazilsince 1951still operatingbears on Research citation impact
Founded in 1951, CAPES built a national postgraduate funding and evaluation system, and by 1995 Brazil's system had more than 60,000 students across over 1,000 master's and 600 doctoral courses.
60,000 students, postgraduate students in the national system, 1995. CAPES, Histรณria e missรฃo, retrieved 2026-08-29.
The move, and what it needed
A federal agency combines funding with peer evaluation and a national standard, so postgraduate programmes coordinate around a common definition of quality instead of growing as isolated local institutions.
Preconditions
A stable public agency with authority over funding and evaluation
A research community willing to participate in a shared assessment system
Long-term public support for postgraduate training
Limits: Student and course counts show the scale of the postgraduate system, not the quality or international impact of its research. CAPES reports the institutional history and its own programmes, while citation impact requires a separate bibliometric source.
๐ง๐ทBrazilsince 1937still operatingbears on Research citation impact
Created in 1937, INEP built a standing education statistics and census system that lets federal, state and municipal actors see schools, students and policy results across Brazil, while its historical archive now spans 218.68 linear metres.
A permanent statistical institution gives education policy a memory and a common set of objects, so schools, students and results can be compared across territories and governments.
Preconditions
A mandate to collect comparable information across levels of government
Schools and secretariats able to report through a shared system
Analytical staff and archival practice that preserve series across reforms
Limits: An archive and a census show that a state can record itself, not that the resulting policy improves learning or research quality. The data system also depends on reporting by schools and education secretariats, and its history includes institutional redesign.
๐ฎ๐ฉIndonesiasince 2020still operatingbears on Adult learning participation
Indonesia's Kartu Prakerja skills programme reached 18.9 million beneficiaries across the country in its first five years, with 347 accredited training providers offering 2,085 online and offline courses in 58 districts and cities in the 2024 report.
Indonesia used an end-to-end digital application, training marketplace and payment rails to move a skills benefit from central government to workers, while accrediting private providers and changing the scheme after the pandemic.
Preconditions
A digital identity and payment channel that can verify applicants at scale
A quality-assurance process for many public and private training providers
Labour-market feedback strong enough to refresh the course catalogue
Where else it has been used: The design has been studied by Cambodia, Thailand, Malaysia and the Philippines. The transferable piece is the operating marketplace and evaluation loop, not the beneficiary total by itself.
Limits: Beneficiary and course counts show the reach of a training marketplace, not whether skills were retained, led to better work or were relevant to each participant. The report is produced by the programme manager, and the 58-district course figure is not the same as equal course availability in every district or a national adult-learning rate.
๐ซ๐ทFrancesince 2015still operatingbears on Adult learning participation
France's Compte personnel de formation generated 1,387,500 training starts in 2024, with 1,232,100 people entering a CPF-funded course through the national account and platform.
France gave individuals a portable training entitlement, then put the account, catalogue, identity checks and payment flow into one national platform managed by the Caisse des Dรฉpรดts. The design lets workers choose a course while public employment services and employers can add targeted funding.
Preconditions
A national entitlement and contribution system that records rights across employers and employment spells
A trusted digital identity, provider catalogue and payment system that can verify courses and prevent fraud
Quality assurance and co-funding rules that keep a large open marketplace from becoming a directory of low-value training
Where else it has been used: Individual learning accounts travel when entitlement portability is paired with provider quality controls and guidance; a wallet without trusted course information can increase transactions without increasing skills.
Limits: Training starts and participant counts show use of a public financing and matching platform, not course completion, learning quality, job mobility, wage gains or equal access. DARES notes major catalogue and contribution-rule changes during 2024, including the motorcycle-licence expansion and a โฌ100 co-payment, so year-on-year totals are not a pure measure of underlying learning demand.
๐ต๐นPortugalsince 2017still operatingbears on Adult learning participation
Portugal's ANQEP reported 1,025,569 cumulative Qualifica registrations by December 2023; 90 percent received a training or recognition, validation and certification referral, and 510,113 registrants obtained some certification.
1,025,569 registrations, cumulative qualifica registrations, 2023-12. 510,113 people, registrants receiving a certification, 2023-12. ANQEP, Plano de Atividades 2024, retrieved 2026-08-31.
The move, and what it needed
A national network of Centros Qualifica gives adults a common entry point for counselling, recognition and validation of prior learning (RVCC), formal training and a portable qualification record, linking fragmented providers to a visible pathway.
Preconditions
Local centres with counsellors and assessors able to validate work-based learning
A qualification framework and records that employers and schools can understand
Funding and referral links to training providers, employment services and employers
Where else it has been used: Adult-learning systems travel when advice, prior-learning recognition and funded courses form one route; registrations without follow-through can otherwise become a throughput statistic rather than capability.
Limits: These are cumulative programme reach and certification totals, not an annual participation rate, completion quality, learning gain or employment and wage effect. The count can include partial certifications and administrative registrations, and it does not identify how many participants would have qualified without Qualifica; ANQEP reports the programme's own administrative data.
๐ต๐นPortugalsince 2006dismantledbears on Adult learning participation
ANQEP's national review records 1,568,490 adult enrolments in Novas Oportunidades between 2006 and July 2011; after the initiative's peak and a period of disinvestment, Portugal launched Qualifica in 2017 and rebuilt a network of more than 300 Centros Qualifica.
1,568,490 adult enrolments, adult enrolments in novas oportunidades, 2011-07. 300 centres (more than), centros qualifica in the replacement network, 2017. ANQEP, Plano Nacional de Literacia de Adultos, retrieved 2026-08-31.
The move, and what it needed
Novas Oportunidades scaled adult qualification through recognition and validation of prior learning (RVCC), adult education and training (EFA) routes and a national delivery network; when political and fiscal support receded, the infrastructure contracted and a successor programme had to rebuild entry points.
Preconditions
A national qualification framework that can turn work experience into recognized credits
Local centres, assessors and training providers able to process large cohorts
Long-horizon funding and cross-party continuity, because adult-learning capacity is slow to rebuild after a stop
Where else it has been used: The interruption travels as a warning: mass adult-learning systems need continuity and credible standards, or a high-enrolment push can leave a capability gap when the delivery network is withdrawn.
Limits: Enrolment is a reach measure, not completed learning, quality, labour-market effect or value for money. The review's 1.57 million figure covers a high-activity period, while the later 300-plus-centre figure describes the replacement network rather than retained Novas Oportunidades capacity; the record evidences interruption and rebuilding, not a causal estimate of the policy's return.
๐ฎ๐ณIndiasince 2017still operatingbears on Adult learning participation
India's SWAYAM platform offered more than 20,370 courses and recorded more than 65 million enrolments by 12 August 2026 as a national channel for flexible higher-education and skills learning.
The Ministry of Education assembled university courses on one open platform, letting learners study remotely and connect course participation to formal credit and certification pathways.
Preconditions
Universities willing to publish reusable course material to a common technical standard
A credit and assessment framework that makes online study count beyond the platform
Connectivity and learner support good enough for participation outside campus
Where else it has been used: Open course platforms travel when enrolment can convert into recognised credit or skills; a large catalogue without completion and credential data measures supply more than continuous learning.
Limits: Cumulative enrolment is not a share of adults participating in learning: one person can enrol repeatedly, and the release does not provide active users, completions or the adult/student split. Course supply and enrolment show a national learning channel, not learning gains, labour-market outcomes or equal access to broadband and devices.
๐ฟ๐ฆSouth Africasince 2000still operatingbears on Adult learning participation
South Africa's Sector Education and Training Authorities reported 45,146 learnership completions, 83,578 skills-programme completions and 99,778 workplace-based learning opportunities for the 2024 post-school education and training intake.
Sector authorities levy and coordinate employer-linked training, then place learners into learnerships, skills programmes and workplace experience tied to occupational demand rather than leaving adult learning entirely to formal colleges.
Preconditions
A statutory levy and sector bodies with a mandate to fund training beyond public campuses
Employers willing to host learners and assess workplace competence
Occupational standards and certification routes that make short programmes legible to workers and firms
Where else it has been used: Workplace learning travels when employers share the cost and the credential is recognised across firms; opportunity counts without learner follow-up cannot show whether skills actually accumulated.
Limits: The ministerial statement aggregates 21 SETAs and counts opportunities and completions, not a harmonised share of adults participating in formal or non-formal learning. It does not report unique learners, demographic reach, completion quality, employment after training or the share of participants who were already in education; the figures are government-reported rather than independently audited here.
๐น๐ทTurkeysince 2024still operatingbears on Adult learning participation
Tรผrkiye's Ministry of National Education reported 7,909,495 participants in public adult-learning programmes in 2024, including 6,493,700 attending in person and 1,415,795 using the HEMBA online platform; 5,305,688 earned a course certificate.
A dense network of public education centres offered locally requested courses while HEMBA moved a subset of the same provision online, allowing adult learners to enter through either a physical centre or a national platform.
Preconditions
A nationwide network of public adult-education centres
A course catalogue and credential system that can be updated centrally
Digital infrastructure and instructors able to support online as well as in-person delivery
Where else it has been used: Adult-learning networks travel when local access and a common credential reinforce one another; participation totals need unique-learner and outcome data before they become a capability score.
Limits: The report counts course participations rather than unique adults, so repeat learners can appear more than once. Certificates indicate completion of a course, not labour-market value or durable skill acquisition, and the ministry's administrative data are not independently audited here.
๐ฐ๐ชKenyasince 1978operating below its peakbears on Adult learning participation
Kenya's state adult education programme, run since the 1978 presidential literacy decree through adult education centres in every county, enrolled 330,340 learners in 2014 and 165,552 in 2025, after a low of 125,402 in 2023, according to the Kenya National Bureau of Statistics.
The presidency decreed a national literacy programme, and the education ministry, and later a social protection directorate, ran it through state-paid instructors in community adult education centres in every district. Enrolment fell by half over a decade as the function moved between ministries and county returns degraded.
Preconditions
A large adult population that had missed primary school
A ministry department with paid instructors at district level
A political mandate that framed literacy as a national campaign rather than a welfare service
Limits: The metric is a head count of learners in adult education centres, mostly basic literacy, not the share of adults in formal or non-formal education and training that the indicator defines: 165,552 is well under one percent of Kenya's adult population, and the count leaves out TVET, university, employer training and private courses entirely. The peak is only the highest year in the series KNBS publishes from 2010 onwards (252,553 in 2010, 330,340 in 2014). The 1980s literacy campaign almost certainly enrolled more, so the true loss from the programme's peak is understated, not overstated. The series is administrative and badly maintained. KNBS revised 2015 from 246,215 (Economic Survey 2016) to 306,228 (Economic Survey 2018). The 2025 figure is provisional. Several counties report identical numbers in consecutive years (Garissa 3,256 in all four years 2022 to 2025, and Samburu and Laikipia repeated), which points to carried-forward returns. One county, Nairobi, accounts for most of the 2025 rebound, rising from 13,194 to 34,506 in one year. Falling enrolment can partly reflect success, since younger cohorts went through universal primary schooling and need less remedial literacy, so the decline is evidence of lost provision only alongside the fact that the programme was reorganised across ministries and is now run by a social protection directorate. The 1978 start year is the presidential decree reported by the state news agency, not a documented first enrolment. The 2014 peak was itself revised: Economic Survey 2015 published 300,462 (provisional) and Economic Survey 2017 restated it as 330,340; against the earlier figure the fall to 2025 is about 45 percent rather than half.
๐ท๐ผRwandasince 2003operating below its peakbears on Adult learning participation
The Ministry of Education set national literacy targets and counted learners through its school census, but relied on churches, NGOs and donor projects to host and staff most centres. When those partners withdrew, enrolment fell, and no state budget line replaced them.
Preconditions
A national school census that reaches non-formal centres
Religious and NGO networks willing to host classes in every sector
Public funding for instructors that would hold up when donor projects ended, which is the precondition that failed here
Limits: The start year is weak. 2003 is the sector policy that committed the state to adult literacy training, not a documented first enrolment, and adult literacy classes ran before it. A reviewer could fairly set the year elsewhere. The programme never came near its own targets: the 2003 policy implied about 500,000 learners a year. The '2017 peak' is only the high of one published five-year table, and earlier years were not checked. Delivery is mixed: the yearbooks blame part of the decline on NGOs and projects closing and on religious buildings ceasing to host centres. Some of the loss is therefore civil society withdrawing rather than the state, even though the ministry counts and partly funds all centres. The 2024/25 yearbook states the latest fall as 5.8 percent in one place and 5.7 percent in another. A head count of literacy learners is far narrower than the share of adults in any education or training that the indicator defines.
๐ช๐นEthiopiasince 2010operating below its peakbears on Adult learning participation
Ethiopia's Ministry of Education rolled out Integrated Functional Adult Education nationally from 2010/11 and enrolled 6,903,065 adults in 2015/16, but its successor Adult Basic Education Program enrolled 1,392,635 adults in 2024/25, about a fifth of the peak.
The Ministry of Education built a national adult literacy programme run through regional education bureaus and community learning centres, then let enrolment fall by most of its peak as the programme was restructured and conflict removed whole regions from reporting.
Preconditions
Regional education bureaus and kebele-level centres able to recruit facilitators and enrol adults at scale
A national campaign target, here ESDP IV's goal to cut adult illiteracy, that kept enrolment a political priority
Peace and administrative reach across regions, which the Amhara and Tigray conflicts removed
Where else it has been used: Mass literacy campaigns travel as enrolment drives; the Ethiopian series shows the reach depends on keeping the delivery network funded and present after the campaign target lapses.
Limits: The 2024/25 count excludes Amhara and Tigray, which reported nothing because of conflict, and Amhara alone held 43 percent of IFAE enrolment in 2015/16; removing Amhara and Tigray, both absent in 2024/25, from the 2015/16 peak still leaves 3,450,597 against 1,392,635 (2,038,631 with Education Light), a figure we computed from Table 6.2 of the 2015/16 abstract and the ministry does not publish. The two numbers are not one series: the two-year IFAE became a one-year Adult Basic Education programme, and the 2024/25 abstract reports a separate Education Light/Torch track of 645,996 adults, so total adult enrolment that year was 2,038,631. All figures are administrative enrolment reported by regional bureaus to the ministry, with no attendance, completion quality or literacy outcome attached, and enrolment is not the construct's share of adults in education and training. The 2024/25 abstract has no stable PDF URL: it downloads from a button on the ministry's annual abstract page, so a checker must download it there and read PDF pages 63 to 64, printed pages 54 to 55 (sections 5.2 and 5.3, Table 5.1). The 2015/16 figure comes from a separate abstract PDF. The 2024/25 abstract also reports that 13,015,641 adults completed or graduated from IFAL/ABE between 2016/17 and 2024/25, so a shrinking pool of illiterate adults is a competing explanation for part of the fall.
๐ฌ๐นGuatemalasince 1972still operatingbears on Adult learning participation
Guatemala's employer-levy-funded training institute INTECAP reported 490,280 participants in 26,387 vocational training events in 2025, 99.72% of its operating-plan target, through centres in all 22 departments.
Guatemala's Congress gave a joint state-employer institute a compulsory 1% payroll levy, collected by the social security institute, and the institute turned it into a national network of training centres and short courses for workers at every occupational level.
Preconditions
A formal payroll base large enough to fund training through a levy collected alongside social security contributions
Employer representation in governance, so firms accept the levy as their own training budget
Centres in every department, which reach formal workers and leave the large informal workforce mostly outside
Where else it has been used: It follows the Latin American levy-funded training model (SENAI, SENA, INA), and like them it reaches the formal sector far better than informal workers.
Limits: This is INTECAP reporting on itself against its own operating plan. Nobody audits the count. Participants are enrolments in events, not unique people: one worker who takes three courses counts three times, so the figure cannot be divided by the adult population to give a participation rate, which is what the indicator asks for. Most of the volume is short: seminars (202,359) and the ACTA line (218,147) make up about 86% of participants, while the long technical careers (CTO, CTM, CTMS) total about 13,400. Some programmes, such as FORJA and the mid-level technical careers, enrol young people, so not every participant is an adult learner. The text layer garbles the event columns of several national rows, though the total row reads cleanly. The figure covers one year only. No series is given here, so this record cannot say whether participation is rising or falling, and it says nothing about completion, certification or labour-market outcomes.
The Dominican Republic's tripartite training institute INFOTEP, created by Law 116-80 in 1980, registered 1,027,791 participants in training actions nationwide in 2025 and 3,613,098 participants between August 2020 and August 2025.
The Dominican state, employers and unions run INFOTEP jointly under Law 116-80, financed by an employer payroll contribution, and since 2020 the institute expanded its own centres and virtual offer to push enrolment past one million a year.
Preconditions
A compulsory employer training levy that gives the institute a revenue base independent of the annual budget contest
Tripartite governance that keeps employers invested in the course offer
A dense network of own and partner training centres, plus a virtual platform, to reach provinces outside Santo Domingo
Where else it has been used: INFOTEP belongs to the Latin American levy-funded training family with SENAI and SENA; El Salvador dissolved its counterpart, INSAFORP, in 2023, so the same model travelled and was then unwound next door.
Limits: The figures are INFOTEP's own administrative counts, relayed by the presidency's news service from the institute's annual report; the report itself was not opened, because infotep.gob.do refused scripted requests (HTTP 403). A participant is an enrolment in a training action, not a person: someone taking several short courses counts several times, so the number cannot be read as a share of adults trained and is not comparable with survey participation rates. The two figures are hard to reconcile: 3.6 million over five years averages about 720,000 a year, against more than one million in 2025 alone, and nothing on the pages explains the jump, whether a definition changed, or how much of it is short or virtual courses. The counts include young people who may not be adults in the survey sense, and they say nothing about course length, completion, certification rates, or labour-market outcomes. The pages are promotional and were published by the government that runs the institute.
๐จ๐บCubasince 1961delivered and closedbears on Adult learning participation
In 1961 Cuba's national literacy campaign mobilized volunteer teachers, student brigades and worker brigades to teach adults to read across the country. By 30 August 1961, 593,651 people were learning and 64,253 had already been made literate, and the government reported illiteracy reduced to 3.92 percent of the population when the campaign closed.
The revolutionary government closed secondary schools for part of 1961 and sent students, teachers and worker brigades to live with illiterate households, after a door-to-door census of illiterates run through mass organizations. It then built follow-on worker-peasant education so the new readers kept studying.
Preconditions
A census of who could not read, collected house by house through neighborhood and mass organizations
A state able to suspend schooling and mobilize about a quarter of a million volunteers under one command
A single primer and teacher manual that untrained volunteers could deliver
Where else it has been used: Cuba exported the method as 'Yo, sรญ puedo', a radio and video version used in about 30 countries; the exported form replaced the volunteer army with recorded lessons and local facilitators, because the mass mobilization did not travel.
Limits: Both numbers come from commemorative articles on Cuban ministry websites, not from a statistical release. The MINED article says it was taken from the Communist Party's website, and both rest on the campaign's own 1961 records, which nobody independently audited. The 593,651 is a mid-campaign count from August 1961, not the final number taught, and the source gives no denominator of adults. A participation share therefore cannot be computed, and the literacy standard reached was basic, a letter written at the end of the course. The 3.92 percent is a government figure against the total population of the time, while the 23.5 percent baseline the MINED article cites is from the 1953 census, eight years earlier, so the two are not a like-for-like comparison. The campaign was a one-year mobilization, not a standing system, and the follow-on adult schooling it created (Superaciรณn Obrera y Campesina) is not measured here. The case is 65 years old and says nothing about adult learning participation in Cuba today.
๐ฆ๐บAustraliasince 2008dismantledbears on Adult learning participation
Australia's VET FEE-HELP loan scheme, introduced in 2008 to fund higher-level vocational courses, grew from fewer than 3,500 loans in 2009 to 196,108 in 2015, and was abolished and replaced from 1 January 2017 by VET Student Loans, which reached 53,892 students in 2019.
The Commonwealth extended income-contingent loans to vocational diplomas and opened them to private providers with light entry controls; providers and brokers enrolled students at scale, Parliament then legislated the scheme's closure and replaced it with capped loans for approved courses and providers.
Preconditions
An income-contingent loan system administered through the tax office, so the state could finance study with no upfront fee
A contestable training market in which private providers drew public money per enrolment
Regulators (ASQA, ACCC, the department) whose powers and data lagged the growth of enrolments by several years
Where else it has been used: The same demand-driven, provider-paid design produced comparable mis-selling in the United States for-profit college sector and in England's Individual Learning Accounts, which were closed in 2001; each time the fix was to cap the loans and vet the providers, which shrinks reach along with fraud.
Limits: Loans are not learners: the count is loans issued, not adults who studied or completed, and the Library's own note says the 2017 onward figures count students while earlier years count loans or places, so the peak and the successor figure are not on the same unit and the fall cannot be read as a precise percentage. Much of the 2013 to 2015 growth was not learning at all: the Senate Education and Employment Committee's report on the replacement bills (https://www.aph.gov.au/Parliamentary_Business/Committees/Senate/Education_and_Employment/VETStudentLoans/Report/c02) records loan value rising from $26 million in 2009 to $2.9 billion in 2015 alongside 'unscrupulous behaviour by a small number of providers and poor student outcomes', so the peak overstates real adult participation and the reversal is partly a correction of fraud rather than a loss of capability. The loans cover only diploma-level and above vocational courses, a small slice of adult learning; the record says nothing about the national share of adults in training, which the ABS measures separately. The Auditor-General's audit of the scheme (ANAO report 31 of 2016-17) is the primary account of the failure, but the ANAO site did not open from here, so its figures are not used. The source is a secondary compilation by the Parliamentary Library from departmental annual reports, and 2016 figures are missing from it.
๐ซ๐ฎFinlandsince 2001dismantledbears on Adult learning participation
Finland paid an earnings-related adult education allowance from 2001 so that working adults could take leave to study, reaching 30,124 recipients in 2022 and 32,518 in 2023, until Parliament abolished it in May 2024 with a transition under which the allowance could be granted until 31 December 2025.
Finland's tripartite system funded study leave for experienced workers out of unemployment-insurance contributions, administered by the Employment Fund; the Orpo government then abolished it to cut contributions and public spending, and Parliament passed the repeal in May 2024.
Preconditions
An earnings-related social insurance fund able to carry an education benefit alongside unemployment cover
Employer acceptance of statutory study leave, through tripartite bargaining
A political majority willing to keep paying for it when fiscal consolidation became the priority
Where else it has been used: Other Nordic states have study-leave income support. Finland's abolition shows how a contribution-funded benefit with no constituency outside its users can be dropped in one budget round.
Limits: The allowance financed study leave; it did not measure participation. About 30,000 recipients a year is a small slice of Finland's working-age population, so its abolition moves the indicator's construct (the share of adults in education or training) by at most a little over one percentage point of the workforce, and no survey figure here shows that participation actually fell. The ministry's own memo, written to cost the abolition, notes that about half of recipients already held a higher-education degree and that the benefit reduced labour supply by an estimated 9,400 to 10,500 person-years, so the case for the allowance as a learning-equity tool is contested in the source itself. The 2022 figure is from a Ministry of Finance memorandum, the 2023 figure from the Employment Fund (Tyรถllisyysrahasto), a statutory fund run by the labour-market parties under Financial Supervisory Authority oversight, reporting on its own benefit; the 2023 number was read from an infographic in a PDF. The source does not print the full 2001-2025 recipient series: only 2020-2024 values are printed, in which 2023 is the highest. Applications surged in early 2024 ahead of the cut-off, so the last years overstate steady-state demand. The abolition was a fiscal decision announced in the 2023 government programme, so the record documents a capability ended by choice, not by collapse.
๐บ๐พUruguaysince 2009still operatingbears on Adult learning participation
Uruguay's Instituto Nacional de Empleo y Formaciรณn Profesional, created by Law 18.406 in 2008 and governed by government, employers and workers, grew from 9,696 training, guidance and technical-assistance places in 2009 to 114,364 in 2019, and its director reported that more than 50,000 people took its training in 2023.
Parliament placed the national labour-retraining fund in a non-state public body governed jointly by the government, employers and unions, which buys courses from training providers and steers them to the unemployed, to workers in firms and to people outside Montevideo.
Preconditions
A payroll-financed retraining fund that predates the institute and gives it its own budget
Employer and union confederations willing to co-govern a public body
A market of training providers the institute can contract instead of teaching itself
Where else it has been used: Tripartite training funds of this kind run across Latin America, from Guatemala's INTECAP to the Dominican Republic's INFOTEP. Uruguay's version is younger and buys training from providers instead of teaching through its own centres.
Limits: The 2023 figure is a rounded "more than 50,000" spoken by the institute's own director in an interview relayed by the presidency's press office, not a published statistical table. The 2019 figure counts places, not people: it includes guidance and technical assistance as well as courses, can count one person several times, and is read from data labels on a chart in a PDF. The two numbers are in different units and must not be read as a fall from 114,364 to 50,000, and the jump from 47,590 in 2016 to 112,183 in 2017 coincides with a change in the source behind the series, so the trend before and after 2017 is not continuous either. The same report also claims "more than 120,000" annual users, which does not match its own chart. Fifty thousand people is roughly 2 percent of Uruguay's adult population, so INEFOP is a national delivery channel but carries only a small part of adult participation in training, and neither number says whether anyone completed a course, learned anything or found work. Half of 2023 participants were unemployed and the reported age band is 20 to 35, so the reach is skewed young.
๐จ๐ฑChilesince 1976operating below its peakbears on Adult learning participation
Chile's SENCE training tax credit, running since 1976, financed 999,732 training participations and 653,939 unique trainees in 2024, down from 1,355,430 participations and 740,152 unique trainees in 2016, while public spending on it nearly doubled in nominal pesos; in July 2026 a Senate committee agreed to cut its ceiling from 1% to 0.7% of payroll.
The state let firms deduct accredited training costs from corporate income tax up to 1% of payroll, with SENCE accrediting courses and intermediaries (OTIC) pooling funds. The volume of training then followed firms' tax appetite rather than a public target, and Hacienda moved to shrink it once DIPRES found no wage effect.
Preconditions
A corporate income tax base large enough for a payroll-linked deduction to matter to firms
An accreditation agency and intermediary bodies able to certify courses and liquidate claims
Administrative records that let a budget office evaluate impact and argue for cutting a tax expenditure
Where else it has been used: Payroll-levy and tax-credit training schemes are common across Latin America and Europe; the Chilean record shows the usual failure mode, where large firms capture most of the subsidy and low-skill workers in small firms stay out.
Limits: The table opens in 2016, so 2016 is the highest year it shows, not a proven all-time peak; earlier SENCE yearbooks were not opened and the true peak may be higher or earlier. Participations count a person once per course, so the 26% fall in participations overstates the fall in people trained; unique trainees fell 12% (740,152 to 653,939). 2020 and 2021 are depressed by COVID and the series had recovered partly by 2023. Figures are SENCE administrative data republished by DIPRES, not a survey of adult participation, and the programme reaches mainly large firms: DIPRES finds about 88% of trainees come from large companies, so the numbers say little about adults in small firms, the self-employed or the unemployed. The same evaluation found no impact on employment or wages, so a hostile reader could call the shrinkage a correction of a low-value subsidy, not a lost capability. The July 2026 cut from 1% to 0.7% of payroll is a Senate committee agreement on the reconstruction bill; whether it was enacted was not verified, and the Executive's original bill eliminated the credit, which its later indicaciones replaced with a modification (Senado, 9 July 2026). Unique trainees peaked at 741,910 in 2018, not 2016.
๐ฐ๐ทSouth Koreasince 1999operating below its peakbears on Adult learning participation
Korea's Employment Insurance-funded employer and in-work training reached 4,916 thousand training participations in 2018, then fell to 2,410 thousand in 2025 after the Ministry of Employment and Labor withdrew support for statutory common training in 2019, COVID closed classrooms in 2020 and a cap on outsourced remote training took effect in late 2023.
The Ministry of Employment and Labor refunded employers from the Employment Insurance fund for training their staff, which drove volume up to a 2018 peak, then cut support for statutory and oversupplied courses and capped remote training to contain cost and quality problems, and participations halved.
Preconditions
An employment insurance fund with a levy earmarked for vocational development
A large market of accredited private training providers, including remote courses cheap enough to scale fast
A ministry willing to trade headline volume for tighter subsidy rules
Where else it has been used: Levy-refund training schemes in many OECD countries show the same arc: volume grows on cheap compliance and remote courses, then falls when the funder tightens eligibility; the volume series alone cannot tell lost capability from removed waste.
Limits: The count is person-courses (์ฐ์ธ์), and the page itself warns that in-work training double-counts heavily, so a halving of participations is not a halving of adults trained. The counting basis changed in 2020 from fund payment date to course start date, so 2018 and 2025 are not measured identically. The aggregate bundles several programmes (employer training subsidy, the in-work Naeil Baeum card, national HRD consortia, work-study), so it is a programme family, not one delivery. Much of the fall is deliberate: the ministry stopped subsidising statutory compulsory courses in 2019 and capped low-quality outsourced remote training in 2023, so a hostile reviewer would read the decline as pruning of low-value box-ticking courses, and the same page shows unemployed and vulnerable-group training growing from 355 thousand in 2020 to 467 thousand in 2024. The series is administrative, not a survey of adult participation, and says nothing about training firms fund without the subsidy.
๐ต๐ญPhilippinessince 1994still operatingbears on Adult learning participation
The Philippines' Technical Education and Skills Development Authority, created in 1994, recorded 1,633,393 enrollees and 1,428,724 graduates in technical-vocational education and training across all regions in 2023, with enrollment 29.61% above 2022.
The Philippines merged its manpower council, technical-education bureau and apprenticeship office into one authority that sets training regulations, registers public and private providers, funds scholarships and certifies competencies, so a national skills system runs through thousands of local providers under one standard.
Preconditions
Promulgated competency standards and an assessment-and-certification system employers recognise
A network of public training institutes plus regulated private providers that reaches every region
Scholarship funding that can be scaled up and down with labour-market needs without collapsing provider capacity
Where else it has been used: Unified TVET authorities with competency-based certification echo Australia's and Singapore's models; the Philippine case shows the scale such a body can reach while leaving labour-market relevance a separate test.
Limits: The gap asks for the share of adults in training in the last 12 months; this is a count of enrollments, not a population share, and the record does not compute one. Enrollments are not unique people: one person taking two courses counts twice, and the figure includes young people entering the labour market for the first time as well as adults returning to learning. The counts are TESDA's own administrative statistics, covering public and registered private providers; informal and employer training outside TESDA registration is missing, and course length ranges from short modules to multi-month programmes. A 29.61% rise in one year partly reflects recovery from the pandemic years and scholarship budget swings, so year-on-year growth is not a trend. The source is a PDF statistical report, not an API, and it shows output, not employment or wage effects.
๐ต๐ฆPanamasince 2006still operatingbears on Adult learning participation
Panama's national vocational training institute, INADEH, restructured by Decree Law 8 of 2006, recorded 156,032 course graduates in 2021, against 90,107 in 2019 and 38,628 in the pandemic year 2020.
Panama's government, through a tripartite board chaired by the labour minister, runs one autonomous national training institute with fixed regional centres, mobile courses and, since 2020, a virtual campus, so short courses can be scaled up in a crisis without new institutions.
Preconditions
A dedicated training levy or budget line that survives changes of government
A network of regional centres and mobile units that reaches comarcas and rural provinces
A course registry able to count enrolment and completion by area and centre
Where else it has been used: National training institutes on the SENA and SENAI model are common in Latin America; INADEH's pandemic shift to a virtual campus shows the model scaling through cheap online short courses, which raises the count faster than it raises skills.
Limits: The construct asks for the share of adults trained in the last 12 months; this is a count of graduations from one public provider, so it is neither a share nor a count of people: one person finishing three short courses counts three times, and private and employer training are missing. The 2021 jump is driven by the virtual campus (55,968 graduates at 'Virtual Sede' that year, none before 2020) and by short courses, so it may reflect lighter course formats rather than more training hours. The institute reports on itself and calls the figures preliminary, extracted from its SISGED database. The table carries visible typos (a 2013 female count printed as 32,64), and nothing after 2021 was found in this series. The started year is the 2006 decree law that restructured INAFORP, a predecessor created in 1983, into INADEH, so national vocational training predates it.
๐ง๐ดBoliviasince 2006delivered and closedbears on Adult learning participation
Bolivia's Ministry of Education reports that its national literacy programmes, run as 'Yo, sรญ puedo' from 2006 to 2008 and continued as a post-literacy programme from 2009, taught more than one million people over 15 to read and write between 2006 and 2017, 74% of them women.
Bolivia's Ministry of Education ran a time-bound national literacy campaign with a standardised audiovisual method and local volunteer facilitators in community centres, then converted its graduates into the enrolment base of a standing post-literacy directorate.
Preconditions
A census-based count of illiterate adults to set targets and find learners
A ready-made, low-skill teaching method that volunteers could deliver without trained teachers
Political priority sustained across the campaign years and a ministry unit to inherit the graduates
Where else it has been used: The same Cuban 'Yo, sรญ puedo' method was used in Venezuela, Nicaragua and Ecuador before and alongside Bolivia, with national literacy declarations in each that independent surveys later qualified.
Limits: The ministry that ran the programmes reports its own cumulative total, as a rounded 'more than one million', with no annual series and no independent verification; the same page gives a second, malformed figure ('1.oo1,372') for 2006 to 2008 alone, which, if accurate, leaves almost no growth in the cumulative total after 2008, so the two figures are either loosely rounded or show the post-literacy phase adding few new literates, and the page is dated 11 May 2016 while citing 2017 data, so it was edited after publication. A count of people made literate over twelve years is not the construct's share of adults in education or training in the last 12 months, and basic literacy is the narrowest form of adult learning; nothing here shows retention of literacy, completion of post-literacy, or participation by adults who were already literate. The page's illiteracy rate of 2.5% in 2017 is not used as a metric: it is not sourced to INE on the page, and higher government-cited rates for 2023 and 2024 appear in state press, so the rate series is contested. The 'Yo, sรญ puedo' method and early delivery relied on Cuban and Venezuelan support, which a reviewer may weigh against test three, though the programme was mandated and run by Bolivia's ministry.
๐ป๐ณVietnamsince 2010delivered and closedbears on Adult learning participation
Vietnam's General Department of Vocational Education reported that, by the end of 2019, about 9.6 million rural workers had received vocational training at various levels under the Prime Minister's Project 1956, 87.2% of its 11 million target, including 5.59 million whose elementary and under-three-month courses the project subsidised, 85.5% of a 6.54 million target for 2010 to 2020.
Vietnam's Prime Minister set a national ten-year quota for rural vocational training, and the labour ministry's vocational directorate financed short courses per head through provinces and districts, which commissioned public and private training providers against annual targets.
Preconditions
A provincial and district administration able to recruit learners and commission courses against a central quota
A network of district vocational centres and private providers to deliver short courses at volume
A per-trainee subsidy schedule set centrally, which made the quota financeable
Limits: The figures come from the directorate that administered the project, reported in the Government's own news outlet in January 2020, so 'by the end of 2019' was at best a preliminary tally and no final 2020 closing figure was opened. The count is cumulative over ten years and does not give the construct's share of adults trained in the last 12 months; it is likely to count enrolments rather than distinct people, and most of the subsidised training was courses under three months. The 9.6 million total includes training at all levels that the project did not fund, so the 5.59 million is the number attributable to the project's own policy. The project missed both targets. The article's employment outcome (more than 80.4% with new or better-paid work after training) is self-reported and includes people who kept the same occupation; no independent impact evaluation was opened. Provincial reports in the same years record uneven quality and courses that did not match local labour demand.
๐ฉ๐ชGermanysince 2008dismantledbears on Adult learning participation
The federal education ministry paid half the cost of job-related further training for lower-income workers through a voucher from December 2008, issued more than 400,000 vouchers, and stopped issuing them on 31 December 2021.
The federal education ministry, with ESF money, handed eligible low-income workers a voucher covering half of a course fee up to 500 euros after a counselling session, putting the purchasing decision in the worker's hands rather than the employer's.
Preconditions
A large market of accredited private and public training providers able to accept the voucher
A nationwide network of counselling points to issue vouchers and check eligibility
Co-financing from the European Social Fund, which tied the programme's life to an EU funding period
Where else it has been used: Individual learning accounts and vouchers were tried widely, from the United Kingdom's ILAs to France's CPF. The German version stayed small and ended; the lesson that travels is that a voucher needs reach and redemption, not issuance, to change participation.
Limits: The figure is the programme's self-report on the BMBF's own programme site, a cumulative total and is a lower bound ('over 400,000'), with no year-by-year series on the page, so the peak and the shape of the decline cannot be read from it. Vouchers issued are not vouchers redeemed, and not people trained. At about 30,000 a year on average, computed from the cumulative total over 13 years, against a workforce of over 40 million, the programme never moved the national participation share the indicator asks for. It was co-financed by the European Social Fund and ended with an ESF funding period, so the end reads as a funding decision rather than a judged failure, and federal support for adult training continued through other instruments; the record documents the loss of this one delivery, not a collapse of adult-learning support. The half-cost and 500 euro terms used in the claim and pattern come from BIBB (https://www.bibb.de/de/742.php), not from the source page.
๐ฉ๐ชGermanysince 1996still operatingbears on Adult learning participation
Federal and state governments paid grants and loans to 186,100 people taking vocational upgrading courses such as master craftsman or certified educator in 2025, 3.8 times the 49,300 supported in 1997, with 1.1 billion euros approved.
Federal law entitles anyone taking a recognised upgrading qualification to support for course fees and living costs, part grant and part loan, so a skilled worker can leave or reduce work to train for a master or technician certificate without employer sponsorship.
Preconditions
A codified ladder of upgrading qualifications (Meister, Techniker, Fachwirt) that employers recognise
Federal and state co-funding under a statutory entitlement rather than a time-limited programme
Chambers and schools able to run courses and examinations at national scale
Where else it has been used: Austria and Switzerland support similar higher vocational qualifications, but the German entitlement model depends on the Meister system existing first; countries without that ladder have little to attach it to.
Limits: A count of funded people is not the share of adults who took part in education or training, which is what the indicator asks for, and it says nothing about people who trained without this support. The 1996 start year is derived from 'thirty years ago' in a 2026 release. 28 percent (52,000) of recipients in 2025 were trainee educators (Erzieher), so the total moves with one occupation's staffing policy. The count has slipped from its 2022 peak of 192,400, a small decline that is noted here but not large enough to call the delivery eroded.
๐ช๐ธSpainsince 2004operating below its peakbears on Adult learning participation
Spain finances employer-organised training through rebates on firms' social security vocational-training contributions, managed by the state foundation FUNDAE. In 2024, 347,442 firms trained 3,378,343 workers, 22.7 percent of private-sector employees, across 5,824,016 participations.
The state ring-fenced a payroll levy for vocational training and let every contributing firm reclaim it as a social security rebate for training it organises itself. A tripartite foundation, now FUNDAE, governed by government, employers and unions, verifies the training and publishes coverage.
Preconditions
A payroll training levy collected through the social security system that can be returned as a rebate
Social partners willing to co-govern the fund and accept its verification rules
Enough administrative capacity in firms to organise and report training, which is why micro-enterprise take-up stays low
Where else it has been used: Levy-rebate schemes run in France and Singapore. Spain's coverage split by firm size shows the design reaches employees of large firms far better than workers in small firms, who need the training most.
Limits: The figures are FUNDAE's own administrative counts of training that firms report in order to claim rebates, so training that is not reported or not rebated is missing. The coverage rate divides people trained by the EPA private-sector employee count, which leaves out the self-employed, the public sector and the unemployed, so it is not the adult population share the gap asks for. The 22.7 percent and the 39.2 percent are different rates: 39.2 percent counts participations, and one person can count several times. The 22.7 and 3,378,343 values are read off an infographic page of the PDF, while the 5.8 million and 347,442 are in prose. Training is short, at 13.3 hours per participant in 2024 against 24 in 2014, and nothing shows skills gained. Reach across firms has eroded: 20.5 percent of eligible firms trained in 2024, against 29.7 percent (471,590 firms) in 2014, both on p.3 of the same Balance 2024 PDF. Micro firms cover 15.3 percent while large firms cover 91.8 percent, and firms used only 53 percent of their assigned credit. Worker participation rose while the firm base and training hours eroded, so the record is filed as eroded, with the 2014 firm-coverage peak as its second metric.
๐ฟ๐ฆSouth Africasince 2015operating below its peakbears on Adult learning participation
South Africa's Department of Higher Education and Training merged the public adult learning centres into nine Community Education and Training colleges on 1 April 2015; the colleges enrolled 273,431 people in 2016 and 120,081 in 2023, against a National Development Plan target of one million by 2030.
DHET took provincial adult learning centres into nine national CET colleges to professionalise public adult education. Enrolments kept dwindling, and some centres had lecturers but no students.
Preconditions
A national department able to fund and accredit adult programmes that adults actually want, not only school-equivalent certificates
An enrolment record that is audited, so that a decline can be told apart from a data clean-up
Physical learning centres and outreach in the communities the colleges are meant to serve
Where else it has been used: DHET's response is to tie CET colleges to SETA and National Skills Fund occupational programmes, which grew 118.8 percent in 2023 from a base of 558.
Limits: The two figures come from two DHET publications, and the peak is not established: 2016 is the earliest CET-college figure found, and the same Monitor puts the predecessor adult basic education system at 297,491 in 2010, so the decline began before the merger. Enrolment is a headcount, not the share of adults in learning that the indicator asks for, and the 2023 count rests on an 83.5 percent response rate, so it is understated by an unknown amount. DHET says enrolments have been dwindling since data moved to audited unit-level records, so part of the fall may be the removal of unaudited or phantom enrolments rather than learners lost. Most enrolments are Grade 12 and GETC, which is second-chance schooling rather than adult training; the largest group is aged 20 to 24; and DHET suggests part of the 2023 Grade 12 drop moved to the Department of Basic Education's second-chance programme. The sector's own report notes centres with lecturers and no students. Both numbers are in PDFs.
๐จ๐ดColombiasince 1957still operatingbears on Adult learning participation
Colombia's Servicio Nacional de Aprendizaje, created by decree in 1957, recorded 5,469,951 learners in short complementary training courses in 2023 (7,035,749 course places), and certified 3,661,218 completions in that strand the same year.
The state funds a national training agency from a payroll levy on employers and lets it run short, free, open-enrolment courses, in person and online, alongside its longer technical programmes, so any working adult can take a module without a qualification track.
Preconditions
A dedicated payroll contribution that funds the agency outside annual budget fights.
A national network of training centres in every region plus a virtual platform to reach rural learners.
Employers and unions on the governing board, which keeps course content close to jobs.
Where else it has been used: The levy-funded national training agency model is shared with Brazil's SENAI and SENAC, which predate SENA and were a reference for its design.
Limits: The indicator asks for the share of adults trained in the last 12 months; SENA reports learner counts from its own enrolment system (Sofia Plus), not a population share, and they cannot be turned into one: the count includes young people below adult age, enrolments through coverage-extension agreements with third parties, and short virtual and bilingualism courses, and a person enrolled in several courses during the year may be counted more than once. Completion is weak: the same table reports 47.29% dropout in complementary training. The started year is the creating decree, not the first course. The www.sena.edu.co copy of the report now serves an empty web app; the figure was read from SENA's own historico mirror, and the December 2024 edition was not found at the same path.
๐จ๐ณChinasince 2019operating below its peakbears on Adult learning participation
Under a State Council plan funded with 100 billion yuan from the unemployment insurance fund surplus, China's human resources ministry organised 32.184 million person-times of subsidised vocational skills training in 2021 alone, the final year of a three-year action that targeted more than 50 million.
The State Council set a three-year national training target and released 100 billion yuan from the unemployment insurance fund surplus into ring-fenced provincial accounts, and the human resources ministry paid accredited providers and employers per trainee against a published target catalogue.
Preconditions
An unemployment insurance fund with a large accumulated surplus that the centre could redirect
A provincial human resources administration able to accredit providers and verify claims at scale
Employers and migrant workers reachable through workplace and village channels
Where else it has been used: Singapore's SkillsFuture and Korea's employment insurance training accounts finance adult training from payroll-linked funds; China's version was a time-bound surge rather than an individual entitlement, which is why volume fell when the fund was spent.
Limits: Person-times are enrolments, not people: one worker trained twice counts twice, so the figure cannot be turned into the share of adults trained that the construct asks for. Even if every person-time were a different adult, 32 million is a small single-digit share of China's working-age population. The count is administrative and self-reported up through provincial human resources bureaux whose budgets depended on hitting the plan target; the bulletin publishes no completion, certification or employment outcome. The 2021 figure coincided with pandemic-era job retention schemes and sits beside 15 million 'training in place of work' subsidies, a wage support instrument. The second metric comes from a gov.cn news article, not a bulletin, and implies by subtraction that annual volume after 2021 fell to roughly half the 2021 level once the special fund ended: more than 92 million person-times from 2021 to September 2025, less 32.184 million in 2021, leaves about 60 million over about 3.7 years, against 32 million in 2021 alone. That drop is computed by subtraction, not published, and it is why the record is marked eroded. The 2023 and 2024 bulletins are on mohrss.gov.cn, which serves a script bot check, and were not read.
๐จ๐ทCosta Ricasince 1965operating below its peakbears on Adult learning participation
Costa Rica's Instituto Nacional de Aprendizaje (INA) recorded 204,773 training enrolments in 2024, down from 313,276 in 2015 at the high point of its 2010 to 2019 plateau, a fall of about 35 percent that the 2020 to 2024 recovery did not reverse.
Costa Rica built a single national training institute in 1965 and financed it largely from an employer payroll contribution, so INA could run free courses across every region without competing for annual budget allocations; after 2019 its enrolments fell back to roughly the level of the mid-2000s.
Preconditions
A dedicated, earmarked revenue stream that does not depend on yearly appropriations
A regional network of training centres able to reach rural planning regions as well as the Central Valley
A curriculum and certification model that employers recognise, which INA has been redesigning around competencies
Where else it has been used: The payroll-levy training institute is a Latin American family (SENA in Colombia, SENAI in Brazil, INA in Costa Rica). Costa Rica's series shows that the protected revenue keeps the institution alive but does not by itself keep participation from contracting when delivery models change.
Limits: Enrolments are a count of course registrations at one provider, not the share of adults in education or training that the gap asks for: a person can enrol more than once, and INA counted 97,870 distinct participants in 2024, roughly 2 to 3 percent of the adult population on our own rough arithmetic, with private, university, employer and online training invisible here. The table is INA reporting on itself and is not independently audited. The series is not clearly comparable across the break: the 2020 to 2024 tables state that enrolments cover programme modules, certifiable modules, technical assistance and competency certification, and the same yearbook reports enrolments in services run by accredited outside entities separately (87,258 in 2023, 65,300 in 2024), so part of the fall may be reclassification rather than lost reach; the yearbook does not reconcile the definitions. The single highest year is 2009 at 382,786, an isolated spike, so the peak named here is the 2015 plateau high and the fall would read as about 47 percent against 2009. The steepest drop (266,679 in 2019 to 172,664 in 2020) is the pandemic year, but the decline had begun in 2019 and enrolments stayed near 205,000 to 212,000 from 2022 to 2024. The number says nothing about training quality, completion or job outcomes, and the PDF has to be opened and read off a table.
๐ธ๐ชSwedensince 1997delivered and closedbears on Adult learning participation
Sweden's government funded municipalities to expand upper secondary adult education from autumn 1997 until 2002, and in spring 1999 municipal adult education produced about 157,000 places, of which about 120,000 full-year places were counted within Kunskapslyftet.
The government paid municipalities earmarked grants for adult education places and paid unemployed adults a special education grant at the level of unemployment benefit, so a year of study cost the learner nothing compared with staying on benefits.
Preconditions
A municipal adult education system (komvux) already present in every municipality
An unemployment insurance system whose benefit level could be matched by the study grant
Fiscal room to fund around 100,000 places a year for five years during a recovery from recession
Where else it has been used: Sweden itself relaunched the idea as a new Kunskapslyft in 2015, with vocational adult education rather than general upper secondary courses at its centre.
Limits: The figure counts full-year places, a volume measure, not people, and not a share of adults: one place can be several part-time students, so it does not translate into the participation rate the gap asks for. It is one semester's figure reported by the committee that ran and evaluated the initiative, citing a Skolverket follow-up, so the programme is partly reporting on itself. The committee's own figure in another passage of the same report is just under 112,000 full-year places, so the count depends on what is included. The same report's Skolverket table gives 228,000 participants (Deltagare) in spring 1999 in municipal upper secondary adult education, which is a count of people but still not a share of adults. The initiative was time-limited to 2002 by design. Effects on earnings and employment are contested in the evaluation literature and are not shown by this number.
๐ฎ๐ชIrelandsince 2011still operatingbears on Adult learning participation
Ireland's Higher Education Authority has run Springboard+ since 2011, funding free or heavily subsidised higher education places for unemployed people, returners and employed adults, with over โฌ400 million spent and over 110,000 people benefiting by 2025, when 7,719 places were offered on 249 courses at 37 institutions.
110,000 people (lower bound, 'over'), people who have benefited from springboard+ since 2011, cumulative, 2025. 7,719 places, springboard+ places offered in the 2025 call (offered, not taken up), 2025. Higher Education Authority (Ireland), Springboard+ 2025 launch, retrieved 2026-10-01.
The move, and what it needed
The Higher Education Authority runs an annual call in which colleges bid to offer short, accredited courses in skills-shortage areas, and the state pays full fees for unemployed applicants and 90 percent for employed ones.
Preconditions
Higher education institutions willing to build short part-time courses on an annual funding cycle
A skills-forecasting function to tell the call which shortage areas to fund
Labour market slack after 2008, which created the demand the programme was built for
Limits: The figure is a rounded, cumulative lower bound from a ministerial launch press release by the agency that runs the programme; 'benefited' is not defined and may count enrolments rather than completions or distinct people. The source is the agency's own launch press release for the 2025 call, and only the cumulative 110,000 figure describes delivery. Over 14 years, 110,000 is a small slice of an adult population of several million, so the programme says little about the national participation rate the gap asks for. Employment outcomes the HEA reports come from its own follow-up surveys and are not used here.
๐บ๐ธUnited Statessince 1966operating below its peakbears on Adult learning participation
Federally funded, state-administered adult basic, adult secondary and English language programmes served 2,629,643 participants in fiscal year 2000, and by fiscal year 2021 served 770,179, after a fall through every year the table shows up to fiscal year 2020 (627,728) and a partial rebound in 2021.
The federal government funds states by formula to run adult basic education, high school equivalency preparation and English classes through local providers, and requires every state to report participants and outcomes into one national reporting system.
Preconditions
A federal formula grant with a state match that keeps every state in the system
A national reporting system that defines a participant the same way across states
Local providers, from school districts to community colleges, willing to run low-margin classes
Where else it has been used: Formula-funded adult basic education is common across OECD countries; the American series shows how a stable statute can still deliver a shrinking service when appropriations flatten and accountability rules raise the bar for who counts.
Limits: The table starts in fiscal year 2000, so the second metric is the highest year it shows, not a proven all-time peak. FY2021 is a pandemic year and a partial rebound from FY2020's 627,728; FY2019, before the pandemic, already stood at 948,934, so most of the fall predates it. The fall is computed from two published table cells (about 71 percent), and part of it may reflect tighter participant definitions and federal performance reporting rather than lost provision; the table carries no note on definitional changes. The count covers only programmes reporting to the federal National Reporting System, not community college, employer or private adult learning, so it is not a national adult participation rate. Funding is shared between federal and state governments and the record does not attribute the decline to a single decision.
๐ณ๐ฑNetherlandssince 2022dismantledbears on Adult learning participation
The Dutch employee insurance agency UWV paid individual training grants of up to 1,000 euros from 2022, approved more than 214,000 applications in five rounds that year, and stopped the scheme on 1 January 2024.
The state replaced a tax deduction for training costs with a first-come grant that any worker or jobseeker could claim once a year from UWV, paid directly to a registered training provider, in fixed application windows with a capped budget.
Preconditions
A national register of eligible training providers and courses
An agency with an online portal and identity check able to process tens of thousands of claims in hours
A fixed annual budget, which made demand visible as queues and exhausted windows
Where else it has been used: Individual training budgets also run in France (CPF) and Singapore (SkillsFuture Credit). The Dutch version shows the failure mode of open first-come windows: budgets ran out within hours, and the scheme was ended after two years.
Limits: The figure is the executing agency reporting on its own scheme, counts approved applications rather than courses completed, and is a lower bound ('ruim 214.000'); the same release puts people with a grant at more than 200,000. Approval is not participation: the release does not say how many approved grants were used, and the record does not show whether the courses improved anyone's position in the labour market. The end date comes from a second UWV page (https://www.uwv.nl/nl/stap-budget), not from the source page. Against a Dutch labour force of roughly ten million, one year of grants is around two percent of workers, so the scheme never moved the national participation share by itself. UWV's page gives no reason for the end, and the 2023 release reports that the cabinet had to tighten the scheme against misuse; the record documents the loss of this channel, not a collapse of Dutch adult learning, which other instruments still fund.
๐ฒ๐ฝMexicosince 1981still operatingbears on Adult learning participation
Mexico's Instituto Nacional para la Educaciรณn de los Adultos certified 651,473 young people and adults as having completed a level of literacy, primary or secondary education in 2024, with 352,600 learners in its services at the close of the year.
A federal institute and state adult education institutes run a semi-open model in which volunteers with a small stipend tutor adults in community spaces, and learners sit module examinations when ready, so a credential can be earned without a classroom timetable.
Preconditions
A modular curriculum and an examination system recognised as equivalent to school certificates
A large network of subsidised volunteer tutors and community learning spaces
Federal and state co-financing, since delivery runs through state institutes
Where else it has been used: The semi-open, examination-at-will model resembles Brazil's ENCCEJA and Spain's adult graduation tests. The Mexican case shows the dependence on volunteer tutors, whose numbers rise and fall with the stipend budget.
Limits: The figures are INEA reporting on itself from its own tracking system (SASA). A completed level is a credential, often obtained by examination after self-study, and says nothing about how much was learned; one person can complete more than one level in a year. Against 27.5 million people aged 15 and over who, by the same document, have not finished basic education, a year's completions are about two percent of the backlog. The same document's indicator series shows completions fell sharply in 2019 and 2020 and rebounded from 2021, and it notes a fall in people served in 2024; chaining those published rates back from the 538,741 completions the document gives for 2023 puts 2019 at about 680,000 and 2018 at about 853,000, so 2024 is below both, by about 4 and 24 percent (our computation, from rounded rates).
๐ฏ๐ตJapansince 1958still operatingbears on Adult learning participation
Japan's national and prefectural governments trained 225,176 people through public vocational training in fiscal year 2023, 95,752 of them jobseekers and 114,552 people in work, and a further 44,699 through job-seeker support training.
The national government, through polytechnic centres run by its employment support agency, and the prefectures run free vocational courses for jobseekers and fee-based upskilling for employees, contract much of the jobseeker training to accredited private providers, and route trainees through Hello Work offices.
Preconditions
A national network of public vocational schools and polytechnic centres
Unemployment insurance and a training allowance so jobseekers can afford to attend
Employment offices that refer trainees and track whether they find work
Where else it has been used: Korea's employer training system and Germany's Federal Employment Agency vouchers do similar work. The Japanese model is distinctive for keeping a large public provider of its own beside contracted private training.
Limits: The figures sit in a table on page 10 of a PDF. The 225,176 total includes 14,872 school-leaver trainees who are mostly young people entering work, not adults returning to learning. Against a labour force of roughly 69 million, public training reaches well under one percent of workers a year; most adult learning in Japan is employer-provided and is not counted here. Trainee counts say nothing about the quality of training beyond the employment rates the same table reports for jobseeker courses.
๐ง๐ทBrazilsince 1996operating below its peakbears on Adult learning participation
Brazil's Educaรงรฃo de Jovens e Adultos, the second-chance basic education modality delivered mostly by public school networks, enrolled 2,962,322 students in 2021 and 2,252,069 in 2025, a fall of about 24 percent in four years, according to INEP's School Census.
The federal education law makes basic education for youths and adults a modality of the public school system, and state and municipal networks deliver it in evening classes, funded through the per-pupil FUNDEB transfers with a lower weight than regular schooling. Enrolment has fallen every year since at least 2020 as networks close evening classes and fewer young adults enter.
Preconditions
A national school census that counts every enrolment by modality each year
State and municipal school networks with buildings and teachers free in the evening
A per-pupil federal-state funding formula that includes the modality, which also sets the incentive to keep or cut it
Limits: Both numbers are read off a bar chart in a PDF slide deck (Grรกfico 24, PDF page 25), not a table, and the chart labels the totals as public plus private. The deck's window only starts in 2021, so the 'peak' is the first year shown and not EJA's historical peak: INEP's 2024 presentation (PDF page 35) shows 3,002,749 enrolments in 2020, and the decline is older than either window. The real loss from the modality's peak is therefore understated, not overstated. Enrolment counts registrations in the census reference week, not adults who attend or complete. A falling count is not by itself a capability loss: part of the fall reflects more young people finishing regular school on time, so fewer need a second chance, and INEP's own 2024 deck notes that EJA also absorbs teenagers who left regular school after repeated failure, so EJA is not purely adult learning. The source gives no count of the adults who lack completed basic education, so the 2,252,069 enrolments cannot be turned into the share of adults the indicator asks for. The 1996 start is the Lei de Diretrizes e Bases (Lei 9.394) that set EJA as a modality; it is not stated at the source URL, and adult schooling in Brazil predates it.
๐ฌ๐งUnited Kingdomsince 1992operating below its peakbears on Adult learning participation
Participation of adults aged 19 and over in government-funded further education and skills in England fell from a recent peak of nearly 3.3 million learners in 2012/13 to 1.6 million in 2020/21, and stood at 1,798,970 in 2024/25, according to the Department for Education.
The Department for Education funds colleges, local authorities and training providers through a national adult skills budget and later loans and apprenticeship levy funding, and counts every funded learner through a single administrative return. Real-terms cuts to the adult budget and the move of level 3 and above provision onto learner loans roughly halved adult participation within a decade.
Preconditions
A national network of further education colleges and adult community learning services funded by central government
A learner-level administrative record that every funded provider must return
A funding settlement that can be cut without legislation, which is what made the erosion possible
Limits: The peak is published only as rounded prose ('nearly 3.3 million learners in 2012/13'), so the second metric is a rounded figure and the fall to 1,798,970 (about 45 percent) is approximate. The department calls 2012/13 the 'most recent peak', which implies earlier years may have been higher; the record does not claim the series' all-time high. The release also warns that the 2024/25 introduction of the Adult Skills Fund and tailored learning breaks the long-running Education and Training time series, although it says total further education and skills volumes are unaffected. Participation counts learners on government-funded or loan-funded provision recorded on the Individualised Learner Record, so adults in employer-paid, private or university learning are outside it, and it is England only, not the United Kingdom. The total includes adult apprenticeships, whose own funding reforms moved volumes independently of the adult skills budget. A headcount cannot show course length or level, so some of the fall may reflect a change in the mix of courses rather than fewer adults learning. The rise to 1.8 million in 2022/23 and 2023/24 came mostly from Multiply and Skills Bootcamps, according to the release, and Multiply funding stopped in March 2025. The 1992 start date is the Further and Higher Education Act that created the funded further education sector; it is not stated at the source URL.
๐ต๐ชPerusince 2003operating below its peakbears on Adult learning participation
Peru's Educaciรณn Bรกsica Alternativa, the second-chance basic education modality for youths and adults, enrolled 226,500 students in 2016, the highest year of the 2014 to 2023 table, and 190,700 in 2023, according to INEI's Compendio Estadรญstico 2024 citing the Ministry of Education's school census.
The 2003 General Education Law converted Peru's old adult schools into a separate modality of basic education, delivered by public and private CEBA centres with flexible schedules and counted in the national school census. Enrolment peaked in 2016 and has drifted down since.
Preconditions
A school census run by the education ministry that reports enrolment by modality
An existing network of adult and evening schools that could be converted rather than built
A legal definition that lets learners certify primary and secondary completion outside the regular track
Limits: The numbers are in a PDF table (Table 5.6, PDF page 323 of 1,046), published in thousands with one decimal, and INEI republishes the Ministry of Education's census rather than collecting it; the Ministry's own ESCALE portal sits behind a bot-verification wall and could not be read. The peak is only the highest year of the ten-year window the table shows (2014 to 2023), and it was a one-year jump: enrolment was 204,500 in 2014 and 203,900 in 2015, so against 2014 the 2023 count is only about 7 percent lower, and the fall of about 16 percent is measured from that jump. The series reached its low of 188,600 in 2022 and rose slightly in 2023. The modality serves adolescents from 14 as well as adults, so not every learner is an adult, and the table does not split public from private centres. Enrolment is a census-day count, not attendance or completion, and a smaller EBA can partly reflect more young people completing regular schooling. A table footnote says the conversion of former adult education centres into the modality has been under way since 2006, so early years mix old and new structures. The 2003 start is the Ley General de Educaciรณn 28044 that created the modality, which is not stated at the source URL.
๐ช๐ชEstoniasince 1990still operatingbears on Adult learning participation
Estonia's secondary schools for adults and the non-stationary departments of other schools enrolled 7,866 learners in upper secondary education in 2025, the highest number in the Statistics Estonia series that starts in 1990, while the number of dedicated adult secondary schools fell from 16 in 2016 to nine.
The state and municipalities keep a network of adult gymnasiums and let ordinary schools run evening, distance and part-time departments that teach the national upper secondary curriculum free of charge to adults. When the network shrank, enrolment moved into consolidated schools and departments rather than disappearing.
Preconditions
A national curriculum and state examinations that are the same for adult and day learners
School funding that follows a learner into non-stationary study
A small, digitally administered school system in which consolidating adult schools did not cut access
Limits: The scale is small: 7,866 enrolments in a country of about 1.4 million people, and non-stationary general education is only one part of adult learning, leaving out vocational schools, universities, employer training and the unemployment fund's courses. Enrolment rose from 5,653 in 2022 to 7,866 in 2025, and the source does not say why; part of the rise may come from refugees from Ukraine and from young people who left day school, so the count is not purely adults returning to study. Basic education enrolments in the same table fell from a 2005 high of 1,722 to 588 in 2025, so the record is not a story of growth across the board. Fewer dedicated adult schools (16 in 2016, nine in 2025) means provision was consolidated, and more schools now teach adults through non-stationary departments (27 in 2025). The 1990 start is the first year of the published series, not the start of adult schooling in Estonia, and the table's reference year should be read as the school year beginning that autumn.
๐น๐ญThailandsince 1993still operatingbears on Adult learning participation
Thailand's Department of Skill Development, through 86 national institutes and provincial offices, trained 91,911 workers in skill-upgrading courses in fiscal year 2023, of whom 88,656 completed, according to its open-data release.
The Ministry of Labour runs a national network of skill development institutes and provincial offices that deliver short, free upgrading courses to employed workers by occupation group, and publishes per-unit results as open data.
Preconditions
A labour ministry department with its own training centres in every region
Occupational skill standards that courses can be built against
An open-data mandate that makes each unit report applicants, entrants and completers
Limits: Both numbers are computed: the department publishes one row per unit and occupation group (494 rows in fiscal 2566), and the totals here are the sums of its male and female completer columns, which the department does not print as a headline. The CSV is encoded in Thai Windows code page 874, and some cells carry a thousands separator, so a sum that parses them as text undercounts. The open data stops at fiscal 2566 (2023) and covers three years only (88,602 completers in fiscal 2564, 63,785 in fiscal 2565, 88,656 in fiscal 2566), so no long trend or peak can be read from it. About 89,000 completers a year is roughly 0.2 percent of a labour force near 40 million, the courses are short, and the count is the agency reporting on itself with no outcome measure. The figures cover the department's own upgrading courses only, not pre-employment training, supplementary occupation courses, employer training claimed under the Skill Development Promotion Act, or the education ministry's non-formal education, whose portals refused connections from this research. The 1993 start date is the founding of the department as given in general references, not stated at the source URL.
๐ณ๐ฌNigeriasince 1990still operatingbears on Adult learning participation
Enrolment in adult literacy education in Nigeria, reported by the National Commission for Mass Literacy, Adult and Non-Formal Education, rose from 322,288 in 2020 to 2,152,146 in 2021 and 2,915,453 in 2022, according to the National Bureau of Statistics.
A federal commission sets adult literacy and non-formal education policy and funds and coordinates state agencies for mass education that run community literacy centres with volunteer and paid facilitators, reporting enrolment upward to the national statistics office.
Preconditions
State agencies for mass education in every state to run the centres
Facilitators paid little or nothing, which keeps the count high and the quality uncertain
An administrative reporting line from centres to the federal commission and on to the statistics office
Limits: The figure is an administrative count supplied by the commission to the statistics office, which republishes it without a definition: it does not say whether enrolment counts people or registrations, whether state-run and donor-funded centres are included, or how many learners completed. A ninefold rise in two years is not explained in the report; 2020 was depressed by pandemic closures, and the jump may also reflect better reporting coverage rather than more learning. The series covers three years only, so no peak or trend before 2020 can be read. The number is in a PDF (Figure 1.1 and its text, PDF page 10), and the newest year is 2022. Even 2.9 million enrolments is small against the large non-literate adult population the commission targets, and literacy classes are basic education, not the training-for-work participation most adult-learning surveys count. The 1990 start is the decree that established the commission, not stated at the source URL.
๐จ๐ฆCanadasince 1996operating below its peakbears on Adult learning participation
Canada's provinces and territories delivered 154,892 employment-benefit services under the federally funded Labour Market Development Agreements in 2024-25, 92 percent of them Skills Development training, in a year when Part II funding fell by $425 million as a temporary top-up in place since 2016-17 ended and Skills Development spending fell 19.7 percent to $798.9 million.
The federal government channels Employment Insurance Part II money to provinces and territories through bilateral agreements, and each province buys or delivers skills training and apprenticeship support for EI-eligible workers, reporting to a common monitoring framework.
Preconditions
A contributory unemployment insurance fund large enough to finance active measures
Provincial employment services able to deliver training under federal money
An annual monitoring report that counts interventions the same way across 13 jurisdictions
Where else it has been used: Temporary federal top-ups raised volumes for eight years; when the top-up sunset the training budget fell with it, which is the usual fate of active measures financed as supplements rather than baseline.
Limits: The counts are services (interventions), not people: one client can hold several, and the department reports 502,007 clients across all LMDA services, most of them in employment assistance rather than training. The erosion rests on money, not on a measured fall in training: the chapter's introduction says Part II received $425 million less in 2024-25 because the temporary additional funding in place since 2016-17 sunset, and Skills Development spending fell 19.7 percent. Employment Benefits services also fell, from 192,365 in 2023-24 (Table 3.1.2.3) to 154,892, but nearly all of that is Skills Development Regular falling 37.8 percent to 62,084, which the table note attributes to a coding correction that moved provincial workshops out of that category; the services fall is therefore not counted as the loss here, and apprenticeship services rose 3.2 percent to 80,376. 2023-24 is the previous year, not a demonstrated all-time peak. The programme reports on itself, and the 92 percent share is stated by the department (62,084 plus 80,376 of 154,892). The started year is the date the first bilateral agreements were signed (Alberta and New Brunswick, December 1996, per the LMDA page on canada.ca), not stated in this chapter. The training is for EI-eligible workers, a subset of all adults, and the record says nothing about outcomes.
๐ญ๐ณHondurassince 1972still operatingbears on Adult learning participation
Honduras's national vocational training institute INFOP trained or certified 326,161 people in 2024 across six programmes and about 12,552 courses, figures the Tribunal Superior de Cuentas checked against instructors' start and end reports.
A levy-funded autonomous training institute, financed by employer and worker contributions, runs its own centres and contracts courses through business chambers and unions, and the supreme audit institution checks its delivery figures each year.
Preconditions
A payroll levy that gives the institute its own revenue
Business chambers and unions willing to host company-agreement courses
An audit office that reviews physical delivery, not only spending
Where else it has been used: The levy-funded national training institute is the Latin American model (SENA, INA, INTECAP, INFOTEP); Honduras's version leans heavily on short courses and outsourced company training.
Limits: The total is computed by us: the audit report (pages 9 to 10 of the PDF) gives six programme results, 24,181 (agriculture), 15,834 (industrial), 135,430 (tourism, commerce and services), 65,035 (ICT), 84,665 (company agreements) and 1,016 (labour competency certification), and no single total; the press figure of 325,145 is the first five. The counts are participations per programme product, not unique adults, so a person who took two courses counts twice. Many courses are short, about 20 hours at the low end, and the list includes soft-skills topics such as customer service and motivation; the industrial programme serves ages 15 to 25, so part of the count is youth. The numbers originate in INFOP's own reports; the Tribunal verified them against signed instructor reports and a sample of 68 participants per product, not by a full audit. Over-execution of up to 165 percent against plan says the targets were low, not that demand was met. The 2021 evaluation records that INFOP cancelled its programme for vulnerable groups after the pandemic, so coverage of the poorest is not shown here. The started year is the 1972 decree law creating INFOP, stated in the report.
๐ฎ๐ฑIsraelsince 2003still operatingbears on Adult learning participation
Israel's Vocational Training Division, in the labour ministry, trained about 61,000 people in 2022, against about 55,000 in 2019 and 59,000 in 2021, according to the Knesset Research and Information Center.
The labour ministry's training division licenses and supervises a large market of paid vocational courses, sets curricula and state certification exams for more than 350 occupations, and funds a smaller set of courses through referrals from the employment service.
Preconditions
State certification exams that employers recognise
Private and franchised training centres that can be supervised
A referral channel from the employment service for funded places
Where else it has been used: The supervised paid-course model keeps volume up when budgets fail, as in 2020, but it also means the state-funded core is a small fraction of the headline count.
Limits: The figures are rounded and reach us through the Knesset's research centre, which republishes labour ministry data; the sentence is on PDF page 11 (printed page 9), in Hebrew. Most of the volume is not state-funded training: the Center's December 2021 paper on the training system says that in 2019, 49,192 of the participants, 88.5 percent, were in non-budgeted courses that learners pay for at private institutions the Division supervises and certifies, and 5,621 were in state-funded courses. The same paper records a fall to 36,522 participants in 2020, which the ministry attributed to the absence of a state budget and the pandemic; the 2022 figure shows recovery, so the record is not filed as a reversal. About 61,000 a year is a small share of Israel's adult population, and the count says nothing about completion or employment outcomes. The started year is not given in either source: 2003 is the year a 2006 European Training Foundation report gives for the Division's predecessor bureau moving to the Ministry of Industry, Trade and Labour, and the bureau ran adult courses before that.
๐ต๐พParaguaysince 1998still operatingbears on Adult learning participation
Paraguay's Ministry of Education and Sciences enrolled 94,436 young people and adults in its four formal and non-formal Educaciรณn Permanente programmes in 2025, as reported in the Ministry of Economy and Finance's budget evaluation for that year.
The education ministry runs basic, secondary and vocational second-chance programmes for over-age learners through a national network of centres, many hosted in ordinary schools in the evening, and reports enrolment through the budget-by-results system.
Preconditions
Schools that can host evening classes for adults
A ministry directorate dedicated to youth and adult education
A budget reporting system that asks for physical results, not only spending
Limits: The number sits in one sentence on printed page 34 (PDF page 35) of a budget evaluation, sourced in its footnote to the education ministry's own reporting (SIME Nยฐ 5.130/2026); the ministry is reporting on itself through the finance ministry. The programmes serve people aged 15 and over, so part of the count is youth rather than adults, and enrolment is not completion. 94,436 is about 2 percent of Paraguay's adult population, so this evidences a standing national second-chance system, not broad adult participation. The four programmes are not broken out, and no earlier year is given on this page, so no trend can be read; the education ministry's open-data portal, which carries the enrolment series, did not respond from here. The started year is the 1998 General Education Law that set out permanent education for young people and adults; it is not stated in the source.
๐ญ๐นHaitisince 2010delivered and closedbears on Adult learning participation
Haiti's national literacy campaigns trained about 250,000 adults between 2010 and 2015, about 10 percent of the illiterate population, according to the education ministry's ten-year plan.
The state ran adult literacy campaigns through the education ministry's literacy apparatus, reaching an estimated tenth of the illiterate adult population over five years.
Preconditions
Volunteer or low-cost literacy facilitators
Foreign cooperation to fund materials and training
Enough stability for community classes to meet
Limits: The only figure is a rounded approximation on printed page 25 (PDF page 28) of the ministry's plan, which cites an unpublished 2017 MENFP review of the 2010-2015 operational plan; no annual series, completion or literacy test results are given. 'Trained' is not 'made literate'. The copy that opens is hosted by the MIT-Haiti Initiative, not by the ministry, whose own site does not carry the PDF. The plan places the adult literacy campaigns, itinerant courses of at most six months, under a Secretariat of State (the Secrรฉtairerie d'รtat ร l'Alphabรฉtisation; printed page 21, PDF page 24), but it does not say who paid for them; Haiti's literacy work in this period drew on foreign cooperation, so the domestic share of the funding is unknown. The same page says non-formal education suffers from inadequate infrastructure, budgets and staff, and the record does not show the campaigns continuing after 2015. 2010 is the start of the reporting period, not the start of literacy campaigns in Haiti.
๐ต๐ฑPolandsince 1999still operatingbears on Adult learning participation
Poland's county governments and cities with county rights ran general secondary schools for adults whose enrolment rose from 26,443 pupils in 2010 to 47,584 in 2013, fell to 16,470 in 2021 and recovered to 26,734 in 2024, according to Statistics Poland's Local Data Bank.
Counties and cities with county rights, which own Poland's public secondary schools, keep evening and extramural general secondary classes open for adults who want the same matura a teenager earns, tuition-free and funded through the public education budget.
Preconditions
A county tier of local government that owns secondary schools and decides which adult classes to open
A national exam and certificate system that treats an adult's matura the same as a teenager's
A per-pupil public education budget that follows adult pupils into local-government schools
Limits: The figures are computed: GUS publishes the two governing-body series separately (2024: 13,857 in country powiats and 12,877 in cities with powiat rights; 2013: 30,667 and 16,917; 2010: 11,917 and 14,526; 2021: 8,046 and 8,424) and this record adds them. The country-powiat series starts in 2010, so the sum cannot be taken further back, and the city series alone shows a break (40,439 in 2000, 5,259 in 2006, 18,038 in 2007) that tracks the restructuring of secondary schooling and a change in classification, not a real swing. These schools are a minority of adult general secondary schooling: in 2024 the same subject counts 102,881 pupils in total, of whom county and city governments hold about 26 percent, gminas, voivodships and central administration another 1,360, social organisations and associations 5,277, and the remaining roughly 69,500 sit in schools run by other bodies, which the variables read here do not break out. The 2013 high was a two-year surge from 23,737 in 2011 that the source does not explain, and the 2024 count is back at its 2010 level after three years of growth, so the record is filed as operating: the fall from 2013 to 2021 reads as a surge unwinding rather than a lost capability, and a reader who treats 2013 as the baseline would call it eroded. BDL labels the school year by its start year; the API response does not say so, so '2024' most likely means 2024/25. A pupil count is enrolment, not completion or a matura pass, and it covers general secondary schooling only: adult vocational schools, post-secondary schools, courses and employer training are outside it. The 1999 start is the year Poland's county tier was created and given secondary schools; it was not read at the cited source.
๐ฒ๐พMalaysiasince 2001still operatingbears on Adult learning participation
Malaysia's public community colleges, run by the Ministry of Higher Education, counted 259,788 part-time students, including participants in Pembelajaran Sepanjang Hayat lifelong-learning courses, in 2024, against 259,810 in 2023, according to the ministry's Statistik Pendidikan Tinggi 2024.
The higher education ministry uses its national network of community colleges, built for post-secondary vocational certificates, to also run short and part-time courses for local adults, and counts them in its annual statistics.
Preconditions
A network of public colleges spread across districts, close to where adults live
Short-course funding inside the college budget so courses can be cheap or free
A ministry statistics unit that collects part-time enrolment college by college
Limits: The ministry reports on its own colleges. The number sits in a PDF: Table 5.4 runs over several pages and the national total is on page 42 of the 72-page chapter file (printed page 143), with the note 'Part Time Student includes PSH programme'. The table does not split PSH short-course participants from part-time certificate and diploma students, does not give ages, and counts enrolments rather than unique people, so one adult taking three short courses can count three times and many PSH courses last hours or days. Against roughly 23 million Malaysians of working age it is about one participation per hundred adults, which is a reach figure, not a participation rate comparable with Eurostat or PIAAC. Earlier editions of the same publication give 238,581 (2018), 222,675 (2019), 118,455 (2020), 140,501 (2021) and 245,909 (2022), so the count halved in the pandemic and has since passed its pre-pandemic level; those figures are in the 2019, 2021 and 2023 chapter files, not at the cited URL. The 2025 edition of chapter 5 no longer carries a part-time table, so 2024 is the latest year published and the status rests on that year. The 2001 start is the year the community college system began operating as commonly reported; it was not read at the cited source.
๐ฎ๐นItalysince 2020still operatingbears on Adult learning participation
Italy's Fondo nuove competenze, set up in 2020 and run by the national labour policy agency ANPAL through two notices, was created to pay for working hours employers turned into training under union agreements, and by the end of 2021 it had covered almost 30 million hours of training for almost 270,000 workers, about 110 hours each.
ANPAL paid employers for working hours converted into training, but only where a collective agreement with the unions set out the cut in hours and the skills plan, which turned a COVID income measure into a negotiated training programme.
Preconditions
A collective bargaining system able to sign thousands of firm-level agreements quickly
European Social Fund money to co-finance the wage cost
Regional bodies able to assess and certify training plans, even if mostly by tacit approval
Limits: The figures come from ANPAL's own thematic review of December 2021, an interim count for the first notice; the final total for that edition is not on this page, and larger figures circulating elsewhere were not verified. The page is an archive: ANPAL was abolished and the fund passed to the Ministry of Labour, and the page says further funding is planned under the Programma nazionale Giovani, donne e lavoro. About 270,000 workers is under one percent of Italian adults (our rough computation), so this is a national programme that barely moves the participation share the indicator asks for. It is a wage subsidy as much as a training programme: the fund reimburses hours employers took out of production, and in 2020 to 2021 that also kept people employed during COVID restrictions. Training content and hours are certified by employers and union agreements, and ANPAL notes that the regions assessing training plans relied widely on tacit approval (silenzio-assenso), so quality control was light.
๐ง๐ฉBangladeshsince 1991operating below its peakbears on Adult learning participation
Bangladesh's Bureau of Non-Formal Education ran national adult literacy projects that reached 9.23 million learners aged 11 to 45 under the Total Literacy Movement project of 1997 to 2003, then 4.46 million aged 15 to 45 under the Basic Literacy Project (64 districts) of 2014 to 2022. Its only programme for learners above school age the bureau lists after June 2022 is a pre-vocational course that reached 4,198 people aged 14 to 24 in 2025-26.
The Bureau of Non-Formal Education used development-budget projects to fund literacy courses, delivered by district administrations or contracted NGOs, in every district. When a project window closed, delivery stopped until the next project was approved, and after 2022 no adult-scale successor followed.
Preconditions
A standing bureau with a district presence to plan and monitor courses
Development-budget approval for each time-bound project, often with donor co-finance
NGOs or district administrations able to run community learning centres at scale
Where else it has been used: Time-bound literacy campaigns financed project by project were common across South Asia in the 1990s. The pattern they share is the one here: large one-off reach, then a gap when the project closes.
Limits: These are the implementing bureau's own counts of learners it reports as reached. The page does not say how a learner was counted or whether literacy was tested. The numbers cover multi-year project windows and not the last 12 months, which is what the indicator asks. Spread over each window they are small against the adult population: roughly 0.5 million a year for 2014 to 2022 and 1.5 million a year for 1997 to 2003, both converted from lakh by us. The peak is itself a shortfall: the 1997 to 2003 project reached 92.25 lakh against a target of 228.89 lakh, about 40 percent. A hostile reviewer would say most of the fall reflects fewer illiterate adults to reach, since primary enrolment rose over the period, and not lost capacity. Nothing here measures that need, so the erosion is in delivery scale and not in a participation rate. The projects run as separate time-bound windows with gaps between them, so calling them one delivery is our reading. Both age bands include non-adults: the 1997 to 2003 project counted learners from age 11 and the 2014 to 2022 project from age 15, so part of each count is children and adolescents. The 2025-26 figure is a different course for ages 14 to 24 and is not a like-for-like successor. Basic Literacy Project teaching was contracted to NGOs. The page is a Bengali table on a government portal, and the values are read off its rows.
๐บ๐ฌUgandasince 1992operating below its peakbears on Adult learning participation
Uganda's Ministry of Gender, Labour and Social Development enrolled 434,664 adult learners in its government-funded Functional Adult Literacy programme in 2004, and enrolment fell to 88,533 by 2007.
The Ministry of Gender, Labour and Social Development ran adult literacy classes through district community development officers and volunteer instructors. Enrolment fell after 2004 while government releases fell only from Shs 3.61bn to 3.08bn a year, so the source does not explain the fall.
Preconditions
District community development staff in every local government
Volunteer instructors willing to teach for little or no pay
A protected budget line that outlasts donor and poverty-fund cycles
Limits: The series stops in 2007. The report is a 2008 ministry submission to UNESCO, and no later official FAL enrolment series was found or checked. Whether it recovered later, under this or another name, is not shown, so the eroded status describes 2004 to 2007 and nothing after. 2004 is the high of one eleven-year table, and the jump from about 149,000 in 2001 to 358,000 in 2002 suggests a one-off funding push rather than a stable peak. The table's printed total (2,090,340) does not match the 2,091,340 in the text above it. The figures are the ministry's own administrative counts, taken from a PDF, with no independent check. A count of literacy learners is much narrower than the indicator's construct, the share of all adults in any education or training. The report's national figures show 43% of adults had some primary schooling, so not every enrolee was starting from zero. It also says FAL relied on UNICEF, UNESCO and DVV funding that stopped around 1999, so part of the swing may come from donor funding rather than from state delivery. The same report's release table shows government FAL releases of Shs 3.61bn a year in 2002/03-2004/05 and 3.08bn in 2005/06-2006/07, a cut of about 15% against a fall in enrolment of about 80%, so the cause of the decline is not shown. Elsewhere the report dates the government's re-launch budget to 1991. The start year is a re-launch: there was adult literacy work before 1992, and FAL began as a pilot in eight districts.
๐น๐ฟTanzaniasince 1993operating below its peakbears on Adult learning participation
Tanzania's Ministry of Education ran Integrated Community-Based Adult Education in all 25 mainland regions from 1993. Enrolment peaked at 1,900,255 learners in 2005 and fell to 786,602 by 2014.
The Ministry of Education ran literacy circles of up to 30 learners in primary school buildings in the evenings, staffed by facilitators and coordinated through local government. Enrolment fell by more than half within a decade as funding and facilitator support thinned.
Preconditions
A primary school in every village to host evening classes
A local government coordinator for adult education in each ward and district
Recurrent funding for facilitators that survives the end of a campaign, which is what eroded
Where else it has been used: Tanzania's 1970s mass literacy campaign is the earlier, larger version of the same move. The literacy gains it is credited with were also reported to slip back in the 1980s and 1990s.
Limits: The table exists only as an image on the UNESCO page, so the numbers were read off a picture and a machine cannot check them. UNESCO's page is a case-study write-up that presents the programme as good practice, and the figures are ministry figures republished without a stated source. The ministry's own 2008 CONFINTEA report prints 1,900,225 for 2004/05, but its male and female columns sum to UNESCO's 1,900,255, so the difference is a ministry arithmetic slip; the ministry's 1999/2000 and 2000/2001 rows reappear as UNESCO's 2000 and 2001, so UNESCO's year labels may be the second year of a school year. The ministry labels its table 'participation in literacy programmes', so UNESCO may have attached the ICBAE name to all adult literacy participation. The series has gaps (2002 to 2004 are missing) and ends in 2014, and later years were not checked, so the eroded status describes 2005 to 2014 only. 2014 ticked up from 734,950 in 2013, and the fall is not monotonic. A count of literacy learners is far narrower than the indicator's construct, the share of all adults in any education or training. The programme had no standardized tests, so the count says nothing about what learners gained.
๐ฌ๐ญGhanasince 1992operating below its peakbears on Adult learning participation
Ghana's Non-Formal Education Division enrolled 312,907 adults in a single National Functional Literacy Programme batch in 2003-2005, and after donor funding ended in 2006 the government-funded batches shrank to 14,029 learners in the 2014-2016 batch.
The Ministry of Education's Non-Formal Education Division ran 21-month local-language literacy classes in every region through volunteer facilitators, scaled them with an IDA credit to about 200,000 to 300,000 enrolments per batch, and then carried the programme on a government budget line that funded a fraction of that after 2006.
Preconditions
A volunteer facilitator network and primers in 15 Ghanaian languages
External credit financing most of the cost at peak
A domestic budget line large enough to replace the donor share, which is what did not happen
Limits: The two numbers come from two different Ministry of Education reports, and both are administrative enrolment counts reported upward by district supervisors; the 2013 report says monitoring visits to verify figures had not been possible since 2010 for lack of funds, and the 2018 report restates Batch 18 at 44,730 where the 2013 report had 51,257, so the series is revised without explanation. A batch runs about 21 months and batches overlap (Batch 10 ran November 2003 to 2005 alongside Batch 9), so a per-batch count is not a 12-month participation rate, and the peak-year total across overlapping batches was higher than the single batch quoted. The programme teaches basic literacy, which is one narrow slice of the adult learning the indicator asks about, and even the peak batch was a few percent of Ghana's adults. The 2003-2005 peak was financed mostly by an IDA credit: US$33.86 million of US$44.94 million in project costs (World Bank ICR535, IDA-3246, P000974), so the erosion partly measures a donor exit rather than a domestic choice. The 2018 report also shows a separate Basic English Literacy track (17,647 learners in 2016/18) that partly offsets the fall. The record carries no figure after the 2014-2016 batch.
๐จ๐ฎCรดte d'Ivoiresince 1996operating below its peakbears on Adult learning participation
Cรดte d'Ivoire's education ministry ran a national network of literacy centres. It counted 75,275 learners in 2,828 centres in 2015-16 and 53,311 learners in 2,707 centres in 2024-25, a fall of about 29 percent in learners over the decade.
The education ministry ran adult literacy through a dedicated service and a network of centres staffed by volunteer or low-paid animateurs. That network contracted by about a quarter in the decade after 2015 while primary enrolment expanded.
Preconditions
A standing budget line for animateurs and materials, which a volunteer-based network lacks
Partner NGOs and donors whose project cycles keep centres open, so closures follow funding cycles
Demand from adults who can take time out of informal work, which literacy alone rarely pays for
Limits: The numbers are enrolments in literacy centres, not a participation rate. At about 53,000 learners against an adult population in the millions, the network reaches well under 1 percent of adults. So it shows little about adult learning participation at national scale. The erosion is measured between two of the ministry's pocket yearbooks. The yearbooks give 2015-16 (75,275), 2016-17 (57,299), 2017-18 (59,111), 2022-23 (55,405), 2023-24 (52,554) and 2024-25 (53,311). 2015-16 is the highest of these years, not a confirmed all-time peak. Most of the fall happened in one year, 2015-16 to 2016-17, which could be a change in coverage or method rather than in delivery. The yearbooks do not explain it. The centres include those run by NGOs, religious bodies and communities as well as the state, and the yearbook does not split them by operator. The age chart in 2024-25 shows some learners under 15. The figures are administrative counts collected through the ministry's own IEPP and DRENA offices. The publisher's domain (mena-desps.org) is the DESPS statistics site, not a gouv.ci address. The start year is when the service that runs the network was created, not when the first literacy class was held.
๐ง๐ซBurkina Fasosince 2002operating below its peakbears on Adult learning participation
Burkina Faso's adult literacy and non-formal education centres, financed through FONAENF and run by contracted operators under the faire-faire strategy, enrolled 178,021 adult learners in 2015. Enrolment fell to 113,721 in 2018 and 86,405 in 2021, when 1,885 adult centres opened against 3,094 in 2018.
The ministry created FONAENF in 2002 to pool state and donor money and pay NGOs and associations to run literacy centres under the faire-faire strategy, so the state finances and supervises while operators deliver. Because each year's allocation decides how many centres open, cuts in the budget line and the spread of insecurity reduced the number of centres and learners almost immediately.
Preconditions
A pooled fund with donor co-financing and a pipeline of contracted local operators in every region
An annual budget allocation large enough to open thousands of centres each campaign
Enough security in rural areas for centres to open and learners to attend, which the country lost after 2016
Limits: The figures are counts of people enrolled in literacy and non-formal education centres, not the share of adults that the gap asks for. The ministry publishes no participation rate, and on any plausible adult population the 2021 count is around one percent of adults or less, so even the 2015 level was a small programme against the construct. The 2021 figure comes from a ministry monitoring report marked 'version provisoire' (provisional), and the copy used is hosted by the Global Partnership for Education, a funder of the plan. The 2015 figure comes from a different ministry document, the SNDENF 2021-2025, and the copy used is on a third-party site (mojaafrica.net), not a government domain. The two documents do not fully agree: the SNDENF prints 2016 adult enrolment as 86,001 in one table and 88,770 in the next. 2015 is the high point of the published 2015 to 2019 series and may not be the programme's true peak, and the series is volatile: it dropped by half in 2016 and partly recovered in 2018, which shows enrolment following each year's funding, not a steady decline. The 2021 figure is up from a 2020 low of 81,350, and the ministry credits the rise to a slight increase in budget programme 060. The 2021 level matches 2016 (86,001), so the decline is from the 2015 and 2018 highs, not a steady fall. The centre counts in Tableau V-4 are for 'jeunes et adultes' centres. 'Jeunes et adultes' covers learners aged 15 and over, and the adolescent centres are counted separately and excluded here. The counts are compiled from forms filled in by the operators the fund pays, so the people delivering the programme are reporting their own enrolment. The ministry attributes the fall to reduced budget allocations, and its own reports say the security crisis since 2016 closed many education facilities, but the series cannot separate the two causes. The record has no figure after 2021, so it cannot say whether the decline continued after the 2022 coups.
๐ฒ๐ผMalawisince 1986still operatingbears on Adult learning participation
Malawi's Ministry of Gender, Community Development and Social Welfare has run the National Adult Literacy Programme since 1986 through volunteer-taught community classes. Enrolment in Adult Literacy and Education reached 171,434 learners in the 2024/25 financial year, against about 82,000 participants in 2006.
The ministry responsible for community development runs adult literacy as a standing national programme. It organises classes of about 25 learners through district community development officers and volunteer instructors that communities choose, and NGOs and churches run parallel classes under the same policy.
Preconditions
A field network of district community development staff to open and supervise classes in every district
Volunteers with at least primary schooling willing to teach for little or no pay
A recurrent budget line, however small, that survives changes of ministry
Limits: The gap asks for a share of adults, and this is a count of learners in one literacy programme. Against Malawi's adult population, roughly 10 million people aged 15 and over, it is under 2 percent. That ratio is our rough computation and is not published. The figure is the ministry's administrative report, carried without method in a budget document. It does not say whether NGO and faith-based classes are included. The series is inconsistent: the Annual Economic Report 2024 gives the same 158,048 as learners who 'graduated' in 2023/24, and the 2025 report uses it as the previous year's enrolment. The 2006 comparison comes from a different ministry, the department's earlier home, and counts public and NGO participants together, so the two numbers may not measure the same thing. Literacy classes are one narrow kind of non-formal adult education. They say nothing about vocational, workplace or tertiary adult learning. The 2008 report records that the programme aimed to reach 2 million of 4.6 million illiterate adults by 2000 but had reached about 860,000 by 2006. It also records 'magnitude of failure', volunteer instructors whose minimum qualification is primary school grade 8, and inadequate funding. The 2025 report names the National Strategy on Adult Literacy and Lifelong Learning, not NALP, so treating the 2024/25 enrolment as the same programme begun in 1986 is our reading. Enrolment says nothing about completion or literacy gained.
๐ฒ๐บMauritiussince 2009still operatingbears on Adult learning participation
The Mauritius Institute of Training and Development, the state technical and vocational training body, enrolled 3,352 trainees on part-time courses across its training centres between July 2022 and June 2023, out of 8,668 people it trained that year in full-time, part-time and apprenticeship modes.
A statutory training institute under the education ministry runs a national network of training centres that sell evening and part-time courses to workers alongside its full-time and apprenticeship programmes, and publishes centre-by-centre enrolment each year.
Preconditions
A network of public training centres spread across the island
A national qualifications framework that part-time courses lead into
Employers and a training levy that make part-time study affordable
Limits: MITD reports on itself. The part-time table is on page 54 of the PDF (printed page 52) and the 8,668 total is in the opening statement on page 6. The counts are enrolments, not unique adults, and the report gives no ages: part-time courses are the mode most likely to serve working adults, but the full-time courses (2,993 trainees on page 50) mainly take school leavers and are not adult learning in the sense of the indicator, which is why the part-time figure is the metric. Professional drivers training (526) and hotel school short courses (787) make up close to two fifths of the part-time count. MITD is one provider: the Human Resource Development Council's training schemes, private providers and the universities are not in it, so it is not a national participation rate. 2022/23 is the latest annual report on the MITD site. The start year is the 2009 Act that created MITD from the former Industrial and Vocational Training Board, as commonly reported; it was not read at the cited source.
๐ณ๐ฆNamibiasince 1998still operatingbears on Adult learning participation
The Namibian College of Open Learning, a state-owned distance education college, registered 27,242 learners in its Grade 11 and 12 secondary programme in 2025, against a peak of 30,999 in 2017 and a low of 23,258 in 2022.
The state set up a distance-learning college that lets anyone who left school without the senior secondary certificate register for individual subjects, study from printed and online materials and attend weekend tutorials at centres across the regions, then sit the national examinations.
Preconditions
A national examination that any registered candidate can sit regardless of age
A network of tutorial centres in schools used after hours
Public subsidy keeping per-subject fees low
Limits: NAMCOL reports on itself. The numbers are read off charts in a PDF laid out as two-page spreads: the 2025 total and the 2010 to 2025 trend (Figure 3) are on PDF page 15 (printed pages 28 to 29). The digest publishes no age profile, so the record cannot show how many learners are adults rather than young people repeating Grade 11 or 12 after leaving school; NAMCOL's secondary programme is a second-chance route and many of its learners are in their late teens and early twenties, so this is the weak point of the record for an adult learning indicator. Learners take an average of about two subjects, part-time, so a registration is not full-time study. The digest attributes the 2018 to 2022 fall, sharpest in 2022 at 17.7 percent, to curriculum changes, and enrolment has risen each year since; 2025 is still 12 percent below the 2017 peak, which is why the peak is carried but the status is operating, not eroded. TVET enrolment at NAMCOL is small (327 in 2025) and is not included. The 1998 start is the year the digest's foreword dates the publication series from; NAMCOL was created by an Act of 1997.
๐ฒ๐ฟMozambiquesince 1975operating below its peakbears on Adult learning participation
Mozambique's education ministry counted 243,655 learners in its public adult literacy and adult primary education classes in the April 2025 school survey (169,341 in literacy and 74,314 in primary education for young people and adults), against 302,207 in 2024 and 529,314 in the March 2014 survey.
The education ministry runs literacy and adult primary classes in community centres staffed by paid or volunteer literacy teachers, counts the learners in its annual school census alongside children, and publishes the totals by province.
Preconditions
A national adult education directorate with provincial and district officers
Literacy teachers paid a stipend or volunteering in each community
An annual school census that also reaches non-formal centres
Where else it has been used: The same census shows the programme concentrated in Nampula, Tete and Sofala in 2025; a programme that depends on stipended community teachers is the first to shrink when the education budget is squeezed.
Limits: The cited URL is an Internet Archive copy: mined.gov.mz timed out on 2026-10-02. Both numbers sit in PDFs: the 2025 total is in the summary table on page 14 of 73 and broken down by province on later pages; the 2014 figure is in the same summary table of the 2014 edition (levantamento_estatistico_do_sector_2014.pdf, page 13, archived copy of 19 August 2026). The comparison is computed by the record (a fall of about 54 percent) and crosses a change in the survey: 2014 was taken on 15 March and labelled 'Educacao de Adultos', 2025 on 30 April and split into literacy and primary education for young people and adults after the 2018 education system law, so part of the gap may be definitional. The intermediate editions read on the same day (290,478 in 2016, 251,887 in 2018, 262,641 in 2019, 229,329 in 2020, 302,207 in 2024) show most of the fall happened between 2014 and 2016 and that the level has moved between about 230,000 and 300,000 since, so 'eroded' rests mainly on one early drop; the 2014 and 2016 editions report the same 2,647 centres, which suggests some figures were carried forward. 2014 is the earliest edition read, not a demonstrated all-time peak. Counts are learners attending at survey date, not completers, and about two thirds are women; nothing here shows literacy gained. Private centres (a few hundred learners) are excluded. The 1975 start is the national literacy campaigns after independence, as commonly dated; it was not read at the cited source.
๐ฆ๐ดAngolasince 2023still operatingbears on Adult learning participation
Angola's national vocational training system, the INEFOP-supervised centres together with the accelerated 'plano intercalar' courses, FUNEA and licensed private centres, enrolled 204,026 trainees in 2024, against 112,478 in 2023, with 152,625 passing, across 1,914 training centres.
The labour ministry's training institute runs its own centres and mobile units, licenses private providers into one national system, and adds an accelerated short-course programme, then counts enrolment, passes and dropouts across all of them in an annual yearbook.
Preconditions
A licensing regime that brings private training centres into the national count
Mobile and rural units that reach trainees outside the main cities
A ministry statistics office that publishes the returns with the national statistics institute
Limits: The numbers sit in a 43-page PDF: the breakdown by channel is on printed page 12 and Tabela 3 on printed page 19. The figures are the labour ministry's administrative returns from its own training institute and from centres it licenses, republished by INE. The text calls the trainees 'jovens' throughout, gives no age split, and the system includes initial vocational training for school leavers, so this is a count of vocational trainees rather than of adults returning to learning. The 81 percent jump from 2023 rests largely on the 'plano intercalar', an acceleration programme that enrolled 56,329, and on 57,127 enrolments reported by private centres, whose returns the ministry does not audit in this document. Counts are enrolments, not unique people. Only two years are compared, so the series cannot show whether 2024 is a lasting scale or a one-year push. The started year is the first year in the cited table, not the start of INEFOP, which is older.
The Democratic Republic of the Congo's Institut National de Prรฉparation Professionnelle recorded 34,376 learners in initial and continuing vocational training in the first half of 2026, against a semester forecast of 42,850, according to the mid-term review of its 2026 performance contract.
A tripartite public training institute financed by an employer payroll levy runs training centres in the provinces for employees and job seekers, and the employment ministry holds it to an annual performance contract with a numeric training target reviewed mid-year.
Preconditions
An employer training levy collected from formal firms
A network of provincial training centres with workshops
A ministry performance-contract regime that sets and reviews numeric targets
Limits: The source is the institute's own news release reporting a mid-term review by a joint commission under the employment ministry; it is self-reporting, not a statistical series, and no annual INPP report with a time series was found at an openable official page. The figure covers one semester and mixes initial training for job seekers with continuing training for employees, without a split. 34,376 learners in six months in a country of over 100 million people is a small reach, and the release gives no completion or placement figures. The execution rate of about 80 percent is the release's own arithmetic. The started year is the creation of INPP by Ordonnance-loi nยฐ206 of 29 June 1964, as stated on the institute's history page, not on the cited release.
๐ฆ๐นAustriasince 1994still operatingbears on Adult learning participation
Austria's Public Employment Service (AMS) had on average 76,496 people a day enrolled in its training measures in 2025, up 1.3 percent on 2024, against an annual average of 65,126 in 2015.
The public employment service buys and runs courses for registered jobseekers, from basic skills to vocational qualifications, and pays an allowance while they attend.
Preconditions
An unemployment insurance system that pays a training allowance
A purchasing budget and a market of adult education providers
Case workers who can assign jobseekers to courses
Where else it has been used: Most public employment services run this model; Austria keeps the training stock at about a fifth of everyone the AMS counts as unemployed or in training (76,496 of 394,036 in 2025, our division), a high share by European standards.
Limits: The agency reports on its own measures, in a five-page PDF (the training table is on page 3). The figure is an annual average stock, the number enrolled on an average day, not the number of different people trained in the year, which is several times larger and is not in this file. It counts only registered jobseekers placed in AMS-funded courses, so employed adults, apprentices and adults in schools or universities are outside it, and it moves with unemployment and with the AMS budget: in 2021 it reached 70,337 after a 23.2 percent rise during the pandemic job programme. The 2015 comparison is from the 2015 edition of the same report (https://www.ams.at/content/dam/dokumente/berichte/001_uebersicht_jahr2015.pdf). The trainees are excluded from the registered unemployment count, which gives the agency an incentive to place people on courses. The 1994 start is the year the AMS was established as a separate agency, not stated in this file.
๐จ๐พCyprussince 1979still operatingbears on Adult learning participation
Cyprus's Human Resource Development Authority subsidised more than 106,000 participations in 11,064 single-company and multi-company training programmes in 2024, at a cost of 35.9 million euros, including 68,064 participations of employed persons in standard single-company programmes against 32,987 in 2023.
A levy-funded public authority reimburses employers part of the cost of training their staff, in their own firm or through accredited training centres, against a list of priority themes agreed with the social partners.
Preconditions
A payroll levy that gives the authority its own income
Accredited private training centres able to run courses across sectors
Social partners on the authority's board who agree the training priorities
Where else it has been used: Levy-grant systems of this kind run in Malaysia, South Africa (zaf-seta-wbl) and Latin America (gtm-intecap, dom-infotep); Cyprus's is among the oldest in Europe and shows the usual sensitivity of claims to the subsidy rate.
Limits: The authority reports on its own schemes, in a 185-page PDF: the 106,000 headline is a rounded floor ('more than') in the summary of results on PDF page 16, and the scheme figures are on PDF page 22. The counts are participations, not people: one employee attending three subsidised courses counts three times, many courses last a day or less, and the report gives no count of distinct individuals. The bulk is employer training of people already in work, subsidised from a levy employers pay; the unemployed and adults outside employment are a small part. The doubling of single-company participations from 2023 to 2024 coincides with what the report calls a substantial increase in subsidy amounts and simplified approval, so part of the rise may be employers claiming for training they would have run anyway. The 1979 start is stated in the report's introduction as the year the authority began operating. The URL carries a cache identifier from the authority's portal and may change.
๐ฉ๐ฐDenmarksince 1960operating below its peakbears on Adult learning participation
Statistics Denmark counted 386,879 participations in Denmark's state-funded adult vocational training courses (AMU) in 2024, about 7,056 full-time equivalents, down from 1,050,944 participations and 15,307 full-time equivalents in 2010.
The state funds short, employer-relevant vocational courses for skilled and unskilled adults, designed by the social partners and delivered by vocational schools and AMU centres, with an allowance paid during the course.
Preconditions
Social partners who agree on course content through sector committees
A national network of vocational schools and training centres
Wage compensation or allowances so workers can attend during working hours
Where else it has been used: AMU is the reference model of a social-partner adult training system; Denmark's own trajectory shows that participation follows the price set for users and the allowance, as in fin-adult-education-allowance.
Limits: A participant in this table is a course participation, not a person: AMU courses are typically one to a few days, so one worker taking several courses counts several times, and the full-time equivalent series (7,056 in 2024 against 15,307 in 2010 and 11,278 in 2005) is the better measure of volume. The 2009 and 2010 peak was inflated by the financial crisis, when laid-off and work-sharing workers were sent on courses; against the pre-crisis level of 621,037 participations in 2005 the 2024 count is still about 38 percent lower, so the erosion does not rest on the crisis peak alone, but it is smaller. The 2025 figure in the same table is 218,057, a fall we have not seen explained and which may reflect incomplete registration, so the record stops at 2024. The decline reflects several things the table cannot separate: user fees introduced and raised in the 2010s, cuts to adult education allowances, and changes in course supply. The 1960 start is the founding of the AMU system as commonly reported, not stated in the table.
๐ญ๐ทCroatiasince 2022still operatingbears on Adult learning participation
The Croatian Employment Service enrolled 23,627 people in training paid for by its education vouchers in 2025, 9,786 of them in digital-skills programmes and 2,558 in green-skills programmes, and 53,106 people in total from the scheme's start on 1 April 2022 to the end of 2025.
The public employment service gives unemployed and employed adults a voucher that pays for an accredited short programme chosen from a national catalogue of skills, booked through an online application.
Preconditions
A catalogue of accredited providers and programmes tied to a national qualifications framework
A digital application through which adults apply and providers are paid
Temporary EU recovery funding large enough to open vouchers to employed people
Where else it has been used: Voucher and individual learning account schemes in the corpus include France's CPF (fra-cpf), Singapore's SkillsFuture (sgp-skillsfuture) and the Dutch STAP budget (nld-stap-budget), which was abolished; Croatia's is the newest and its continuation after the recovery plan is not yet funded.
Limits: The service reports on its own measure, in an 82-page PDF (printed page 43, PDF page 45). 'Included' means enrolled in a programme, not completed. The yearbook says 54, 13 and 33 percent of users 'completed' digital, green and other programmes; the three shares sum to 100, so they describe the mix of programmes and no completion or employment rate can be read from them. It also splits users into 70 percent 'traลพitelji zaposlenja' (jobseekers) and 30 percent unemployed without defining the first group, which presumably holds employed people seeking training or a new job; the share of employed adults is therefore not stated. The vouchers are financed from Croatia's EU Recovery and Resilience Plan (reform C4.1 R3), so the scale depends on money that runs out in 2026, and the yearbook gives no budget beyond that. The 2025 intake is not compared with earlier years in this file.
๐ฑ๐ปLatviasince 2017delivered and closedbears on Adult learning participation
Latvia's State Education Development Agency (VIAA) trained almost 77,000 employed adults aged 25 and over through its ESF-funded 'Mฤcฤซbas pieauguลกajiem' project between 2017 and 2023, in ten rounds and 2,309 programmes offered by 124 education institutions, with 90 to 100 percent of fees covered.
The national education agency bought places on formal and non-formal programmes from colleges, universities and vocational schools and offered them to employed adults in competitive rounds, with EU money paying almost the whole fee.
Preconditions
European Social Fund money large enough to cover 90 to 100 percent of fees
Education institutions willing to adapt programmes for working adults, including distance formats
A national agency able to run open application rounds and pay providers
Where else it has been used: Portugal's Novas Oportunidades (prt-novas-oportunidades) and Croatia's vouchers (hrv-vauceri) are other EU-funded adult training pushes; Latvia's ended with its funding period and was replaced by a different instrument, individual learning accounts.
Limits: The agency reports on its own project, and the headline is a rounded 'almost 77,000' credited to 'project data' with no table behind it; whether it counts people or enrolments is not stated, and an adult who took part in two rounds may be counted twice. The figure covers seven years, roughly 11,000 a year against fewer than a million employed people, and only employed adults aged 25 and over, so the unemployed and young adults are outside it. 'Improved their competence' is the agency's wording; the page does not say how many completed a programme or gained a qualification, beyond 1,909 people compensated for a qualification exam. The project was financed by the European Social Fund and ended in 2023; the agency has since moved to individual learning accounts funded by the EU recovery plan, which this record does not cover.
๐ณ๐ดNorwaysince 2002still operatingbears on Adult learning participation
Norway's municipalities, delivering the statutory right of adults to primary and lower secondary education, taught 10,223 adult participants in 2025, 7,588 in ordinary basic education, 2,435 in special education and 200 in both, against a high of 13,946 in 2019.
National law gives every adult who lacks it a right to primary and lower secondary education, and the municipality must provide it free, reporting participants to the national municipal statistics system.
Preconditions
A statutory individual right that municipalities must honour regardless of budget
Municipal adult education centres, often combined with Norwegian language teaching for immigrants
A common municipal reporting system (KOSTRA) that counts participants the same way everywhere
Where else it has been used: Sweden's Komvux and Estonia's adult gymnasiums (est-adult-gymnasiums) rest on the same idea of a legal right to second-chance schooling; Norway's version is small because few native-born adults lack compulsory schooling.
Limits: The URL is the table's API endpoint; the national row ('EAK The whole country') is returned by a POST query on it, or can be read in the StatBank interface for table 12237. The counts come from municipalities' own KOSTRA reports. They cover only compulsory-level schooling for adults (and special education), a small part of adult learning: upper secondary education for adults, vocational training, study associations and employer training are outside it, and it reaches about 10,000 people in a population of 5.6 million. Most ordinary participants are recent immigrants and refugees completing basic schooling, so the series follows arrivals: the rise to 2019 tracks the 2015 to 2016 refugee intake and the fall since then reflects fewer such learners as much as any change in provision, which is why the record is not classed as an erosion. The 2002 start is the year the adult right to primary and lower secondary education took effect, not stated in the table, and the series itself starts in 2015.
๐ธ๐ฐSlovakiasince 2024still operatingbears on Adult learning participation
Slovakia's Central Office of Labour, Social Affairs and Family enrolled 17,008 clients in its national project 'Zruฤnosti pre trh prรกce' in 2025 and paid training contributions for 16,853 of them, 14.56 million euros in all, with 9,714 registered jobseekers starting retraining courses against 5,904 in 2024.
Labour offices let registered jobseekers pick an accredited course and provider themselves, approve a contract, and reimburse the fee and travel costs once the course is completed.
Preconditions
A market of accredited adult training providers across regions
Labour offices with case workers able to assess applications individually
EU co-financing to pay course fees in full
Where else it has been used: The demand-led design resembles voucher systems such as hrv-vauceri, with the difference that the Slovak office pays after completion rather than issuing a voucher up front.
Limits: The office evaluates its own project, in a 76-page PDF (printed pages 36 to 37, PDF page 37). The 17,008 mixes three activities, retraining of registered jobseekers, skills for young people not in employment or training, and training of registered job applicants (zรกujemcovia o zamestnanie), who may be in work, and the client counts are enrolments and payments, not completions; the retraining contributions are paid after a course ends, so some 2025 payments relate to agreements from 2024. The client chooses the course and provider from the open market and the office only reimburses it, so course quality is not controlled by the office. The six-month placement rate the office reports (53.83 percent for retrained jobseekers) is not a measured effect, since there is no comparison group. The project began in April 2024, so the 2024 comparison covers only part of a year; earlier demand-led retraining schemes of the labour offices (known as REPAS+ and KOMPAS+) are not covered in this report and their scale was not checked.
๐ธ๐ฎSloveniasince 1991still operatingbears on Adult learning participation
Slovenia's publicly accredited upper secondary programmes for adults, run by schools and adult education centres, enrolled 17,858 adults in the 2024/25 school year, up from a low of 13,851 in 2020/21 but below the 20,580 enrolled in 2015/16.
The state accredits upper secondary programmes adapted for adults, delivered part-time by secondary schools and folk high schools (ljudske univerze), so adults can earn the same vocational or general qualification as young people.
Preconditions
A national qualification framework that lets adult programmes award the same certificates as youth schooling
A network of folk high schools and school adult units in every region
Public co-funding or employer payment of fees, which adults otherwise bear
Where else it has been used: Poland's licea dla dorosลych (pol-licea-dla-doroslych) and Estonia's adult gymnasiums (est-adult-gymnasiums) are the same instrument; the Slovenian series shows the usual dip in the pandemic and a partial recovery.
Limits: The URL is the table's API endpoint; the national total is returned by a POST query selecting the totals for sex, age and type of education, or can be read in SiStat. The count is enrolment in formal upper secondary programmes leading to a state-recognised qualification, from short vocational to the matura course, so it leaves out the non-formal adult education that makes up most adult learning in Slovenia, and it reaches about 1.5 percent of the working-age population. Enrolment is not completion: SURS publishes adult graduates in a separate table that was not used here. The series restarts in 2015/16 under a new methodology; the old series (table 0953356S) gives 13,327 in 2014/15 and 15,963 in 2011/12, which cannot be joined to the new one without a break. The 2020/21 low coincides with the pandemic. Adult upper secondary education predates 1991; the start year is that of the independent state's school system and is not in the table.
๐ฆ๐ฑAlbaniasince 2014still operatingbears on Adult learning participation
Albania's National Agency for Employment and Skills (AKPA) registered 11,250 trainees in 2021 in vocational courses at its ten regional public vocational training directorates, 7,052 of them on free places for registered jobseekers and other priority groups by the bulletin's text, after 15,143 in 2019 and 7,427 in the pandemic year 2020.
The national employment and skills agency runs regional public training centres that offer short national and provider-level vocational courses, free for registered jobseekers and paid for others.
Preconditions
A network of regional public training centres under the employment agency
A national catalogue of short qualifications that centres can certify
Referral from employment offices so jobseekers reach free places
Limits: The agency reports on itself, on page 60 of an 83-page PDF in Albanian. The bulletin contradicts itself on the free places: the text on page 60 says 7,052 (about 63 percent), while Table 27 on page 61 and Table 29 on page 65 give 7,811 trainees on free places, 6,631 of them registered unemployed, so the free-place figure is uncertain by about 750 and the record carries the text figure. 2021 is the latest annual bulletin found at an open URL: the agency's publications page renders by script and no 2022 to 2025 bulletin was found, so the record cannot show the current scale, and 'operating' rests on the agency's site still listing course offers and centre notices on 2026-10-02, not on a count. The 2019 (15,143) and 2020 (7,427) figures are in the agency's 2019 and 2020 bulletins on the same host, not at the cited URL; the 2021 bulletin itself says 2021 was 50 percent above 2020, which fits the 2020 figure, and the 2019 figure was not re-checked. 2021 sits below 2019, but with pandemic disruption in both 2020 and 2021 the record does not call that an erosion. Counts are course registrations, not completions or unique people: Table 29 gives 6,180 certified of the 11,250. Most courses are short provider-level courses, but national-level courses can run up to two years. The started year is the earliest labour market bulletin the agency still hosts (2014); the public vocational training centres are older, and their founding year was not read at any source, so delivery began no later than 2014.
๐ง๐ฌBulgariasince 2002still operatingbears on Adult learning participation
Bulgaria's Employment Agency (Agentsia po zaetostta) included more than 155,450 unemployed and employed adults in training in 2025, against just over 87,540 in 2024, with 97,911 people certified through digital skills vouchers financed under the national Recovery and Resilience Plan.
The national employment agency issues training vouchers to unemployed, inactive and employed adults, who choose a licensed provider, and pays the provider on completion, with EU recovery and social funds paying for most of the volume.
Preconditions
A register of licensed vocational training providers under the national VET agency
An online voucher application system run through the labour offices
EU funding lines (RRF, ESF+) large enough to pay for courses at national scale
Limits: The agency reports on itself, and the cited URL is a zip archive from the agency's 'Otchet za deynostta na AZ' page holding two PDFs; the headline counts are on page 81 of the 84-page text volume (Otchet_Plan za deistvie_2025_text.pdf), and both are stated as 'over' and 'just over', so they are rounded. The 2025 jump is mostly vouchers: 97,911 people certified in basic and advanced digital skills under component 2 of investment C1.I3 of the Recovery and Resilience Plan (page 34), and 22,140 employed people trained under the ESF+ project for employed workers; much of the money is EU, though the agency delivers it. Digital skills courses can be short and online, the counts mix enrolment and certification, and one person may appear in more than one scheme. Training for the registered unemployed is far smaller: 6,705 unemployed people were included in training under the programmes and measures of the national employment action plan (page 17). The RRF component ends with the facility in 2026, so 2025 is unlikely to be a stable level. The 2002 start is the year of the Employment Promotion Act under which the agency trains adults, not read at the source.
๐ฒ๐ฉMoldovasince 2025still operatingbears on Adult learning participation
Moldova's National Employment Agency (ANOFM) enrolled 1,751 registered unemployed people, 1,223 of them women, in voucher-based vocational training courses between January and August 2026, and 952 unemployed people graduated from voucher courses in the same months, some of them enrolled in 2025.
The employment agency gives registered unemployed people a training voucher redeemable with an approved provider through an online platform, and reports enrolments and graduates monthly.
Preconditions
A register of approved vocational training providers in the regions
An online voucher platform linked to the unemployment register
A state budget line for active measures under the employment law
Limits: The agency reports on itself, in a monthly statistical PDF (rows 15 to 18); the figures are year-to-date for eight months, not a full year, and no December 2025 report was found at an open URL, so no full-year or prior-year comparison is made. The agency's 2025 activity report is a one-page scan with no text. Separately, 95 unemployed people enrolled in workplace training (art. 33) over the same months, not added. The scale is small: about 1,750 enrolments in eight months in a country of about 2.4 million, limited to the registered unemployed (6,010 at the start of August), so it says nothing about adults in work. The 2025 start is the date of the agency's launch event for its voucher platform for training the unemployed (17 April 2025, on the agency site); training of the unemployed under the employment law existed before vouchers, and that earlier delivery is not counted.
๐ท๐ดRomaniasince 2002operating below its peakbears on Adult learning participation
Romania's National Employment Agency (ANOFM) enrolled 15,634 people in vocational training courses, skills assessment and apprenticeships under its national training plan in 2025, down from 27,843 in 2023, with unemployed trainees falling from 25,420 to 11,082 while the registered unemployment rate rose.
The national employment agency sets an annual national training plan and delivers it mainly through its own regional adult training centres, free for the registered unemployed and paid for others.
Preconditions
An unemployment insurance budget that can pay course costs
A network of regional adult training centres run by the agency
County agencies that refer registered jobseekers into courses
Where else it has been used: Romania's plan targets fell with delivery (28,877 places planned for 2023, 24,440 for 2025), so the agency planned the decline as well as reporting it.
Limits: The agency reports on itself; the figure is on page 17 of a 59-page PDF, and the plan target (24,440) is on the same page. The 2023 peak is from the agency's 2023 activity report (https://www.anofm.ro/wp-content/uploads/2024/09/Raport_activitate_ANOFM_2023.pdf, 27,843 people, 25,420 unemployed), and the 2024 report gives 18,859, so the fall is over two years; earlier years were not read because the 2021 and 2022 report files are not text-searchable, so 2023 is the highest year checked, not a demonstrated all-time peak. The count mixes short courses, competence assessments and workplace apprenticeships (2,257 apprentices in 2025 against 591 in 2024, so the classroom fall is steeper than the total), and counts enrolments, not completions. The registered unemployment rate is from the same reports (2.89 percent average in 2023, 3.24 percent in 2025), so the fall is not explained by fewer unemployed. Against roughly 12 million Romanians aged 25 to 64 the programme reaches about one adult in 800, so it is a delivery record, not a participation rate. The 2002 start is the year the agency's training remit was set in law (Law 76/2002), not read at the source.
๐ท๐ธSerbiasince 2009operating below its peakbears on Adult learning participation
Serbia's National Employment Service put 1,613 unemployed people into labour market training (obuke za trziste rada) in 2024, but in 2025 it delivered none of the 1,100 places planned for that year for lack of accredited providers, and only 626 people trained under contracts carried over from the 2023 and 2024 plans.
The public employment service buys short vocational courses for registered unemployed people from a catalogue it procures each year, and since the adult education law only from providers accredited as publicly recognised adult education organisers.
Preconditions
A national catalogue of training matched to local vacancies
Enough accredited private and public training providers in each district
A procurement cycle that closes early enough in the year for courses to run
Where else it has been used: Tying public training money to an accreditation register raises quality on paper; Serbia's 2025 report shows the cost when the register is thinner than the demand.
Limits: The agency reports on itself, in a 127-page PDF in Serbian Cyrillic: Table 6 is on page 23 and the statement that the measure was not realised in 2025 ('dovelo do nerealizacije ove mere u 2025. godini') is on pages 28 to 29, where the agency attributes it to too few training programmes with publicly recognised adult education provider (JPOA) status at the locations needed. The 2024 figure is in the 2024 report (Izvestaj o radu NSZ za 2024. godinu, Table 6, on the same agency page), not at the cited URL. 2024 is the previous year, not a demonstrated peak: the 2019 report gives 1,257. The scale is very small for a country of about 6.6 million: other NSZ measures in the same table (internships, functional basic adult education with 1,157 participants in 2025, employer-requested training with 601) are separate counts and were not added in. The counts are people included, not completions or job outcomes, and cover only the registered unemployed. A one-year procurement and accreditation failure may reverse; the record documents a dated loss, not a permanent end. The started year is approximate: the measure appears in the agency's reports from at least 2010 and the 2009 start is the year the current employment law took effect, not read at the source.
๐น๐ณTunisiasince 2000still operatingbears on Adult learning participation
Tunisia's Ministry of Social Affairs, through its national adult education centre and regional coordinations, taught 16,560 adult learners in the 2021-2022 school year, 10,320 in the basic cycle and 6,240 in the completion and intensive cycles, 9,799 of them women, against a forecast of 22,710.
The social affairs ministry runs a national adult education centre that hires contract teachers to give literacy and post-literacy classes in every governorate, and reports learners by governorate and cycle in its performance budget.
Preconditions
A performance-budgeting system that makes ministries report indicators and enrolment annually
A pool of contract teachers paid from the ministry budget
Local venues and regional coordinators close to rural learners, especially women
Limits: The ministry reports on its own programme in its annual performance report, in Arabic. The numbers sit in a PDF: the narrative on page 43 and the governorate table on page 78; the text layer of this PDF maps digits wrongly (it extracts 16.362 and 22.912), so the figures were read from the rendered pages. The count is learners enrolled in literacy (basic) and post-literacy (completion and intensive) classes, not completions, and literacy classes are only part of adult learning. Against roughly eight million Tunisian adults it is a small reach, about two learners per thousand adults, in a country whose own report puts adult illiteracy at 16.8 percent in 2022. The report attributes the shortfall against forecast to the loss of contract teachers (954 at end 2021 to 943 at end 2022) and to difficulty attracting learners, and the 2024 report still speaks of 'targeting 22 thousand beneficiaries' with 885 contract and 113 tenured teachers, so the programme has not regained its forecast scale; no earlier enrolment peak was verified, so the record is not filed as an erosion. The 2000 start is the commonly reported launch year of the national programme and was not read at the cited source.
๐ฒ๐ฆMoroccosince 2015still operatingbears on Adult learning participation
Morocco's national literacy agency (ANLCA) and its partners registered 763,744 adults in literacy programmes in 2023-2024, 570,507 at level 1 literacy and 193,237 in post-literacy, with about 81 percent of level 1 registrants women and 62 percent rural.
A national agency funds and certifies literacy classes that associations, the religious affairs ministry and other departments run in villages, mosques and neighbourhoods, and counts registrants each programme year.
Preconditions
A dense civil-society network able to host classes close to where adults live
State subsidy that pays associations per class, with EU budget support at the margin
A registration system that collects counts by programme, region, sex and residence
Limits: The agency reports on its own programmes and the page carries no publication date. The counts are registrations, not completions or people who became literate: the Cour des comptes (annual report 2023-2024, principal axes, pages 49 to 52) found an average attendance index of about 40 percent and an assiduity index of 43 percent in 14,263 inspected classes over 2019 to 2022, so attendance is far below registration. Delivery runs mostly through civil-society associations under agency contracts (385,442 of the 570,710 level 1 registrants in the page's partner table) and through the Ministry of Habous mosque programme (165,290), so the agency finances and certifies rather than teaches; the partner table total (570,710) differs by 203 from the headline level 1 figure (570,507), and the page does not explain the gap. Literacy courses are basic education, which is only part of the adult learning construct. Press reports, citing the agency and the Cour des comptes, put registrations at over one million in 2023; that figure was not found at an official URL this protocol could open, so no peak is claimed, but if it holds for 2022-2023 the 2023-2024 count is about a quarter lower. The 2015 start is the year the Cour des comptes gives for the launch of the agency's programmes, not a date on the cited page.
๐ท๐บRussiasince 2019delivered and closedbears on Adult learning participation
Under the federal project Sodeystvie zanyatosti of the national project Demography, which paid for vocational training and further professional education of selected groups of adults, more than 80,000 citizens had started training and 21,125 had completed it by the end of the third quarter of 2024.
The labour ministry paid regional employment centres and accredited providers to train target groups of adults in short courses tied to regional demand, under a national project with quarterly reporting to the deputy prime minister.
Preconditions
A national project budget line with targets reported quarterly to the centre of government
Regional employment centres able to enrol and refer trainees
A pool of accredited training providers, including universities and colleges
Limits: The source is a government press release reporting a meeting, not a statistical table, and the counts cover only nine months of 2024, the project's final year, so they are a partial year and understate annual scale. An earlier release (government.ru/news/50674, 22 January 2024) says more than 190,000 people had completed short courses under the Demography retraining programme, without stating a period, so the two figures cannot be compared as an erosion. The training is for eligible groups (jobseekers, people aged 50 and over, women on parental leave and others defined by government resolution), not all adults. The national project Demography ended in 2024 and the release says its lines move into new national projects from 2025, so the status is concluded; the successor was not verified. The 2019 start is the first year of the national project Demography; the federal project took its current training form in 2021.
๐บ๐ฟUzbekistansince 2018still operatingbears on Adult learning participation
Uzbekistan's Ministry of Poverty Reduction and Employment expanded short-term vocational skill training of unemployed people from 20,000 in 2018 to 110,000 in 2022, counting graduates of the regional monocentres and district vocational training centres, according to ministry data reported by the World Bank.
The employment ministry built a network of regional training monocentres and district vocational centres that give registered unemployed people short, free skill courses, and counts graduates centrally.
Preconditions
A labour ministry with a district-level presence and registration of the unemployed
Budget for free short courses and for equipping training centres
Donor co-financing for equipment and curricula in the early years
Limits: The numbers are the ministry's, passed on in a World Bank appraisal document as background, rounded to thousands and without a table; the ministry's own release was not found at an open URL. The footnote says the numbers include graduates of both the regional monocentres and district centres, so the figures are graduates rather than enrolments, and the text calls them 'coverage'. Only two years are given, so the path between 2018 and 2022 and anything after 2022 are unknown; 'operating' rests on the appraisal describing the system as current in 2024 and planning further support to it. The training is for registered unemployed people, not all adults, and courses are short (weeks to a few months). The 2018 start is the base year of the series, not a programme launch date read at the source.
๐ฆ๐ฟAzerbaijansince 2021still operatingbears on Adult learning participation
Azerbaijan's State Employment Agency trained 24,399 successful applicants to the national Self-Employment Program in business skills and business planning between May 2021 and November 2024, 2,155 of them in the third quarter of 2024, as a condition of receiving start-up assets.
The employment agency selects vulnerable jobseekers into a self-employment programme, trains them to write a business plan, and transfers start-up assets to those whose plans pass, with mentoring visits afterwards.
Preconditions
A registry of vulnerable jobseekers and a selection process for applicants
Contracted business trainers able to run batches across regions
Budget for in-kind asset transfers, here partly loan-financed
Limits: The count is from the World Bank's monitoring of a loan it finances, reported by the implementing agency, and covers only programme participants financed under the project from May 2021, not the Self-Employment Program's earlier cohorts, which the state budget financed from 2017. The training is short business training that gates an asset transfer, not vocational or general education, so it is a narrow slice of adult learning. The same report gives 27,452 SEP beneficiaries 'in training or have completed trainings' and 'over 27,000' participating in business training, so the denominator varies by indicator; the completed count is used. About 3,000 people a year passed through the ministry's two vocational training centres, by the 2020 appraisal document's account, and the vocational certification and on-the-job training windows of the restructured project had not started by December 2024.
๐ฏ๐ฒJamaicasince 1982still operatingbears on Adult learning participation
Jamaica's HEART/NSTA Trust, the national training agency financed by a 3 percent payroll levy, enrolled 116,918 people in its own technical and vocational, work-based and second-chance programmes in the 2022-23 financial year, including 13,918 in the High School Diploma Equivalency programme, and 129,957 when Ministry of Education and private providers are counted.
A statutory trust collects a payroll levy from employers and spends it on free institutional, community and workplace training and on a second-chance secondary diploma, setting enrolment targets every year.
Preconditions
A payroll levy collected by the tax authority and ring-fenced for training
A network of training institutions and community sites across the island
A national qualifications framework (NVQ-J) so short courses lead to recognised certificates
Limits: The agency reports on itself in a 132-page PDF; the table is on page 39 and the HSDE figure on a later programme page. Counts are enrolments in programmes and interventions, not unique adults, and one person can enrol twice in a year. The total includes the HOPE National Service Corps and the School Leavers' Training Opportunities Programme, which serve school leavers and young people, and the agency gives no age breakdown, so the share of adults beyond initial education is unknown. The headline 129,957 also counts 11,104 enrolments in Ministry of Education CAP and other schools, a secondary-school programme, which is why the record uses the 116,918 HEART total instead. Enrolment is not completion: the same report gives 51,541 NVQ-J certificates for the year. The figure is one year against planned targets, not a series, so the record says nothing about trend. The 1982 start is the year the original HEART Act created the Trust as commonly reported; the report itself cites the HEART Act 2021 under which the merged HEART/NSTA Trust now operates.
๐ณ๐ตNepalsince 2018delivered and closedbears on Adult learning participation
Nepal's Ministry of Education, Science and Technology, through the EVENT II project secretariat, put 111,062 people aged 16 to 40 through market-relevant short-term vocational training between 2018 and 2023, delivered by 142 private and public training providers in 68 of the 77 districts and paid partly on trainees' employment six months after completion.
The education ministry bought short courses from accredited private and public providers and paid them per trainee, with part of the payment withheld until the trainee was verified as employed six months later, and paid more for women, disadvantaged youth and the poorest districts.
Preconditions
A pool of private training providers able to open courses in rural districts at short notice
A joint monitoring team to check enrolment, completion and employment before payment
A skills testing body (NSTB) to certify trades nationally
Where else it has been used: The model came from the Swiss-funded Employment Fund and ran through EVENT I before EVENT II; each version was a time-bound donor project, which is the weakness the pattern carries.
Limits: The figure is in a PDF completion report written by the World Bank about its own project, from data supplied by the ministry's project secretariat and checked by a joint monitoring team of local education officers, CTEVT, the ministry and SDC-Helvetas, which is not fully independent of the delivery; the Bank rated its own outcome. It counts trainees over a five-year project, not participation in the last 12 months, and amounts to roughly 22,000 a year against a labour force in the millions. The project defined youth as 16 to 40, so the count mixes recent school leavers with older adults and gives no age split. The total was 96.5 percent of a 115,000 target and training stopped for about 18 months during the pandemic. The training was financed by IDA credit, not the national budget, and the report does not say whether the government continued the results-based short-term training after the July 2023 closing, so 'concluded' describes the project window, not a measured end of the model. Employment rates of 70 to 83 percent six months after training are from the project's own tracer verification.
๐ณ๐ฟNew Zealandsince 2012operating below its peakbears on Adult learning participation
New Zealand's adult and community education (ACE) delivered by universities, polytechnics, wฤnanga and private training establishments enrolled 23,610 domestic students in 2021, down from 49,365 in 2012, according to the Ministry of Education's Education Counts tables.
The Tertiary Education Commission funds short, often non-credit community courses for adults through tertiary providers, and the education ministry counts enrolments each year from provider returns.
Preconditions
A dedicated ACE funding pool inside the tertiary funding system
Providers with community reach willing to run low-margin courses
A single student enrolment return that captures non-credit learners
Where else it has been used: Headcount halved while study load held, so the remaining ACE places went to fewer people taking longer courses; whether that was a deliberate shift toward longer, credit-bearing courses is not shown by the table.
Limits: The cited file is an Internet Archive copy: educationcounts.govt.nz answered 403 behind a bot check on 2026-10-02, and the archived September 2026 page that links a newer edition of the same spreadsheet returned 410 for the file, so the series stops at 2021 here although the ministry has published later years. The table counts only ACE delivered by tertiary providers; it leaves out ACE run by schools and community organisations, and 2012 is the first year of the table, not a demonstrated peak. Any funding change before 2012, including the reduction of ACE funding in schools commonly dated to 2009, is outside the series and was not read at a source for this record. Most of the fall is two steps: universities from 13,180 to 3,040 between 2012 and 2013, and polytechnics (now Te Pลซkenga) from 35,235 in 2016 to 27,975 in 2017 and 20,890 in 2020; the 2020 and 2021 counts are also depressed by COVID-19 lockdowns. Equivalent full-time student units held near 3,200 to 4,100 throughout, so fewer people took longer courses, and the fall in headcount is not a fall in hours of learning. Counts are students enrolled at any time in the year, rounded to the nearest 5, and students in two sub-sectors count in each. The started year is the first year of the published table, not the start of ACE funding, which is decades older.
๐ต๐ฐPakistansince 2002operating below its peakbears on Adult learning participation
Pakistan's National Commission for Human Development opened 19,399 adult literacy centres enrolling 482,295 learners in 2009-10, but its literacy programme 'came to a halt since 2013 due to scarcity of funds', and in 2021 it ran a phase of 1,490 centres, financed by the Pakistan Human Development Fund, that enrolled 36,529 learners, mostly women aged 15 to 45.
A federal commission recruited local teachers to run six-month literacy centres in villages, mostly for women, funded from the federal development budget; when that budget line lapsed, delivery fell to small donor-funded rounds.
Preconditions
A federal agency with district offices able to recruit and supervise thousands of local teachers
A recurring development-budget allocation, since each centre runs only one cohort
A political mandate for adult literacy that survives the devolution of education to provinces
Where else it has been used: Bangladesh's Bureau of Non-Formal Education ran the same project-by-project model (bgd-adult-literacy) and shows the same pattern of large reach followed by a gap when funding windows close.
Limits: The commission reports on itself. The 2021 figure is in the foreword of the 2021 annual report (PDF page 10, printed page 8, 'Around 36,529 learners, mostly women benefitted') and in its programme chapter (PDF page 38), which says 1,490 centres were established in 116 districts 'wherein, 36529 learners were enrolled' and 33,377 graduated; 1,500 centres were allocated and ten in North Waziristan could not be opened. The peak comes from a different document at the same host, NCHD Annual Report 2012 (https://www.nchd.org.pk/SiteImage/Misc/files/NCHD_Annual_Report_2012.pdf, PDF page 16), whose table gives 19,399 centres and 482,295 learners for 2009-10, 8,317 and 197,907 for 2010-11 and 8,270 and 202,349 for 2011-12; 2009-10 is the highest year in that table, not a demonstrated all-time peak, and the commission's cumulative claim of about 4.07 million people made literate since 2002 implies that earlier years were also large. The statement that the programme halted in 2013 for lack of funds is in NCHD Annual Report 2016 (https://www.nchd.org.pk/SiteImage/Misc/files/Annual%20Report-2016.pdf, PDF page 8), which also says the 2017 project approval cut the planned centres from 12,000 to 6,000. Enrolment is not completion or tested literacy. The 2021 centres were financed by the Pakistan Human Development Fund, a support fund set up under the commission's ordinance, rather than the commission's core development budget. Part of the fall may reflect the devolution of education to the provinces after the 2010 constitutional amendment, which moved literacy work to provincial departments that this record does not count. No annual figure after 2021 was found at the commission's site.
Trinidad and Tobago's YTEPP Limited, the state company that trains people aged 15 to 35 who are out of work or school, enrolled 1,601 persons in October 2024 to June 2025, 61 percent of its target, against 2,301 trainees in the Youth Training Programme over the same October to June window of 2019-20, according to the Ministry of Finance's Social Sector Investment Programme reviews.
The state funds a training company that runs short vocational and life-skills courses in centres across both islands for young people outside work and school, and reports enrolment against a yearly target in the budget's social sector review.
Preconditions
A recurrent transfer from the budget, which in a hydrocarbon economy moves with energy revenue
Training centres spread across districts, including Tobago
A national vocational qualification (CVQ/NEC) that short courses can certify against
Where else it has been used: Caribbean youth training agencies founded in the 1980s (HEART in Jamaica, YTEPP in Trinidad and Tobago) share the design. YTEPP depends on a yearly budget transfer that fell after 2016, while Jamaica's levy-financed HEART reported growing enrolment (jam-heart-nsta).
Limits: Both counts are the ministry's own programme reporting, collected in budget documents. The earlier figure is in a different edition, Social Sector Investment Programme 2021 (https://www.finance.gov.tt/wp-content/uploads/2020/10/Social-Sector-Investment-Programme-2021-1.pdf), and the two may not count the same thing: 2021 counts trainees in the Youth Training Programme, 2026 counts persons in 'youth training and employment programmes', and 2026 separately reports 1,323 youths in CVQ/TVET, NEC or YTEPP-certified programmes. 2019-20 is not a demonstrated peak; it is the earliest comparable count found, and its window already includes the first COVID-19 closures from March 2020, so the true fall is probably larger. The money moved the same way: the SSIP 2017 table gives an estimate of TT$111 million for YTEPP in 2016, SSIP 2019 gives TT$64.4 million actual spending in 2017, and SSIP 2026 lists TT$41.4 million, but transfers between ministries make those columns hard to align, so they are context, not the metric. The programme serves ages 15 to 35, so part of it is young people just out of school rather than adults returning to learning. Enrolment is not completion. The 1988 start is YTEPP's founding year as commonly reported; it was not read at the cited source.
๐ฆ๐นAustriasince 1998dismantledbears on Adult learning participation
Austria's Public Employment Service (AMS) paid Weiterbildungsgeld from 1998 to employees taking up to a year of educational leave, an average of about 14,000 recipients in 2021 at a cost that rose from 236 million euros in 2019 to 646 million in 2024, until the Nationalrat ended the benefit on 31 March 2025 in a budget consolidation law.
The federal government let employees suspend their contract for training and had the AMS pay an unemployment-benefit-level allowance out of the unemployment insurance fund; parliament then cut the allowance outright when costs tripled and evaluations found weak employment effects.
Preconditions
An unemployment insurance fund large enough to carry a training benefit without a separate appropriation
Labour law that lets an employee suspend a contract by agreement and return to the job
Loose eligibility on course content, which drove both growth and the case for abolition
Where else it has been used: The successor scheme keeps the leave but caps the budget at 150 million euros a year, tightens eligibility and makes higher-earning employees' employers co-fund the allowance, which is the targeting the Court of Audit asked for in 2023.
Limits: The recipient count is an annual average of people on leave at once, not a share of adults who trained in a year, so it cannot be set against the indicator's unit. The Court of Audit found the training requirements so weak that the leave could fund low-value courses or plain time out of work, and that uptake skewed toward people already highly educated, so the programme's peak is partly a peak in subsidised leave rather than in learning. The 2024 spending figure is the parliament's summary of the government's justification, it includes social insurance contributions, and it is a cost, not a participation number. Abolition is not the end of state support for training leave: a successor Weiterbildungszeit with a Weiterbildungsbeihilfe capped at 150 million euros a year took effect in January 2026 with stricter eligibility (PK0907, 16 October 2025), so a reviewer could call this erosion rather than dismantling. The record documents that the benefit ended by decision; it does not show that adult learning participation in Austria fell as a result.
The United Kingdom's financial regulator built the first regulatory sandbox in 2016, and by the end of 2022 had accepted 168 firms and products to test with real customers under supervision.
The regulator wrote a bounded testing regime: a firm without full authorisation can trial a product on real customers for a fixed period, with disclosure requirements and an exit plan, while the regulator watches from inside.
Preconditions
A single conduct regulator with statutory room to waive its own rules
A fintech sector dense enough to fill cohorts
Political cover for a regulator explicitly helping firms it will later police
Where else it has been used: The sandbox is Britain's most copied regulatory export, and several dozen jurisdictions now run one. Most copies keep the name and drop the supervisory intensity that makes it work.
Limits: The count is the regulator reporting on its own programme, and acceptance is not success: the FCA does not publish how many tests led to authorisation, funding or a product that survived. One hundred and sixty-eight firms over seven years is small next to the sector, so the number evidences that the mechanism runs, not that it moves the industry.
๐ฎ๐ฑIsraelsince 1993delivered and closedbears on Venture deals
Israel put 100 million dollars of public money into ten hybrid venture funds from 1993, with foreign partners and a buy-out option, and by 2002 the country had 131 venture funds with around 10 billion dollars under management.
The state capitalised new private funds rather than picking companies, required each fund to bring a foreign venture firm as partner, and sold the government's stake cheaply to the private partners once a fund worked, so the subsidy was the upside, and it expired by design.
Preconditions
A stock of investable technology and founders already present
Foreign venture firms willing to teach the craft for cheap equity
A government able to exit a working programme instead of scaling it
Where else it has been used: Yozma is the most copied venture policy in the world, and most copies fail on the same two points: no credible foreign-partner requirement, and no willingness to sell the state's stake at the moment of success.
Limits: Attribution is the weak joint: the ten funds were one cause among several, alongside a million-person skilled immigration wave, military technology spillovers and a US stock market that wanted Israeli listings, and the OECD paper credits Yozma as catalyst rather than cause. The 10 billion figure is capital under management, not investment made or returns earned, and roughly 70 percent of it was foreign money. The programme was deliberately closed once private capital led, which is why the status is concluded.
๐ฐ๐ทSouth Koreasince 2019still operatingbears on Firms through regulatory sandboxes
South Korea expanded its regulatory-sandbox system from ICT and industrial-convergence fields in January 2019 to eight sectoral regimes by 2024, adding finance, regulatory-free zones, smart cities, R&D zones, mobility and circular economy testing.
South Korea created a common regulatory-sandbox architecture with temporary permits, demonstration exemptions and rapid regulatory confirmation, then added sector-specific tracks and dedicated support bodies so smaller firms could test products without negotiating every rule from scratch.
Preconditions
A legal framework that allows time-, place- and scale-limited exemptions while retaining safety controls
Specialist ministries and intake bodies able to assess experiments across finance, mobility, ICT and industrial sectors
A route from a successful test to regulatory revision, licensing or procurement rather than leaving pilots stranded
Where else it has been used: Sandbox systems travel when experimentation, safeguards and rule-change pathways are designed together; multiplying sector labels without review capacity creates a catalogue rather than a learning system.
Limits: The regime count shows institutional breadth, not the number of firms approved, experiments completed, commercial products launched, safety outcomes or permanent rule changes. The government portal describes the system's design and expansion but does not independently evaluate whether each sectoral track is active at the same intensity or produces net social benefit.
๐ธ๐ฌSingaporesince 2016still operatingbears on Firms through regulatory sandboxes
Singapore's Monetary Authority operates three FinTech Regulatory Sandbox variantsโSandbox, Sandbox Express and Sandbox Plusโgiving firms distinct routes to test regulated products with real customers under defined safeguards.
MAS made experimentation a staged regulatory service: firms needing tailored controls use the original sandbox, lower-risk activities can use Sandbox Express, and promising proposals can receive a streamlined route with support. Each variant keeps a live test bounded while giving the regulator evidence before a full licence.
Preconditions
A financial regulator with authority to vary requirements temporarily and monitor live customer activity
Clear entry, boundary and exit criteria so a sandbox does not become indefinite unlicensed operation
A dense financial and technology ecosystem in which tested products can find counterparties, capital and a path to licensing
Where else it has been used: Regulatory sandboxes travel best as a portfolio matched to risk: one universal process can be too slow for low-risk tests and too loose for products that move real money.
Limits: Counting three sandbox routes shows regulatory design and breadth, not applications received, firms graduating to full licences, customer protection, failure rates or the cost of supervision. The EDB account is an official ecosystem description and includes illustrative company cases rather than an independent evaluation of all experiments.
The British Business Bank reported that government-owned British Patient Capital had committed ยฃ2.287 billion through 80 fund commitments, 10 co-investments and 18 Future Fund deals by 31 March 2024, with an underlying portfolio of 1,350 companies.
The UK used a public, long-horizon investment platform to anchor private funds, make selected co-investments and keep capital flowing into innovative companies that can be underserved by shorter-horizon finance. A fund-manager network spreads sourcing and specialist judgement beyond the state balance sheet.
Preconditions
A public investment institution with a mandate, capital and governance that can tolerate long time horizons
Private fund managers and co-investors able to select, price and support high-growth firms
A pipeline of investable companies and follow-on capital so an anchor commitment can compound rather than stand alone
Where else it has been used: Public venture catalysts travel when they crowd in capable private managers and preserve commercial decision-making; a state fund without deal flow, follow-on finance or disciplined governance can become a subsidy ledger rather than a capital market.
Limits: Commitments, assets and portfolio-company counts are administrative programme totals, not venture capital as a share of GDP, additionality, exits, jobs or causal economic impact. The portfolio includes third-party fund capital and co-investments, and the figures are reported by the government-owned delivery institution rather than independently evaluating what would have happened without it.
UK Research and Innovation's ICURe impact report says spinout companies from the programme secured ยฃ326 million in investment funding and created more than 1,400 jobs from 2014 onward, after research teams were trained and funded to test markets before deciding whether to spin out.
ICURe gives research teams funding, coaching and a disciplined market-discovery period in which they conduct more than 100 customer interviews before a university-backed go/no-go decision. Teams advised to spin out can then apply for follow-on funding, turning technology-transfer work into an investable proposition rather than treating a patent as a company.
Preconditions
Universities and research institutes with discoveries, translational staff and authority to support commercial routes
Researchers willing to test a problem with users and founders able to lead a company after the discovery phase
Follow-on grants, investors and technology-transfer offices that can carry a validated opportunity beyond the programme
Where else it has been used: Spinout pipelines travel when market discovery, founder formation and follow-on capital are designed as one chain; a grant or patent alone does not create the commercial capabilities and decisions needed for a durable firm.
Limits: The impact report's cumulative investment and job figures cover teams supported by ICURe and do not show the full spinout population, survival, profitability, quality of jobs, counterfactual or the programme's independent contribution. They are programme impact estimates, not a national spinout rate or a guarantee that every participating discovery should become a company.
Portugal Ventures, the public venture-capital company in the Banco Portuguรชs de Fomento group, reports โฌ233.2 million invested in 256 new companies since 2012 and โฌ283 million under management.
A state-backed investor supplies pre-seed, seed and Series A capital, often alongside private investors, using staged minority investments and portfolio support to bridge companies that are too early for ordinary bank finance.
Preconditions
A pipeline of founders and research-derived firms large enough to fill investment rounds
Investment professionals with authority to make commercial decisions inside a public mandate
Follow-on private capital and exit markets so public money can recycle rather than become permanent subsidy
Where else it has been used: Public venture capital travels when it crowds in follow-on investors and accepts portfolio risk transparently; a state fund that only replaces missing private money can preserve firms without building a market.
Limits: These are company-reported cumulative portfolio figures, not venture-capital investment as a share of GDP, additionality, exits, survival, jobs or a counterfactual. The homepage does not give a full reference-year methodology, and the totals can include private co-investment effects as well as Portugal Ventures' own capital.
๐ต๐นPortugalsince 2022still operatingbears on Firms through regulatory sandboxes
ANI's National Innovation Report records 47 recognised Portuguese test beds and โฌ150 million allocated to the national network for developing and testing new products and services at technology-readiness levels 5 to 9.
Portugal funded and recognised shared facilities where SMEs and startups can develop and test near-market products, lowering the cost of moving an experiment from prototype toward demonstration without requiring every firm to build its own laboratory.
Preconditions
Facilities with technical staff, equipment and safety procedures that can host external firms
A recognition and funding process that makes test results credible to customers and investors
A pipeline of SMEs and startups with products mature enough to test at TRL 5 to 9
Where else it has been used: Test-bed networks travel when access, measurement and procurement pathways connect a facility to real adopters; a list of subsidised facilities without utilisation data can become an infrastructure inventory rather than an experimentation capability.
Limits: The count and allocation show a national experimentation infrastructure, not the number of live trials, firms served, products commercialised or regulatory changes that resulted. ANI's report says 47 were recognised at publication but gives no common utilisation definition, and the โฌ150 million is an allocation rather than completed spending or additional private investment.
๐ต๐นPortugalsince 2021still operatingbears on Firms through regulatory sandboxes
ANI's National Innovation Report records four Portuguese Zonas Livres Tecnolรณgicas (technological free zones) by the end of 2024, where innovators can test technology-based processes in real or near-real environments under continuous regulator supervision.
Portugal created a legal route for regulators and innovators to run controlled tests in defined physical environments, using supervised evidence from those tests to clarify or adapt rules where existing law is uncertain.
Preconditions
Regulators willing to supervise experiments while protecting safety and public rights
A legal framework that permits controlled derogations or clarifications without suspending accountability
Test environments with monitoring, reporting and a route to ordinary market authorisation
Where else it has been used: Regulatory sandboxes travel when the evidence gathered in a zone can change a rule or unlock a normal licence; a protected pilot site with no route out simply postpones the regulatory decision.
Limits: Four zones are an institutional count, not a measure of trials, firms admitted, time to authorisation or successful regulatory adaptation. ANI's report uses a 31 December 2024 cutoff and does not provide a common utilisation or outcome series, so the record shows that a sandbox channel exists rather than that it is broad or effective.
๐ง๐ทBrazilsince 2021delivered and closedbears on Firms through regulatory sandboxes
Brazil's central bank opened its regulatory sandbox to applications in February 2021, received 52 projects, admitted seven to test under relaxed rules from 6 December 2021, and ran the cycle for one year with one permitted extension.
The National Monetary Council and the central bank set a joint sandbox rule in 2020, and a standing committee of seven department heads screened applications by cycle, admitting only projects that needed a specific rule relaxed to be tested.
Preconditions
A central bank that already runs the payment rails it is willing to test against, as Pix shows
A legal basis that lets the regulator waive its own rules for a limited period
Enough supervisory staff to take on a cycle of bespoke oversight
Where else it has been used: The design follows the UK model of time-bound cohorts. Brazil kept the cohort small and the entry test strict, which limits both the risk and the reach.
Limits: The central bank is reporting on its own programme. Seven admissions out of 52 shows a selective regime, not a broad one: the report records that the bank found 24 projects not innovative enough or not needing any rule relief, judged five outside its remit, and disqualified 16 more. The report covers the first cycle only and says nothing about what the seven tests changed in regulation or which products reached the market. No second cycle appears in the bank's published sandbox documents as of the retrieval date, so the status is concluded on that absence, not on a closing statement. The securities regulator CVM runs a separate sandbox that this record does not count. The source is a PDF and the counts are on page 11.
๐ฎ๐ณIndiasince 2020still operatingbears on Firms through regulatory sandboxes
The Reserve Bank of India ran four themed sandbox cohorts between 2020 and 2024, retail payments, cross-border payments, MSME lending and fraud prevention, admitting 28 entities to test and finding 13 products viable in the three cohorts whose exits it published, and keeps the cohorts open on tap.
The RBI published an enabling framework in 2019 and ran the sandbox as themed cohorts, each opened on one policy problem, so the regulator chose the questions and firms competed to answer them; a closed cohort reopens on tap.
Preconditions
A central bank that sets policy for a fast-growing digital payments system it also oversees
Willingness to publish names of admitted entities and test outcomes
A pipeline of startups and incumbents with products held back by a specific rule
Where else it has been used: Themed cohorts are the RBI's variation on the UK model. The theme concentrates supervisory attention but caps volume, which is why the counts stay in single digits per cohort.
Limits: Both numbers are computed from four RBI press releases, not stated in one: six entities admitted and six viable in the first cohort (13 September 2021), eight and four in the second (19 July 2022), eight admitted to the third (6 June 2022) and six admitted with three viable in the fourth (18 June 2024). The third cohort's exit result was not found, so the viable count leaves it out. The regulator is reporting on its own programme, and viable within the boundary conditions of a test is not adoption: the releases say only that regulated entities may consider the products. Twenty-eight entities over four years is small next to India's payments sector, so the record evidences that a working sandbox exists, not that it shaped the market.
๐ฐ๐ชKenyasince 2019still operatingbears on Firms through regulatory sandboxes
Kenya's Capital Markets Authority opened a regulatory sandbox under a 2019 policy guidance note and by January 2024 had let 11 firms exit it after live testing, with new admissions continuing into January 2025.
The Capital Markets Authority issued a policy guidance note that let fintech firms and incumbents test capital markets products with real customers under relaxed rules, then used each exit as the moment to grant a licence or approval under the Capital Markets Act.
Preconditions
A regulator with statutory power to grant no-objections and tailored approvals
A domestic fintech sector already built on mobile money rails
Peer regulators abroad, through the Global Financial Innovation Network, to copy procedures from
Limits: This is one regulator's sandbox for capital markets products only. It says nothing about the Central Bank of Kenya, the Communications Authority or any other regulator, so it covers little of the breadth the indicator asks for. The count is the regulator's own, taken from a press release, and the CMA publishes no running register. The 11 exits are cumulative over almost five years, a small number that says nothing about what the exiting products did in the market. 'Successful exit' is the CMA's own description. The figure is from January 2024, and no later official total was found.
๐ณ๐ฌNigeriasince 2023still operatingbears on Firms through regulatory sandboxes
Nigeria's Securities and Exchange Commission opened a one-year Regulatory Incubation sandbox in September 2023 and had admitted seven firms across cohorts 001/23 and 002/24 to test capital-market products, then cleared seven more entities into its Accelerated Regulatory Incubation Programme in July 2026.
The SEC created a time-limited sandbox with conditional rule waivers and a year-long test period, then added an accelerated track to pull digital-asset firms that were operating informally into supervised testing.
Preconditions
A regulator with statutory room to grant conditional waivers, here strengthened by the Investments and Securities Act 2025
A pipeline of fintech and digital-asset firms already serving Nigerian users outside the rules
Supervisory staff able to review participants and decide graduation or wind-down
Where else it has been used: Built on the FCA sandbox template; Nigeria's variant adds an accelerated track aimed at bringing crypto exchanges and tokenisation platforms already in the market under supervision, rather than only testing new products.
Limits: The count of seven is ours, read off the directory table as retrieved on 2026-10-01, which carries no publication date and may not be current; the SEC publishes no summary statistic, application count, graduation rate or outcome for any participant. Three of the seven RI firms (Trovotech, Wrapped CBDC, Blockvault) reappear among the seven ARIP admissions of July 2026, so the two metrics overlap and must not be added, and an ARIP Approval-in-Principle is by the SEC's own wording not a final licence. This is one regulator's sandbox in one sector: the Central Bank of Nigeria's separate sandbox framework is not counted, and the record says nothing about whether testing changed any rule.
The Financial Services Regulatory Authority of Abu Dhabi Global Market opened its RegLab fintech sandbox to applications on 2 November 2016 and by August 2019 had admitted 32 firms from 83 applications from 20 countries, of which two had graduated to the live market.
The FSRA wrote a separate, time-limited authorisation category into its rulebook that let fintech firms test with real customers under tailored limits for up to two years, then admitted them in cohorts and moved the successful ones to full licences.
Preconditions
A financial free zone with its own common-law regulator able to change rules quickly
A federal state that tolerates several competing financial regulators and sandboxes
International fintech firms looking for a regional base
Where else it has been used: The design follows the UK FCA sandbox closely. In the UAE it was followed by the DFSA in Dubai and by the federal RegLab launched in January 2019 with the Dubai Future Foundation, so the country now runs several parallel sandboxes instead of one.
Limits: The count is the regulator announcing its own programme in a press release, and it dates from August 2019: no later cumulative total was found at an ADGM URL, so the record says nothing about the sandbox's size today. ADGM is a financial free zone in Abu Dhabi with its own regulator, so this is one jurisdiction's sandbox, not a federal regime. The UAE government portal (https://u.ae/en/about-the-uae/digital-uae/regulatory-framework/regulatory-sandboxes-in-the-uae) lists at least five others, including the federal RegLab launched in January 2019 with the Dubai Future Foundation and the DFSA's Innovation Testing Licence, and no official count of firms across all of them exists. Admissions measure regulatory openness, not whether tested products reached customers safely: two graduates out of 32 admitted by 2019 is a thin outcome, and the release does not say how many firms left without a licence.
๐ฎ๐ฉIndonesiasince 2018still operatingbears on Firms through regulatory sandboxes
Indonesia's Financial Services Authority (OJK), running a regulatory sandbox since 2018 and under POJK 3/2024 since February 2024, had received 26 applications and approved 9 sandbox participants by December 2025, four of which had completed testing and been declared passed.
OJK wrote the sandbox into binding regulation, first in 2018 and again in 2024 when it took over crypto and digital-asset supervision, and used graduation from the sandbox to open a registration route for whole business models rather than for one firm.
Preconditions
A financial regulator with a statutory mandate over the new activity, here extended to crypto assets by the 2023 financial-sector law
Staff able to supervise live trials and write graduation decisions that set precedent
A pipeline of firms willing to test under supervision instead of operating offshore
Where else it has been used: Regulatory sandboxes spread from the UK FCA's 2016 model; Indonesia's version shows the sandbox being reused as the entry gate when a regulator inherits a new sector.
Limits: The count of 9 covers only the regime under POJK 3/2024, so it omits every participant tested under the 2018 regulation, and the press release gives no cumulative total since 2018. The regulator reports on its own sandbox. The release's own breakdown names 4 in testing and 4 passed, which sums to 8, so one approved participant is unaccounted for in the text. The cohort is concentrated in digital financial assets and crypto (tokenised gold, tokenised bonds, tokenised property), so breadth across financial activities is narrow. A sandbox count shows that trials run, not that rules changed because of them, although the release does say that business models matching a passed participant may register without a trial. Bank Indonesia runs a separate payments sandbox that is not counted here. The OJK site rejects requests that do not send browser-like headers (a plain curl returned 'Request Rejected'), so automated URL checks may flag it.
๐ฒ๐พMalaysiasince 2016still operatingbears on Firms through regulatory sandboxes
Bank Negara Malaysia has operated a fintech regulatory sandbox since October 2016 and added an accelerated Green Lane for financial institutions in February 2024; as of June 2026 it reported 139 Standard and Green Lane applications submitted to date and five solutions in live testing.
Bank Negara Malaysia published a sandbox framework that lets fintech firms test regulated financial services live under tailored conditions, then added a faster Green Lane so incumbents with a strong risk-management record can test with lighter procedures.
Preconditions
A single integrated prudential and conduct regulator for banking, insurance and payments that can grant case-by-case flexibility
Legal powers to vary requirements for a defined test without amending statute
Supervisory staff able to assess and monitor novel business models one by one
Where else it has been used: BNM launched in the same year as Singapore's and after the UK's, and its 2024 Green Lane revision follows the wider move from start-up sandboxes toward faster testing tracks for incumbents.
Limits: The page is the regulator reporting on its own programme, as a live dashboard with no archived history, so the figures can change without notice and earlier values cannot be checked. 139 counts applications (the page also calls them applicants), not admissions: BNM publishes no count of participants admitted or graduated, so the share rejected, withdrawn or still in assessment is unknown. The same page's table of solutions currently in the sandbox lists six firms (one dating from June 2022) against the counter's five, so the live-test count is itself inconsistent on the source page. Five live tests after almost ten years is a small number, and the count says nothing about whether tested products reached the market or led to rule changes. It covers one regulator's sandbox; other Malaysian regulators' sandboxes are not counted here.
๐น๐ญThailandsince 2016still operatingbears on Firms through regulatory sandboxes
The Bank of Thailand, which opened a fintech regulatory sandbox in late 2016, listed 10 sandbox project tracks as of April 2026, whose participant counts sum to 90, from QR payment (14) and biometrics (19) to letters of guarantee (18), the National Digital ID (12) and programmable payment.
10 sandbox project tracks, project tracks listed under the bank of thailand regulatory sandbox, 2026-04. 90 participants, participants summed across the listed sandbox project tracks, 2026-04. Bank of Thailand, Regulatory Sandbox page, retrieved 2026-10-01.
The move, and what it needed
The Bank of Thailand let banks, fintechs and technology providers test payment, identity and lending services under limited scope and temporary relief, then wrote what worked into rules.
Preconditions
A central bank with legal room to grant temporary, conditional relief and to supervise tests closely
Incumbent banks willing to test shared standards inside the sandbox rather than lobby against them
A path from a successful test to a licence or a published standard, so the sandbox is not a holding pen
Where else it has been used: Sandboxes spread across ASEAN after 2016; Thailand's variant is notable for running thematic tracks (QR, biometrics, digital ID).
Limits: The count is the regulator's own listing on a web page rendered from its CMS, not an audited statistic, and the page does not say whether the participants are live today or cumulative since each track opened: the linked P2P list includes firms that exited or discontinued (one discontinued in 2022, one in 2023), so 90 is likely cumulative, not live. The total of 90 is computed by summing the table. A track is the Bank's own grouping, so 10 tracks is not comparable to a count of sandbox regimes elsewhere. The figures show breadth of testing, not how many tests became licensed products, what was learned, or whether consumers were protected; the started year rests on a Bank of Thailand research note, not the 2016 guideline itself. Other Thai regulators' sandboxes are not counted here.
๐ช๐ธSpainsince 2021operating below its peakbears on Firms through regulatory sandboxes
Spain's Treasury ran a national financial regulatory sandbox under Law 7/2020 from 2021. Its first cohort drew 66 applications, 12 of which entered testing. By 2025 demand had fallen to two applications across the ninth and tenth cohorts, one of which received a favourable assessment, and the Government approved a draft law (Anteproyecto, 17 December 2024) that replaces closed cohorts with year-round intake and cuts pre-assessment to 45 working days.
The Treasury built a cohort-based sandbox in which the Bank of Spain, the securities regulator CNMV and the insurance supervisor DGSFP assessed and monitored tests under one statute, then fed the conclusions back into legislation. Demand collapsed once the first cohort's backlog cleared, and the Government answered with a draft law for year-round intake and a shorter pre-assessment.
Preconditions
A statute that lets supervisors grant time-limited testing protocols without licensing
Supervisors with staff to assess applications and write public conclusions for each test
A steady supply of firms whose innovation is blocked by a specific rule, rather than one-off pent-up demand
Where else it has been used: The UK FCA's sandbox model was copied widely after 2016. Spain's cohort data show the common afterlife of such regimes: an opening surge, then thin demand once EU-level regimes and innovation hubs offer cheaper routes.
Limits: Applications measure demand for the regime. They do not measure the regime's regulatory value, and the Treasury reports its own programme. The first cohort's surge was partly pent-up demand from the two-year legislative delay before Law 7/2020, so 66 is an unusual peak and not a steady state. Some of the later fall may reflect substitutes rather than lost capability: the EU DLT Pilot Regime, MiCA, the Bank of Spain's innovation hub and new Spanish sandboxes such as the AI pilot under RD 817/2023, Eco-Lab and the mobility sandbox. Spain opened further sandboxes (AI pilot, Eco-Lab, mobility) modelled on this one. The same report counts 117 applications over ten cohorts, four projects fully in the market and three legal changes traced to tests: Law 6/2023 on DLT securities and fund units, and RD 668/2023 on pension contributions. Read as one regime, the decline is real. Read as Spain's whole sandbox capacity, it overstates erosion. Every figure is a PDF read and there is no machine-readable series. The cohort chart in section 1.1 gives per-cohort counts only as unlabelled values, so the record uses the two figures stated in prose.
๐ฟ๐ฆSouth Africasince 2020still operatingbears on Firms through regulatory sandboxes
South Africa's Intergovernmental Fintech Working Group, a joint body of the financial regulators, opened a cross-regulator sandbox in April 2020 that took 54 unique applications and admitted nine sandbox tests from eight firms in its first cohort, then moved to a rolling intake.
The Reserve Bank, the FSCA and other financial regulators ran one sandbox through a joint working group, so an innovator whose product crossed mandates such as crypto cross-border payments was tested by every relevant regulator at once.
Preconditions
Regulators willing to share one intake and one set of case managers across mandates
Enough supervisory staff to assess applications; the report says the volume of applications strained capacity
An Innovation Hub with a guidance unit to absorb the many applicants who do not need a sandbox
Where else it has been used: The IFWG moved from a cohort-based window to an adjusted rolling-based approach after its first round.
Limits: Nine tests is a small count, and the source is the regulators reporting on their own sandbox. The report does not show that any rule changed because of a test: it says two tests confirmed compliance with existing rules and seven produced insights for possible future frameworks. The 'operating' status rests on the October 2022 statement that the IFWG voted to move to a rolling intake. No official page that opened on the retrieval date counted participants after the first cohort, and the IFWG's own site (ifwg.co.za) did not respond to scripted requests. The sandbox covers financial services only, so it shows nothing about controlled trial regimes in other sectors. The figures are in a PDF.
๐ฏ๐ตJapansince 2018still operatingbears on Firms through regulatory sandboxes
Japan's cross-government regulatory sandbox, created in 2018 under the Productivity Improvement Act and made permanent in 2021 under the Industrial Competitiveness Enhancement Act, had approved 33 demonstration plans involving 152 businesses by 19 March 2025, across mobility, IoT, fintech and healthcare.
The Cabinet Secretariat opened one government-wide window where a firm can propose a time- and participant-limited trial, and the minister responsible for each affected law approves it after checking with an expert committee, so ministries must answer within fixed timelines and use the trial data to review the rule.
Preconditions
A cabinet-level secretariat with authority to route a proposal to every regulating ministry
A statute that lets ministers approve trials without first amending the underlying law
Firms willing to publish trial plans and results for the data to inform rule changes
Where else it has been used: Built on the UK FCA's fintech sandbox but widened from one regulator to all ministries; Korea's 2019 regulatory sandbox took the cross-sector form further and reports far higher approval counts.
Limits: The count comes from the sandbox's own secretariat in a promotional briefing, not a statistics release. Thirty-three plans over more than six years is about five a year, a modest rate for an economy of Japan's size, and the count says nothing about how many trials led to a permanent rule change; the annex lists some special measures (for example a 2020 insurance-law exception for P2P insurance) but the briefing gives no total of rules amended. A plan is counted at approval, not completion, and several approved plans are small single-firm demonstrations. The briefing's annex dates each approval: 19 of the 33 plans were approved by October 2020, and about three a year since, with one in 2024, so the scheme's use slowed sharply after its first two years. Japan runs other experimentation regimes (national strategic special zones, the grey-zone resolution system, the FSA fintech hub) that are not in this count, so it understates sandbox-type activity as the construct defines it while also overstating it as evidence of reform.
๐บ๐ธUnited Statessince 2019dismantledbears on Firms through regulatory sandboxes
The Consumer Financial Protection Bureau opened a national Compliance Assistance Sandbox in September 2019, issued two sandbox approval orders in December 2020, terminated one of them in June 2022 and rescinded the policy with effect from 30 September 2022; a reissued version of January 2025 was withdrawn in May 2025.
A federal consumer-finance regulator offered firms a statutory safe harbour and a written approval that a described product complied with named laws, in exchange for sharing data during the test. A later leadership judged that the policies did not advance innovation and lacked transparency, let the information-collection authority lapse and rescinded the policy.
Preconditions
Statutory safe-harbour provisions the agency could invoke for good-faith reliance on its approval
Leadership continuity across administrations, which this sandbox did not have
Guardrails stopping a firm from marketing an approval as an endorsement
Where else it has been used: Sandboxes that rest on an agency policy statement rather than a statute can be switched off by the next director; the CFPB's was rescinded, reissued and withdrawn within three years.
Limits: The count of two is ours, made by counting the approval orders the CFPB lists as granted (Payactiv and Synchrony Bank); the page also lists one template (Build Commonwealth) and separate No-Action Letters and Trial Disclosure waivers, which are not counted here. The rescission and the 2025 withdrawal are read from Federal Register notices on govinfo.gov (87 FR 58439, 90 FR 1974 and 90 FR 20086), not from the cited page. The CFPB page is blocked to some automated clients. Two approvals over three years is evidence that the federal consumer-finance sandbox barely operated before it was ended, not that it failed on its merits: the Bureau's stated reasons were that the policies did not advance consumer-beneficial innovation and failed to meet standards for transparency and stakeholder participation. State sandboxes (Arizona, Utah and others) are subnational and are not counted.
๐จ๐ฆCanadasince 2016delivered and closedbears on Firms through regulatory sandboxes
Canada's provincial securities regulators, acting jointly as the Canadian Securities Administrators, authorised 13 firms through the CSA Regulatory Sandbox between 2016 and 2021, before it was succeeded by the CSA Financial Innovation Hub and its decision list stopped being updated in December 2023.
Thirteen provincial and territorial securities regulators agreed to process fintech registrations and exemptive relief faster and on a time-limited basis through a shared sandbox, then folded it into a broader innovation hub once crypto platforms moved to their own registration track.
Preconditions
A passport-style system letting one principal regulator's decision apply across provinces
A wave of novel business models (crypto funds, token offerings) that did not fit existing categories
Regulators willing to publish each decision so the sandbox could be audited
Where else it has been used: Canada shows a sandbox absorbed once its main user group gets a dedicated rulebook: the crypto platforms that filled it moved to a pre-registration undertaking regime.
Limits: The 13 is ours, counted from the firms the page lists under 'Former CSA Regulatory Sandbox'; the crypto trading platform decisions on the same page belong to a separate pre-registration process and are not counted. Most authorisations were crypto funds and token offerings from 2017 to 2019, and the page says nothing about outcomes. The sandbox was succeeded by the CSA Financial Innovation Hub, whose page lists research, workshops and 'testing environments' among possible activities, so the testing function may not be wholly gone; what ended is the named regime and its public register, which is why the record is concluded rather than dismantled. The earliest listed decision (AngelList, October 2016) predates the sandbox's 2017 start and is listed under it by the CSA. The site sits behind a JavaScript challenge that blocks simple scripted fetches.
๐ฆ๐บAustraliasince 2020still operatingbears on Firms through regulatory sandboxes
Australia's government introduced the enhanced regulatory sandbox on 1 September 2020, letting firms test financial services and credit activities for up to 24 months without a licence, and ASIC's register lists 23 users, four current and 19 past.
Parliament and Treasury put a licensing exemption into regulation, so an eligible firm can notify ASIC and start testing within set client and exposure limits, rather than applying for case-by-case relief.
Preconditions
A statutory power to exempt from the licensing regime by regulation
Hard caps on clients and exposure so a notification regime stays safe without vetting
A public register so customers can check that a firm is a sandbox user
Where else it has been used: Australia moved from a regulator-run sandbox to a legislated notification exemption; low take-up suggests automatic entry does not by itself create demand.
Limits: The 23 is ours, counted from the two expandable lists ('Current users' 4, 'Past users' 19) on the day of retrieval; the page carries no headline total or reference date, and the lists change as exemptions start and lapse. Users notify ASIC rather than being selected, so the count measures take-up of an exemption, not supervised trials, and the page does not say how many past users went on to obtain a licence or simply stopped. Twenty-three users in six years is small for Australia's financial sector. ASIC's earlier sandbox, which the ERS superseded, is not counted.
๐ฉ๐ชGermanysince 2017delivered and closedbears on Firms through regulatory sandboxes
Germany's federal government ran a time-limited regulatory experimentation clause for the SINTEG smart-energy showcases from June 2017 to June 2022, under which the Federal Network Agency received 49 notifications of experiments but only two claims for compensation of regulatory costs.
The federal government wrote a time-boxed ordinance letting showcase participants trial flexible grid use and be compensated afterwards for network charges and levies they would otherwise have lost, with the network regulator verifying each claim.
Preconditions
A large funded demonstration programme to supply the experimenters
A regulator able to verify claims against network operators
State-aid clearance from the European Commission for the compensation
Where else it has been used: The lesson in the report is that ex post compensation with audited claims deters use, so a clause must waive rules up front to be taken up.
Limits: Both figures are inside a PDF (pages 4 and 5). The clause was open only to participants in the five funded SINTEG showcases, not to every firm in the country. A notification is an announcement of intent, not a completed trial, and the government's own report says use of the compensation mechanism fell well short of expectations because participants had to bear losses first and faced high administrative cost. The regulation expired by design in 2022. Because its core was ex post compensation, the report says the general legal framework applied essentially unchanged to participants, apart from the ยง 5 derogation for flexibility platforms.
๐ณ๐ฑNetherlandssince 2015operating below its peakbears on Firms through regulatory sandboxes
The Dutch economy ministry let energy cooperatives and owners' associations deviate from the Electricity Act 1998 through 19 exemptions granted in four rounds between 2015 and 2018, then closed the scheme to new entrants; by 2024 only 10 of the experiments were operational.
The ministry used a decree under the Electricity Act to let local cooperatives run their own project grids and supply members for up to ten years, testing whether the full act needed to apply to local renewable generation.
Preconditions
An enabling article in the Electricity Act permitting time-limited derogations
Energy cooperatives able to carry grid-operator duties
An agency (RVO) to vet applications and monitor results
Where else it has been used: The Dutch case shows a sandbox closed to entry while a successor law was drafted; experiments outside the new law's scope must either fit it or end when their exemption expires.
Limits: Both figures are inside a PDF (section 3, page labelled 4 of 12, PDF page 5). Nineteen is the total ever granted, used as the peak because no entry was possible after 2018. It counts exemptions and the 10 counts operating experiments, so the gap mixes projects that never started with ones that stopped; the sources do not show 19 operating at once. The only same-unit loss they document is from 19 exemptions granted to 15 still held. RVO's scheme page gives a third figure, 15 current exemption holders, which counts exemptions still held rather than experiments running. The minister closed the scheme on the advice of the Raad van State pending the new Energiewet (in force 1 January 2026), so part of the fall is a planned run-down rather than failure. The experiments are small local grids, and the report itself says their number and size limit what can be concluded.
๐ซ๐ทFrancesince 2020still operatingbears on Firms through regulatory sandboxes
Under the 2019 energy and climate law, France's energy regulator CRE and the energy directorate DGEC granted temporary derogations from network access rules to 30 projects, 25 by CRE and 5 by DGEC, of which six were running and reporting indicators in May 2025 and two had been generalised into regulation in 2024.
Parliament gave the energy regulator and the ministry a power to grant time-limited derogations from network access conditions, first through application windows and since November 2022 on a rolling basis, with annual public progress reports.
Preconditions
A statutory derogation power covering named titles of the energy code
An independent regulator able to assess projects and publish follow-up
Network operators willing to host and measure non-standard arrangements
Where else it has been used: Two experiments (Enedis Reflex and an EDF project) were generalised by regulation in 2024, the route from sandbox to rule that many sandboxes lack.
Limits: The figure is inside a PDF (page 3). A derogation granted is not a trial run: the same report says only six of the 30 were in progress and producing indicators, and 14 further requests were awaiting a decision, mostly at DGEC. Progress data come from project holders and network operators. The sandbox covers network access and use in electricity and gas only, so it says nothing about experimentation elsewhere in French regulation.
๐ฎ๐นItalysince 2021operating below its peakbears on Firms through regulatory sandboxes
Italy's FinTech sandbox, run by the Bank of Italy, Consob and IVASS under the MEF's FinTech Committee, admitted 16 projects from its first application window (November 2021 to January 2022) but only 3 from its second (November to December 2023), after applications fell by about 80.7 percent.
The MEF's FinTech Committee opened timed application windows in which each supervisor admitted projects to tests of six months, extendable to at most 18 months, with tailored conditions, and demand fell once operators learned how heavy admission was and found lighter informal channels at the Bank of Italy.
Preconditions
A statutory sandbox shared across three supervisors with separate admission procedures
Lighter informal innovation channels at the central bank competing for the same applicants
Short, infrequent application windows that make the sandbox slow relative to product cycles
Limits: The counts are projects, not firms: the IVASS projects in the second window involved several firms each (for example ITAS Mutua with ANIA, Autovia and Connexa InsTech), while the one Bank of Italy project was a single FinTech operator, so the indicator's firms-per-million figure would need a firm count the report does not give. Per million people the sandbox admitted roughly 0.27 projects from the first window and 0.05 from the second; those ratios are our computation with a population the source does not state. The sandbox runs in windows rather than continuously, so the two numbers are two cohorts, not two years. The MEF attributes the fall to an admission process that operators and the supervisors themselves found burdensome and to the Bank of Italy's informal channels (Canale Fintech, Milano Hub) absorbing demand, so part of the decline is innovation dialogue moving outside the sandbox, which the indicator would not see. The Bank of Italy's sandbox page shows both cohorts as concluded and lists no window after 2023 as of retrieval; a new window or a redesigned regime would change the status. The report is the MEF reporting on a regime it chairs, and the MEF itself judges the sandbox experience 'soddisfacente' while calling for procedural simplification.
๐ฌ๐ญGhanasince 2026still operatingbears on Firms through regulatory sandboxes
The Bank of Ghana, which launched a regulatory and innovation sandbox pilot in 2021, admitted six firms into its Regulatory Sandbox for one year in January 2026 to test the exchange, custody, administration and issuance of virtual assets.
The Bank of Ghana's FinTech and Innovation Office admits named firms for a fixed one-year test under supervision, using each cohort to validate rules it has not yet issued, here for virtual assets.
Preconditions
A central bank FinTech office with staff to supervise individual tests
A legal basis to grant time-limited, conditional permission outside existing licences
A pending regulatory question concrete enough to define a cohort around
Limits: The metric is a count of six firms in one cohort, taken from a press release rather than a statistical series; divided by Ghana's population it is well under one firm per million people. It counts admissions only, not completions or exits. The 2021 launch was a pilot run with EMTECH Service LLC, not a documented admission, so the start year is the first documented admission, 2026. The indicator sums every regulator in the country, and this record covers the Bank of Ghana only. A sandbox at this scale is evidence that the instrument exists and is used, not that it operates at national scale in the sense of a mass programme.
๐ฐ๐ฟKazakhstansince 2018still operatingbears on Firms through regulatory sandboxes
The Astana Financial Services Authority granted five FinTech Lab sandbox licences in 2022, saw the first participant graduate to a full licence that year, and had 18 licensed firms in its sandbox at the end of 2022.
5 firms, fintech lab licences granted (firms admitted to the sandbox) in 2022, 2022. 18 firms, licensed companies in the fintech lab at end of year, 2022. Astana Financial Services Authority, 2022 Annual Report, retrieved 2026-10-02.
The move, and what it needed
The Astana Financial Services Authority grants individual, time-limited FinTech Lab licences with requirements tailored to each firm, and graduates a firm to full authorisation when its test succeeds.
Preconditions
A financial centre with its own regulator and rulebook separate from the national one
Demand from foreign fintech and crypto firms for a regional base
Supervisory staff able to write firm-specific requirements
Limits: The regulator is reporting on itself in an annual report, and the counts stop at 2022. Admissions plus completions in 2022 come to six firms, which is well under one per million people. AFSA regulates only the Astana International Financial Centre, a special jurisdiction under English-style common law, so the sandbox sits outside Kazakhstan's ordinary financial regulation; the indicator sums every regulator in the country, and the National Bank and the Agency for Regulation and Development of Financial Markets are not counted here. Many participants are crypto exchanges drawn by the AIFC regime, and eight licences were withdrawn in the same year five were granted, so the counts measure churn as much as successful testing. The start year is inferred from 'fifth year' rather than stated.
๐น๐ณTunisiasince 2020delivered and closedbears on Firms through regulatory sandboxes
The Banque Centrale de Tunisie launched a regulatory sandbox in January 2020, admitted four fintech solutions to its first cohort in September 2020, ran tests with real customers and partner banks for 13 months, notified exit in September 2022, and drafted a circular on digital enrolment and remote bank account opening from the results (BCT Rapport annuel 2022).
The central bank set up a fintech committee, published an access guide, selected four solutions against its own strategic priorities, exempted them from specified rules for a bounded test with partner banks, and then wrote the tested remote-onboarding practice into a banking circular.
Preconditions
A central bank with rule-making power over the banks the fintechs had to test through
Partner banks willing to embed outside firms in a protected test environment
A strategic plan that named innovation as an objective, so staff time was allocated to it
Where else it has been used: The design follows the cohort model of the UK FCA and Bahrain's central bank; what did not travel was the cadence, since Tunisia ran one cohort and stopped.
Limits: The count is the regulator's own report of its own programme. The number does not say how many of the four completed, only that exit plans were validated. Only one regulator is covered; capital markets and insurance regulators are not. Four solutions in one cohort over 2020 to 2022 is about 0.33 admissions per million people (our computation against the World Bank's 2020 population of 11,974,057), and the record shows one cohort only.
๐ฒ๐ฟMozambiquesince 2018still operatingbears on Firms through regulatory sandboxes
Banco de Moรงambique admitted 11 participants, chosen from 19 applicants, to the sixth edition of its regulatory sandbox in March 2025, and at the same ceremony gave participation certificates to the five fintechs that had completed the fifth edition.
The central bank runs its sandbox in annual editions: it opens a call, selects a cohort, runs roughly nine months of preparation, testing and reporting, and certifies the firms that finish so they can apply for a licence.
Preconditions
A sandbox regulation issued by the central bank with fixed phases and exit rules
A financial inclusion office with staff to supervise each cohort
A licensing route for payment and credit services that a graduate can apply to
Where else it has been used: The cohort model with a public certificate ceremony is the format several southern African central banks adopted later; Mozambique is among the earliest in the region, with a first cohort in 2018.
Limits: The regulator reports on its own sandbox in a news item, not in a statistical series, and the page gives one cohort only: it does not give cumulative admissions or exits across the six editions, nor how many of the 11 went on to complete testing. 'Participants' can include incumbents as well as start-ups (the five graduates named include Vodafone M-Pesa, a large mobile money operator), so the count is not a count of new firms. Completing the sandbox makes a firm eligible to apply for a licence; the page does not say how many were licensed. The 2018 start is the establishment date in the World Bank's Appendix 3 database (May 2018, five firms in the sandbox at the time), not stated on the cited page. The sandbox page (bancomoc.mz, BM Regulatory Sandbox) shows a call for an eighth edition dated September 2026 and a seventh-edition invitation dated 1 October 2025 for testing from February 2026, so the regime was running at retrieval, but neither document carries a participant count. Against a population of about 34 million, 11 admissions is about 0.3 per million people.
๐ฒ๐บMauritiussince 2018still operatingbears on Firms through regulatory sandboxes
Mauritius's Economic Development Board, operating a regulatory sandbox licence regime for innovative financial projects since September 2018 by the World Bank's dating, had nine firms in its sandbox when the World Bank compiled its global database of sandboxes in 2020.
The investment promotion board issued time-bound sandbox licences to firms proposing financial activities for which Mauritian law had no framework, on the recommendation of a national committee, before the sector regulators took on their own sandbox authorisations.
Preconditions
A statute letting a non-regulator body license activities outside existing financial law
A cross-agency committee to vet applications on behalf of several regulators
Regulators willing later to absorb the function under their own statutes
Limits: The figure is a stock of firms in the sandbox at an unstated date in 2020, compiled by the World Bank from EDB sources, not an annual count of admissions and exits, and it sits in a wide table in the plain-text version of the report (the row lists 'Operational', 9, 'Regulatory Sandbox license', established Sep-18). EDB's own annual reports do not give admissions: the 2020/21 report says only that the EDB 'registered 8 new applications for fintech RSL' and that licences were issued when positively determined, and the 2021/22, 2022/23 and 2023/24 reports read today do not mention the licence at all. A licence under the EDB regime could cover activities with no legal framework at all, so some licensees used it as a market-entry route rather than a time-limited test. Since 2021 the sandbox function has also sat with the sector regulators: the Financial Services Commission issued its first Regulatory Sandbox Authorisation under section 14B of the Financial Services Act on 30 March 2022 and reported that entity exiting after successful testing in its 2022/23 annual report, and the Bank of Mauritius issued a Guideline for Regulatory Sandbox Authorisation in May 2024 with no participant count in its 2024 annual report or December 2024 Financial Stability Report. Status operating rests on those successor regimes, not on evidence that the EDB licence itself still admits firms.
๐บ๐ฌUgandasince 2021still operatingbears on Firms through regulatory sandboxes
The Bank of Uganda reviewed five applications to its regulatory sandbox in the year to June 2025, issued deficiency letters to all of them, and reported one firm, Rexial Payment Systems Limited, exiting the sandbox after testing and applying for a payment systems operator licence.
The central bank screens sandbox applications against its framework, sends deficiency letters rather than admitting weak proposals, supervises the experiments it has admitted, and steers firms that finish testing into its licensing regime.
Preconditions
A sandbox framework issued under the central bank's payment systems law
A licence category (payment systems operator) that a graduate can move into
Supervisory staff to monitor experiments for scalability and market suitability
Limits: The bank reports on itself, in a 188-page PDF (PDF page 59). The report gives no count of firms admitted in the year: that none of the five applicants was admitted is our reading of 'Deficiency letters were issued to all applicants', so the year's admissions may be zero, and the count of firms still testing is not given. One exit is the whole completion figure, and the report does not say whether the licence was granted. The 2021 start is the year the sandbox framework launched, not stated in this report. The same report lists 'review the regulatory sandbox framework' among next-year priorities. The Capital Markets Authority launched separate Regulatory Sandbox Rules in October 2025 (cmauganda.co.ug news, 8 October 2025) with no participant count, so the record covers only the central bank's sandbox. Against about 48 million people, one completion is a rate near zero per million.
๐ฟ๐ฒZambiasince 2021delivered and closedbears on Firms through regulatory sandboxes
Zambia's Securities and Exchange Commission admitted four innovations to the first cohort of its capital markets sandbox, with testing starting on 19 November 2021, and announced on 15 July 2024 that live testing had been completed and directed all four to suspend operations inside the sandbox pending evaluation and exit.
The securities regulator admitted a single cohort of capital markets innovations, let them take real investors under parameters and caps, then froze onboarding at the end of testing so it could evaluate each before deciding on commercial deployment.
Preconditions
A sandbox guideline issued under the Securities Act
Firms, including the national exchange, willing to test under caps
A licensing decision process after testing, which here had not been published
Limits: The regulator reports on itself in a one-page press release (PDF). The four are named innovations from four firms (Kukula Capital, Lupiya Circle, Lusaka Securities Exchange and Premier Credit Zambia), one of them the national stock exchange. The release says the framework was issued on 18 December 2022 but that testing began on 'Wednesday, 19th November 2021', which cannot both be right, and neither date matches its weekday (19 November 2021 was a Friday, 18 December 2022 a Sunday); the start year 2021 is the earliest date the release gives and is not confirmed by any other source read. Completion of testing is not a successful exit: the release says each innovator will or will not be permitted to deploy commercially after review, and no outcome statement was found on the SEC site; its list of sandbox participants page is empty. If the 2021 date is right, testing ran over two and a half years for one cohort of four, and no second cohort was found. The Bank of Zambia runs a separate sandbox and reports that 'one financial innovation exited the Sandbox after a successful test' in 2024 (Annual Report 2024, printed page 69, PDF page 86, https://www.boz.zm/sites/default/files/2026-01/Bank_of_Zambia_2024_Annual_Report.pdf); summed, the two regulators show about five firms through a sandbox over three years for 21 million people.
๐ฟ๐ผZimbabwesince 2026still operatingbears on Firms through regulatory sandboxes
The Securities and Exchange Commission of Zimbabwe published a list of seven innovative financial technology solutions approved for testing in its regulatory sandbox in July 2026, among them asset, securities and infrastructure tokenisation platforms, a crowdfunding platform and a blockchain capital raising platform.
The securities regulator approves named solutions into a timed, supervised sandbox under its own guidelines and publishes the list as a public notice so investors can tell a sandbox participant from a licensed firm.
Preconditions
Sandbox guidelines issued by the securities regulator
A fintech unit to monitor and evaluate the tests
Demand from tokenisation and crowdfunding firms with no licence category to apply under
Limits: The notice is a one-page PDF listing approvals; it calls itself an update but gives no earlier list, no admission date per firm, no testing period and no exits, so the figure is a stock of approved participants at July 2026, not an annual admissions count, and started is set to the notice year because no earlier admission was read at a source. Approval for testing is not testing: the notice warns that representations beyond testing are misleading, and some entries (Financial Securities Exchange, a licensed exchange) are incumbents. The Reserve Bank of Zimbabwe has run a separate fintech regulatory sandbox since March 2021, whose admissions are not counted here because no current count was read at an official source today. Seven approvals against about 16 million people is about 0.4 per million.
๐ฆ๐นAustriasince 2020still operatingbears on Firms through regulatory sandboxes
Austria's Financial Market Authority admitted eight FinTechs to its Regulatory Sandbox between its launch on 1 September 2020 and January 2023, five seeking authorisation as European crowdfunding service providers and three offering crypto-asset or securities services, and one participant had obtained a licence for a digital trading platform by the start of 2022.
Parliament amended the Financial Market Authority Act to create a sandbox at the FMA, with a federal grant for running it and an advisory board at the finance ministry that comments on each application's economic interest and readiness before the FMA decides.
Preconditions
A statutory basis and earmarked federal funding
An advisory board that can vet applications quickly
An EU rule change (here the Crowdfunding Regulation) that creates a cohort of firms needing supervised preparation
Limits: The count is the regulator's own and dates from January 2023; no later total was found, and an FMA notice on a later admission returned 403 to a script and was not read. Eight admissions in 28 months is below the five a year the federal contribution of EUR 500,000 a year was sized for (impact assessment, Vorblatt, https://www.ris.bka.gv.at/Dokumente/Begut/BEGUT_COO_2026_100_2_1632524/COO_2026_100_2_1632539.pdf, page 1), so the regime ran under its design volume. The sandbox grants no relief from regulatory requirements: the impact assessment says a core principle is that requirements are not lowered, so it is supervised testing with guidance rather than derogation. Five of the eight were crowdfunding platforms preparing for one EU regulation, which inflates the count with a one-off wave. Admission is not completion: the release names one licence obtained. The page was read through WebFetch because the FMA site returns 403 to curl.
๐ฉ๐ฐDenmarksince 2024still operatingbears on Firms through regulatory sandboxes
Denmark's data protection authority and digitalisation agency selected two projects from 23 applications for the first round of their regulatory sandbox for AI in July 2024, two from 18 applications in June 2025 and three more in July 2026, seven projects in all.
Two agencies run annual application rounds, pick a handful of AI projects whose questions others share, give each months of guided dialogue on data protection, and publish final reports so the answers reach firms that never applied.
Preconditions
A data protection authority willing to give advance guidance outside enforcement
A digitalisation agency to co-fund and co-select
Applicants willing to have their compliance questions published
Where else it has been used: The model follows Norway's Datatilsynet sandbox (nor-datatilsynet-sandkasse), running since 2020, with a smaller intake per round.
Limits: The total of seven is our sum of three announcements (https://www.datatilsynet.dk/presse-og-nyheder/nyhedsarkiv/2024/jul/to-ai-projekter-udvalgt-til-foerste-runde-af-den-regulatoriske-sandkasse and https://www.datatilsynet.dk/presse-og-nyheder/nyhedsarkiv/2025/jun/to-ai-projekter-udvalgt-til-anden-runde-af-den-regulatoriske-sandkasse); the third announcement gives no application count. Not every participant is a firm: round one included a project led by three municipalities with Systematic, round two the charity Boerns Vilkaar, and round three Region Hovedstaden with Kintella, so firms are fewer than seven. The sandbox offers guidance on data protection law, not derogations, and it is a GDPR sandbox rather than the AI Act sandbox the EU requires. Applications fell from 23 to 18 between the first two rounds. The Danish FSA's FT Lab, the financial sandbox since 2018, publishes admissions only as individual news items and no total, so it was not used; the national figure across regulators is therefore higher than seven by an unknown amount.
๐ณ๐ดNorwaysince 2020still operatingbears on Firms through regulatory sandboxes
Norway's data protection authority ran eight sandbox projects at the end of 2025, five of them jointly with the Financial Supervisory Authority on data sharing against financial crime, and completed two projects (Akershus University Hospital and Mobai's SALT biometric verification) during the year.
The data protection authority takes a few projects a year into a structured dialogue on how privacy law applies to a new use of data, then publishes a final report and podcast so the answer reaches the whole sector, and in 2025 opened a joint sandbox with the financial supervisor.
Preconditions
A ministry mandate and allocation letter that name the sandbox as a priority area
Regulators willing to give advance interpretations in public
Applicants whose question is general enough that a published answer helps others
Where else it has been used: Denmark's data protection authority opened a similar AI sandbox in 2024 (dnk-ai-sandkasse).
Limits: The authority reports on its own sandbox. It gives dialogue-based guidance and no exemption from law, so it is a controlled advisory regime rather than a derogation sandbox. Participants include public bodies (a hospital) and banks working as a group, so projects are not firms. The report is internally inconsistent on the joint financial sandbox: the chapter says five projects, the director's foreword in the PDF version (https://www.datatilsynet.no/contentassets/112abf21f900417f859c66c8f7d838de/datatilsynets-arsrapport-for-2025.pdf) says four were taken in; Finanstilsynet's own announcement of 2025 names five. The joint projects are counted here and are not added again from Finanstilsynet. Finanstilsynet's separate fintech sandbox admitted few firms (two of 12 applicants in the first pool in 2020, then single firms such as Abendum, R8Me and Qbig) and was not summed in. The sandbox is staffed by one full-time coordinator, per the same PDF, which limits its scale. The 2020 start is from the authority's earlier evaluation material, not read today.
๐ฑ๐นLithuaniasince 2018still operatingbears on Firms through regulatory sandboxes
The Bank of Lithuania opened its regulatory sandbox to applications on 15 October 2018 and, by the page's last update on 27 June 2025, listed one innovation ever tested in it: UAB Workpower's peer-to-peer insurance platform, authorised on 13 June 2019 and tested from 24 October 2019 to 23 October 2020, which led the bank to draft proportionate rules for P2P insurance.
The central bank offered live testing under supervision for financial innovations new to the Lithuanian market and used its one completed test to write proportionate guidelines for peer-to-peer insurance.
Preconditions
A regulator with authority to set test conditions under existing licences
Firms whose model does not fit existing licences, which in Lithuania were few because licensing was already fast
A route from test results to guidelines
Where else it has been used: The single test did travel into rules: the bank's 2021 news says it would issue P2P insurance guidelines, the sandbox-to-rule step that most sandboxes lack.
Limits: The live page (https://www.lb.lt/en/regulatory-sandbox) returned a bot challenge and 403 to scripts, so the archived copy was read; it carries the bank's own 'last update' date of 27 June 2025. The count is one firm in about seven years, so the regime exists but barely delivers; the page's heading 'Financial innovations currently tested' still shows a test that ended in 2020. The bank's better-known LBChain blockchain platform, which 11 fintechs from 8 countries used during development to May 2020 (https://www.lb.lt/en/lbchain, archived), was a technical test environment without regulatory relief and is not counted. The bank also runs a newer sustainability-report test under the GFIN initiative (news of 16 July 2021) that is not listed as a firm test on the page. Lithuania's high fintech licensing volume runs through ordinary licensing, not this sandbox.
๐ธ๐ฐSlovakiasince 2022still operatingbears on Firms through regulatory sandboxes
Slovakia's central bank opened its regulatory sandbox on 1 January 2022 and lists one former participant, Crowdberry Investment Platform, which entered on 22 November 2023, tested an electronic bulletin board for secondary trading of crowdfunded securities with selected clients from 15 December 2024 to 18 April 2025, and exited on 15 May 2025.
The central bank admits firms year-round into a preparation phase of up to six months of consultations and a testing phase of up to six months on real clients under a test plan, ending with a final report and meeting.
Preconditions
A published methodology for entry criteria
Staff in a fintech department to run long consultations
Firms with a product that needs real-client testing, which in a small market are few
Limits: The page lists former participants only and gives no count of current participants or applications, so admissions in progress are unknown; one completion in nearly four years of operation is the figure the page supports. The sandbox waives no regulatory requirement: the page says requirements cannot be waived, so it is supervised testing with consultation. The bank states that its innovation hub handles far more interactions than the sandbox, which reflects the design rather than a measured failure. The regulator reports on its own programme.
๐บ๐ฆUkrainesince 2025still operatingbears on Firms through regulatory sandboxes
Ukraine's Ministry of Digital Transformation and the Ukrainian Startup Fund opened a sandbox for AI and blockchain products in 2025 and, in its first year, received more than 80 applications and admitted 15 companies to a first cohort that completed testing.
The digital ministry, with a state startup fund, invites AI and blockchain firms to submit products for an expert audit across law, technology and business, and feeds what it learns into drafting AI legislation aligned with the EU AI Act.
Preconditions
A ministry with a standing AI policy unit and a lawmaking agenda to feed
A state startup fund able to recruit and pay outside experts under NDA
A pipeline of domestic AI and blockchain firms seeking EU market access
Limits: The ministry reports on its own programme in a news item, not a statistics page, and 80 is a floor ('more than 80'). The sandbox grants no statutory derogation that this check found: what participants receive is a free legal, technical and business audit by outside experts with recommendations, and the ministry's statement that firms test 'without fines or legal risks' is not backed by a legal instrument read today, so this is closer to an advisory sandbox than to a regime that relaxes rules, and it sits at the weak edge of the D152 construct. 'Completed testing' means completed the audit cycle; the ministry's page of 21 October 2025 names Artellence as the first company to finish it. Higher figures in the press (over 100 applications, 25 companies under review) were not found at an official publisher and are not used. The National Bank of Ukraine runs a statutory regulatory platform under the payment services and financial services laws, open only to firms it already licenses, but its page (promo.bank.gov.ua/sandbox) publishes no participant count, so the figure covers one programme and not every regulator.
๐ท๐บRussiasince 2018still operatingbears on Firms through regulatory sandboxes
The Bank of Russia decided in 2025 to create the legal conditions for introducing 4 digital financial services piloted in its regulatory sandbox, out of 10 pilot applications it considered that year in identification, artificial intelligence, insurance, distributed ledgers and other areas.
The central bank runs an isolated environment in which any organisation can pilot a financial service that existing law would not allow, and when a pilot succeeds it approves a roadmap of legal changes so the service can be introduced.
Preconditions
A central bank with rule-making power over most of the financial market
An expert review process for applications
A legislative route to turn a successful pilot into law, which the experimental legal regimes law of 2020 added
Limits: The central bank reports on its own sandbox, in a 351-page PDF of 57 MB, on PDF page 213 (printed page 211). The unit is applications, not firms: the report does not say how many firms stand behind the 10 applications or the 4 decisions, and one firm can file several. The 4 are read as completions because the Bank's sandbox page (https://www.cbr.ru/fintech/regulatory_sandbox/) says it approves a roadmap for legal conditions when piloting is completed successfully; the annual report itself does not say the pilots ended in 2025, and the 10 considered are not admissions, since an application can be reviewed without a pilot running. The sandbox launched in April 2018 (the Bank's sandbox page) and the Bank said in August 2023 that 86 services had been submitted since launch, so 10 in 2025 is a modest flow, about 0.07 per million people, and 4 completions about 0.03. Experimental legal regimes under the 2020 federal law on digital innovation, which the Ministry of Economic Development registers for every sector, are a second channel the column should sum; the ministry's register did not open from here (connection refused) and its last archived copy is from 2022, so they are not counted. Participants are mostly large incumbent banks and infrastructure operators, so this is not a startup programme.
๐ฏ๐ดJordansince 2024still operatingbears on Firms through regulatory sandboxes
The Central Bank of Jordan approved three companies on 31 October 2024, out of nine applications to the first cohort of its JoRegBox regulatory sandbox launched in February 2024, to begin testing digital financial products with real customers, among them the escrow platforms E-Damana and Watheq.
The central bank admits fintech firms in cohorts to test mature products with real customers under its supervision, and links the sandbox to the national financial inclusion strategy and the economic modernisation vision.
Preconditions
A central bank sandbox framework with published eligibility and exit rules
Partner banks willing to hold escrow and settlement accounts for test firms
A government programme that makes fintech a named priority
Limits: The live cbj.gov.jo page refused the connection, so the number was read on a Wayback Machine copy of the central bank's own release. Three is the admission decision for one cohort, the first under the JoRegBox name, and the release gives no count of firms completing the sandbox; the bank's annual report, which would carry later cohorts and exits, was not located today. The central bank had an earlier sandbox framework, which the World Bank's 2020 survey lists as operational from April 2018; the release treats JoRegBox as a new launch on 18 February 2024, and whether firms passed through the earlier framework is not shown here. The governor has been quoted in the press giving a cumulative count of tested products, which this record does not use because no official page carrying it was opened. Three admissions against about 11 million people is about 0.3 per million. The release frames the cohort as part of the government's economic modernisation programme, so the timing is partly political.
๐ฐ๐ฌKyrgyzstansince 2020still operatingbears on Firms through regulatory sandboxes
The National Bank of the Kyrgyz Republic approved six special regulatory regimes between October 2020 and August 2026, letting a microcredit company, three payment and e-money firms and two banks test operations outside its standing rules, judged testing successful and closed for four of them between April 2022 and February 2025, and ended one early with the licence revoked in March 2022.
The central bank grants a firm a time-limited special regime, licensed and entered in a public register with start and end dates, to test a named banking or payment operation, then records whether testing succeeded and amends its rules where it did.
Preconditions
A banking law provision allowing the national bank to set special regulatory regimes
A public register kept up to date row by row
Payment and e-money firms with products that standing rules did not allow
Where else it has been used: After the regime run by Regional Payment Systems, the board amended its regulations in May 2024 to remove obstacles to card acquiring by payment system operators, which is the route from test to rule.
Limits: Both counts are ours, made by reading the rows of the register page (dated 2 September 2026): it lists six regimes with approval, start and end dates, the words 'testing judged successful (regime completed)' against four, 'ended early, licence revoked' against the Bakai Bank regime (September 2021 to March 2022), and a sixth, Eldik Bank, approved on 7 August 2026 and testing until August 2027; the page itself states no total. Some regimes list partner institutions (the Interbank Processing Centre) as co-participants, so the count is of regimes, not of every firm involved. Six regimes in six years, about one a year in a country of about seven million people, is a thin flow, and the regime appears only once a firm asks for it rather than through open cohorts. The register says the national bank amended its regulations in May 2024 after the Regional Payment Systems regime, but it does not show whether the other completions led to rule changes. The register is a central bank page only; other Kyrgyz regulators were not checked.
๐ต๐ฐPakistansince 2020still operatingbears on Firms through regulatory sandboxes
Pakistan's Securities and Exchange Commission approved 31 business models for testing in its Regulatory Sandbox across four annual cohorts from 2020 to 2023, out of 125 fintech applicants, with intake rising from six selected in the first cohort to 11 in the fourth.
The securities regulator issued sandbox guidelines under the Companies Act and ran annual cohorts in which fintech, insurtech and capital-market applicants were screened and the selected ones tested for six to nine months under agreed limits, then switched to year-round applications.
Preconditions
A statutory power to issue sandbox guidelines, here section 458A of the Companies Act 2017 as cited in the reissued guidelines
A pipeline of fintech applicants large enough to select from
Regulator staff to evaluate applications and supervise tests
Limits: The regulator reports on its own sandbox. The figures are in a PDF on PDF page 82 (printed page 81), with a cohort chart giving applicants and selections per year (33 and 6, 38 and 6, 28 and 8, 26 and 11); the live secp.gov.pk site returns a Cloudflare challenge to scripts, so the report was read from a Wayback Machine copy of the SECP download link. 'Approved for testing' counts admissions, not completions: the report does not say how many business models finished testing or went to market. The cohort figures sum to 31, matching the stated total. After 2023 the SECP moved to an always-open format, and the same page says it had then received three applications with one testing, a much smaller intake than cohort IV, but over a shorter and unstated period, so it is not read as a decline. The State Bank of Pakistan runs a separate sandbox whose admissions are not counted here, so the figure understates the country total the construct asks for.
๐ฏ๐ฒJamaicasince 2020still operatingbears on Firms through regulatory sandboxes
The Bank of Jamaica's FinTech Regulatory Sandbox, effective from March 2020, had received 50 applications and approved 29 of them for testing by the end of 2025, including four new applications from three entities in 2025.
The central bank opened a sandbox for deposit-takers, cambios, remittance providers, credit unions and locally incorporated fintechs partnered with a bank, admitting products to a controlled live test before authorisation.
Preconditions
Guidelines and an application and fit-and-proper process run by the payments oversight function
Licensed deposit-takers willing to partner with fintechs
Fintech and payment firms with wallet and card products ready to test
Limits: The central bank reports on its own sandbox. The figures are in a 33 MB PDF served at that address, on PDF page 92 (printed page 66). They count applications, not firms: one entity can file several, as the three entities filing four applications in 2025 show, and 29 is approvals since 2020, not the number testing at year end or the number that completed and launched. The report gives 46 applications at end-2024, so only four arrived in 2025, but it gives no yearly approval figures. The approvals are described as mobile wallets, debit cards and prepaid card offerings; the same section of the report covers the bank's JAM-DEX digital currency roll-out, but it does not say whether any sandbox approval relates to it. The Financial Services Commission's activity is not included.
๐ฒ๐ณMongoliasince 2021still operatingbears on Firms through regulatory sandboxes
Mongolia's Financial Regulatory Commission admitted three products and services to its sandbox for one year in the fourth quarter of 2022 and released one, the MonPay digital bond, on 30 December 2022 after a year of testing, leaving four under test.
The non-bank financial regulator adopted a sandbox regulation in 2021 and a sandbox council that admits fintech products for a one-year test and issues a recommendation to release each one.
Preconditions
A regulation on the sandbox environment adopted by the commission
A council to decide admission and exit
Domestic fintech and virtual asset firms seeking a route to licensing
Limits: The regulator reports on itself, in Mongolian, in a PDF on PDF page 29 (printed page 30). It counts products and services, not firms. The same page breaks down requests submitted to the sandbox by type (6, 2, 2, 1, 1, 1, 1, 1), which sum to 15 by my addition, but it does not say over what period they were received or how many were refused. The admission figure covers one quarter, and the report gives no full-year admissions for 2021 or 2022, when MonPay's test began on 1 November 2021. Only one year was read: later FRC reports were not found today, so the record says nothing about whether the sandbox kept admitting products after 2022. The products under test were a repo platform, a peer-to-peer lender, an insurtech service and a crowdfunding platform, and virtual asset providers made up the largest group of requests. The Bank of Mongolia's activity, if any, is not included.
๐ณ๐ฆNamibiasince 2023still operatingbears on Firms through regulatory sandboxes
Namibia's non-bank financial regulator NAMFISA launched its Regulatory Sandbox in May 2023, received 10 applications for its first, microlending cohort, and had admitted three participants (ReferredBy Financial Solutions, Ergo Analytics trading as Verime, and the microlender Express Credit) with a fourth application under review by its 2025 annual report.
The non-bank supervisor opens industry-specific cohorts, admits licensed firms and start-ups to test a product under supervision where current legislation would not otherwise allow it, and feeds what it learns into regulation, with a FinTech Square event as its engagement channel.
Preconditions
A supervisor with power to grant conditions or exemptions to the entities it licenses
A cohort design narrow enough for a small supervisory team to monitor
A pipeline of local fintech applicants, here built partly through the FinTech Square events
Where else it has been used: NAMFISA plans to re-establish the sandbox under the Financial Institutions and Markets Act (FIMA) once it commences; the Bank of Namibia runs a parallel framework for banking and payments with no published participant count.
Limits: NAMFISA is reporting on its own programme. Both figures are inside PDFs: the three admitted participants and the fourth applicant are on printed pages 87 to 88 of the Annual Report 2025 (PDF page 100), with the same count in the risk section ('three active participants and a fourth in an advanced stage'); the 10 applications are in the Annual Report 2024 (printed page 81, PDF page 48, https://www.namfisa.com.na/download/annual-report-2024/), whose chart of successful against unsuccessful applications prints labels that cannot be read reliably as counts, so the record does not use it. The reporting year ends 31 March 2025, but the text calls the third participant 'recently admitted', so the asOf is the report year, not a day. The same report's division summary for the Market Conduct and Operations function says that in the first cohort two participants were actively regulated and monitored during 2024/25, which fits the third, Express Credit, being admitted late in the year and still in mobilisation. The count is cumulative admissions over about two years, not admissions in one year, and no firm is reported as having completed or exited the sandbox; the 2024 report notes one selected entity whose start was delayed for market readiness. Three firms in a population of about three million is small, and it covers only the non-bank sector NAMFISA supervises: the Bank of Namibia's separate FinTech innovation framework publishes no participant count and is not included. A second cohort for crowdfunding and peer-to-peer lending closed applications on 23 February 2026 with no admissions published yet.
๐ฆ๐ฟAzerbaijansince 2024still operatingbears on Firms through regulatory sandboxes
The Central Bank of the Republic of Azerbaijan began testing its first regulatory sandbox participant in December 2024 and admitted three more whose testing began in 2026, out of 14 applicants.
The central bank wrote rules on a special regulation regime that exempt participants from licensing, opened cohort application windows, and admitted a handful of banks and crypto firms to test products with capped users under its supervision.
Preconditions
A central bank that is also the integrated financial regulator, so one body can grant the exemption
Demand for products the existing licence categories do not cover, here Islamic finance and crypto exchange
Limits: The year count of three is computed from testing start dates in the registry; the central bank publishes only undated cumulative counters (14 applicants, 4 participants), so the series has no official annual figure and earlier years cannot be reconstructed if entries are removed. The registry's completed projects section shows no entries although the first test was scheduled to end in February 2026 and is still listed as current, so firms completing the regime, half of what the indicator asks for, cannot be counted. Only one regulator is covered. A sandbox of four firms is not a national flagship in the sense of a programme policymakers name first; it is filed because it evidences this gap's construct literally.
Brazil ended four decades of high inflation with a staged currency reform, taking annual consumer price inflation from 2,075.9 percent in 1994 to 3.2 percent in 1998.
3.2 % per year, annual consumer price inflation, 1998. World Bank, from IBGE, retrieved 2026-08-26.
The move, and what it needed
A stable unit of account was introduced alongside the failing currency, prices were re-anchored to it while it was still only a reference, and it became legal tender only once the anchoring had happened.
Preconditions
A technical team with authority to sequence the reform
Enough reserves to defend the new currency
Public exhaustion with inflation deep enough to accept the transition
Where else it has been used: The staged unit-of-account move is the transferable idea and it has been studied everywhere. The exchange rate anchor that held it in place broke in 1999, which is the part not to copy.
Limits: Disinflation was held in place with an overvalued exchange rate and high interest rates, and that arrangement broke in the 1999 devaluation and again in the 2002 confidence crisis. One macroeconomic outcome is not a general measure of how fast Brazilian institutions respond to new conditions. The design and the sequencing are the capability here, and no number in this record captures them.
๐ง๐ทBrazilsince 1975still operatingbears on Institutional responsiveness
Brazil answered the 1973 oil shock by mandating ethanol blending and building a national fuel alcohol industry, taking production from 580 thousand cubic metres in 1975 to 38,199 in 2025, and from 2003 flex-fuel engines moved the choice to the driver.
A blending mandate created guaranteed demand, subsidised credit built the mills against it, and flex-fuel engines two decades later moved the choice from the state to the driver.
Preconditions
A crop and land base able to supply at scale
A state able to hold a mandate for decades
Domestic vehicle manufacturing that can be made to adapt
Where else it has been used: The mandate-then-credit sequence is transferable and has been tried widely. The flex-fuel step is what made it durable and it required the car industry to move.
Limits: The programme was heavily subsidised in the 1980s and nearly collapsed when oil prices fell in the 1990s, so the line is not a story of steady success. Production volume says nothing about land use, cane labour conditions, or how the emissions accounting works out. What this evidences is a state changing an entire fuel system in response to an external shock, which is the construct the indicator asks for.
๐ฆ๐ทArgentinasince 1991dismantledbears on Institutional responsiveness
Argentina ended hyperinflation by fixing the peso to the dollar by law in April 1991, taking annual inflation from 4,923 percent in 1989 to 3.9 percent in 1994, and the regime held until it collapsed in the 2001-2002 crisis.
Congress fixed the exchange rate one-to-one to the dollar by statute and barred the central bank from printing unbacked pesos, which converted a monetary promise nobody believed into a law that was costly to break.
Preconditions
Dollar reserves large enough to back the monetary base
A legislature willing to bind itself and its successors
Prices and contracts already partly dollarised, so the anchor was credible on day one
Where else it has been used: Currency boards were copied widely in the 1990s, from Estonia to Bulgaria, and they worked as stabilisers. Argentina is the case every copy now studies for the exit problem: the law had no honest way out.
Limits: The IMF and World Bank series for Argentina do not reach these years, so the numbers come from a teaching case that republishes the official table, and the 1989 figure is a December-to-December change where other publications give 3,079 percent as the annual average. Ending inflation is what the indicator asks for, but the fixed parity that delivered the stabilisation also removed the tools that could have absorbed the shocks of 1999-2001, so the case evidences a response that consumed its own durability. The collapse later cost Argentina its statistical credibility too, which is why the official series is hard to cite today.
๐ฉ๐ชGermanysince 2008delivered and closedbears on Institutional responsiveness
Germany answered the 2008-2009 collapse in manufacturing orders by paying firms to cut hours instead of jobs, and short-time work covered 1.44 million workers at the May 2009 peak, about 5 percent of insured employment, while unemployment barely rose.
The employment agency reimburses firms most of the net wage for hours not worked when demand collapses, converting the existing unemployment-insurance fund into a payroll bridge, and the 2009 government lengthened the entitlement and cut the paperwork.
Preconditions
A standing legal instrument and fund, so scaling needed a regulation rather than a new law
Works councils and employers used to negotiating hours
A shock read as temporary, since the bridge only pays off if demand returns
Where else it has been used: Short-time work schemes spread across the OECD after 2009 and nearly everywhere in 2020. The lesson that travelled worst is fiscal: the bridge is cheap only when the crisis is short.
Limits: The count says how many workers were covered, not what the scheme cost, how much of the covered work would have survived anyway, or whether it slowed reallocation to better firms, which are the standard criticisms. The instrument long predates 2008; what the episode evidences is the speed at which an existing legal instrument was expanded inside one year. The same instrument covered over six million people in 2020, so the 2009 number is not its ceiling.
๐น๐ทTurkeysince 2001operating below its peakbears on Institutional responsiveness
Turkey answered its 2001 banking collapse with central bank independence, bank restructuring and fiscal rule, taking inflation from 54.4 percent in 2001 to 8.6 percent in 2004, and after 2018 the same framework was overridden until inflation reached 72.3 percent in 2022.
A crisis government handed monetary policy to an independent central bank by law, recapitalised and closed broken banks through an autonomous agency, and held a primary surplus, which together made the anti-inflation promise credible for fifteen years.
Preconditions
A crisis deep enough to make the political cost of reform payable
External anchors, the IMF programme and EU accession prospect, that rewarded holding the framework
A technocratic team with authority to execute across ministries
Where else it has been used: The package was the standard post-crisis prescription and worked as designed. What the case adds is the second half: formal independence survived on paper while governors were replaced until policy followed the executive, a path other countries have since repeated.
Limits: Inflation is an outcome with many causes: the 2001-2004 fall also rode an IMF programme and a global disinflation, and the post-2018 rise includes a currency crisis and imported energy prices, so neither number isolates the institutions. The record runs both ways by design: the same state that built a working stabilisation framework later demonstrated that the framework had no defence against its own executive. The institutions were not abolished, which is why the status is eroded rather than dismantled.
๐ง๐ทBrazilsince 1972delivered and closedbears on Institutional responsiveness
Created in 1972, Telebras coordinated Brazil's state telecom system until its 1998 privatisation, when 54 concessionaires were split out and the state moved from operating the network to regulating and universalising service.
The state first assembled a fragmented telecom sector into one planning and investment system, then dismantled that holding structure when a new regulatory and private operating model was judged more suitable for expansion.
Preconditions
Authority to consolidate regional operators under one national plan
A public investment and research base large enough to build the network
A legal and regulatory framework able to transfer operating assets without abandoning the public objective
Limits: The number of controlled companies shows the structure of the old system, not whether privatisation delivered affordable, reliable service everywhere. The transition also left a continuing Telebras with a different role, so the record concerns a model that ended rather than a company that disappeared.
๐ช๐ชEstoniasince 1990still operatingbears on Institutional responsiveness
The Estonian Defence League was re-established in February 1990, before the country regained independence, and operates as a voluntary national defence organisation under the Ministry of Defence with territorial units and its own training structure.
A state rebuilt a territorial volunteer defence organisation before independence was restored, preserving a civic channel for mobilisation alongside the formal defence forces.
Preconditions
A population willing to sustain a voluntary defence organisation
A legal route from civic organisation to public defence responsibility
A ministry able to train and coordinate territorial units
Limits: The re-establishment date and organisational description show institutional intent and continuity, not readiness, effectiveness or public support. The source describes the Defence League itself, so it does not independently test its performance.
๐ฐ๐ทSouth Koreasince 2009still operatingbears on Institutional responsiveness
Korea established the Korea Internet and Security Agency in July 2009 by integrating three predecessor organisations, giving the country one public body for internet development, security and international information technology cooperation.
Three specialised organisations were combined into one agency so that internet development, security and international cooperation could be handled through a shared public capability.
Preconditions
A government willing to merge overlapping specialist bodies
Technical staff and institutional knowledge that can survive the merger
A mandate broad enough to connect infrastructure, security and cooperation
Limits: The integration date and mandate show a public response to a changing digital environment, not the agency's effectiveness against cyber threats or its performance across every programme. The agency is the source for its own institutional history.
๐ฉ๐ชGermanysince 1955dismantledbears on Disaster preparedness and recovery
West Germany built a federal warning service after 1955; after the Cold War the federally owned service was dissolved, its ten warning offices were abandoned and the nationwide siren alarm network was dismantled or transferred to municipalities, ending a system that had relied on about 1,700 volunteer helpers.
After the Cold War, the federal government treated a nationally coordinated warning service as obsolete, dissolved its dedicated organisation and shifted much of the remaining equipment and responsibility to local authorities without a ready replacement with the same audible reach.
Preconditions
A federal system able to finance and coordinate warning infrastructure across municipalities
Local authorities willing to accept ownership and operating costs after transfer
A replacement communications plan that works when electricity or mobile networks fail
Where else it has been used: The post-Cold War drawdown was common across Europe. Germany's later investment in warning apps, cell broadcast and renewed siren funding shows the cost of dismantling a capability before the next hazard makes its value visible.
Limits: The number of offices and volunteers describes the architecture that was removed, not how many people the sirens actually reached or how well the replacement warning channels worked. Responsibilities moved to Lรคnder and municipalities rather than disappearing from every locality, and some sirens were retained for fire services. The Bundestag report is an official retrospective, not an independent test of warning performance.
๐ณ๐ฑNetherlandssince 2006delivered and closedbears on Disaster preparedness and recovery
The Netherlands ran Room for the River from 2006 to 2019, delivering 39 measures at 34 locations along the IJssel, Waal, Nederrijn and Lek for about four million residents at a total cost of โฌ2.3 billion.
After the near-floods of 1993 and 1995, central government and regional authorities moved dikes inland, lowered floodplains and dug side channels so high water could spread safely, using one programme to coordinate Rijkswaterstaat, water boards, provinces, municipalities, landowners and residents.
Preconditions
A national water authority with technical and land-management capacity
Water boards and local governments able to share decisions and long-term maintenance
Multi-year funding and a credible process for acquiring land and compensating residents
Where else it has been used: The approach has been studied and adapted in other deltas. What travels is the portfolio logicโmany local measures under one hydraulic modelโnot a single signature project, and it depends on institutions that can keep negotiating for fifteen years.
Limits: The measure and location counts show delivery, not the ex post reduction in flood risk, maintenance, cost-effectiveness or local equity. The official programme account does not independently test hydraulic performance, and the four-million figure is the population in the programme area rather than a safety outcome.
๐ณ๐ฑNetherlandssince 2004operating below its peakbears on Institutional responsiveness
The Netherlands' tax administration put tens of thousands of childcare-benefits parents into debt between 2004 and 2019; by February 2026 the recovery operation had recognized more than 43,000 parents as harmed among about 69,000 applicants and guaranteed each recognized parent at least โฌ30,000.
A rigid fraud-control regime treated small administrative errors as evidence of wrongdoing and allowed repayment demands to compound across years; after public and parliamentary scrutiny, the state created a separate recovery organisation, integral assessments, debt relief and a standard โฌ30,000 first payment.
Preconditions
Benefits data and automated risk rules powerful enough to stop payments at scale
Weak internal challenge and fragmented appeal routes that let an error persist across agencies
A parliamentary and judicial route capable of forcing an apology, new rules and a large-scale redress operation
Where else it has been used: The transferable lesson is institutional: auditability, human review and a single route for appeal matter as much as detection. A recovery fund can pay money after failure, but it cannot recreate the years of security that a responsive administration would have protected.
Limits: The application and recognition counts describe the scale of harm and the administrative response, not whether compensation restores lost homes, jobs, health or trust. The โฌ30,000 floor is a legal payment rule rather than a measure of full loss, and the government reports the recovery operation about itself. The case therefore shows a serious failure and a continuing repair effort, not a settled measure of responsiveness.
๐ณ๐ฑNetherlandssince 2013still operatingbears on Institutional responsiveness
On 1 January 2013 the Netherlands merged 25 autonomous regional police corps and the national KLPD into one organisation with ten regional units, a national unit and a shared Police Services Centre, centralising specialist capacity while retaining local teams.
The government replaced autonomous regional corps with one national command and shared services, while keeping districts and frontline teams close to municipalities; specialist intelligence, organised-crime and technology functions could therefore be coordinated across the country without abolishing local policing.
Preconditions
A legal mandate strong enough to overcome the autonomy of existing regional forces
A common information, training and human-resources system for the merged organisation
Local commanders and municipalities able to preserve neighbourhood accountability inside a national structure
Where else it has been used: National police reforms are common where crime and digital threats cross borders. The Dutch design shows the trade-off clearly: centralise scarce expertise and procurement, but keep frontline teams and local authority visible enough that nationalisation does not become withdrawal.
Limits: The number of units shows an organisational response to cross-border crime and technology, not faster emergency response, better investigations or lower crime. A merger can also create transition costs and distance from local accountability that this architecture count cannot see. The police account describes the reform and does not independently evaluate its outcomes.
๐ณ๐ฑNetherlandssince 2014still operatingbears on Disaster preparedness and recovery
The Netherlands tests NL-Alert twice a year and, in the 1 December 2025 national test, 92 percent of residents aged 12 and over said they received the message on their mobile phone; 90 percent received it only on that channel and 2 percent received it on mobile plus public digital signs.
The government uses cell broadcast through every telecom provider, adds public transport and advertising displays, tests the channel twice a year and measures reach by age and safety region so a warning can carry both a location-specific message and an action to take.
Preconditions
Telecom operators able to broadcast one authorised message to a defined geographic cell area
A legal and operational chain from national ownership to local safety regions that decide when to alert
Public channels and campaigns that can reach people who are outside, at a border or unable to rely on a smartphone
Where else it has been used: Cell-broadcast warning systems are spreading internationally. The Dutch lesson is to measure reach repeatedly and keep non-phone channels in the design; a technically elegant alert that leaves the same people unreachable is not a resilient warning system.
Limits: The reach figures are self-reported survey responses to a test message, not observed receipt during a live crisis, comprehension of the instruction or compliance with it. The survey covers residents aged 12 and over and still found 8 percent not reached; people without a working phone, people abroad and older residents are not interchangeable gaps. The report also notes a method break in the historical series.
๐ณ๐ฑNetherlandssince 2005still operatingbears on Institutional responsiveness
The Netherlands built delegated authority into its public-service login so a family member, carer or adviser can act for someone else without taking their password; in 2024 DigiD Machtigen handled 15,563,897 queries from 798 connected services at 99.84 percent availability against a 99.5 percent norm.
15,563,897 queries, digid machtigen queries by connected services, 2024. 798 services, connected services, 2024. Logius, 2024 service figures, retrieved 2026-08-31.
The move, and what it needed
The state separated authentication from authority: a person grants a named representative access to one service for a defined period, while the representative uses their own credential and the system can record or revoke the mandate.
Preconditions
A trusted identity rail for both the person granting access and the representative
Service-specific permissions and expiry rules rather than one blanket delegation
Postal, telephone and assisted-digital routes for people who cannot create a mandate online
Where else it has been used: Delegated digital access is increasingly important as governments move services online. The Dutch lesson is to design representation as core infrastructure, not as an informal workaround that forces people to surrender credentials.
Limits: Queries, connected services and uptime show a maintained delegation rail, not whether the right person receives help, whether abuse is prevented or whether delegated transactions resolve well. A service can be technically available while a person remains unable to nominate a representative or understand the scope of access. The operator's counts also do not show how many residents still rely on paper or informal password sharing.
๐ณ๐ฑNetherlandssince 2004dismantledbears on Institutional responsiveness
The Netherlands ended DigiNotar's PKIoverheid service after the 2011 breach generated at least 531 forged certificates; 333 could be revoked and the government tightened certificate requirements, role division and supervision after the incident.
531 certificates, forged certificates generated during the breach, 2011. 333 certificates, forged certificates that could be revoked, 2011. Dutch Safety Board, DigiNotar incident investigation, retrieved 2026-08-31.
The move, and what it needed
When a private certificate provider was compromised, the state took operational control, moved users to other providers and rewrote the PKIoverheid requirements so Logius and the regulator had clearer roles and stricter oversight. The response improvised around the risk that revoking every certificate at once could interrupt essential government data flows.
Preconditions
A government-managed trust framework with contractual authority to remove a compromised provider
Alternative certificate providers able to absorb public-sector services during a forced migration
Incident investigation and supervisory institutions that can turn a breach into changed requirements rather than a one-off fix
Where else it has been used: Certificate-authority failures elsewhere led to similar moves toward tighter root-store governance and incident-response playbooks. The Dutch lesson is that a trust system needs a rehearsed provider-failure path; audits and a nominal single point of trust are not a recovery plan.
Limits: The certificate counts measure the scale of a compromise and the clean-up problem, not the number of people harmed or the security of the replacement trust chain. The Safety Board found that an unknown number of forged certificates entered circulation and that one was known to have been used in Iran; it did not establish direct financial or privacy damage in the Netherlands. Ending one provider and tightening oversight shows a response to failure, not proof that digital-government security became permanently robust.
๐ณ๐ฑNetherlandssince 2015dismantledbears on Institutional responsiveness
The Netherlands operated the Programmatic Approach to Nitrogen (PAS) from July 2015 until the Council of State invalidated its legal basis in May 2019; by the end of 2023, 2,557 affected PAS reporters had sought permits and only 115 had a recorded solution.
The state tied permission for new nitrogen-emitting activity to promised future reductions and ecological recovery; when the Council of State found that the programme did not meet EU nature-law requirements, the legal shortcut collapsed and ministries and provinces had to rebuild permitting and legalisation around demonstrable emissions space.
Preconditions
A judicial review route able to test a national programme against binding European environmental law
A monitoring and modelling system such as AERIUS that can allocate and revisit emissions space
Cross-government authority and budget to redesign permits while protecting affected operators and restoring ecological conditions
Where else it has been used: The case travels as a warning about policy built on discounted future action: adaptive capacity includes the ability to abandon a legally unsafe shortcut and support those stranded by it, not only the ability to announce a new target.
Limits: The figures show the administrative aftermath of a court-invalidated permitting framework, not the total nitrogen trend, the condition of every Natura 2000 site or the full distribution of responsibility for the policy. The Court of Audit counts requests and solutions reported through the government process; a recorded solution can mean a permit is unnecessary as well as that one was granted. The reversal demonstrates a failure to make future mitigation legally certain, while the later legalisation effort remains a separate and unfinished response.
๐จ๐ฑChilesince 2010delivered and closedbears on Disaster preparedness and recovery
After the 2010 earthquake and tsunami, Chile's reconstruction programme repaired or replaced more than 223,000 homes across seven regions; MINVU reported 99.93% progress by February 2020 and changed protocols, seismic standards and risk-mitigation requirements.
223,000 homes repaired or replaced, homes repaired or replaced in 27f reconstruction, 2020. 99.93 % of reconstruction programme, reported progress of 27f reconstruction programme, 2020. Ministry of Housing and Urbanism (MINVU), 27F reconstruction report, retrieved 2026-08-31.
The move, and what it needed
Chile coordinated a multi-region housing and urban recovery programme through national standards, subsidies and local delivery, then used the reconstruction experience to revise emergency protocols, seismic requirements and mitigation practice.
Preconditions
A national housing ministry with the budget and authority to coordinate across regions and municipalities
Construction and engineering standards suited to earthquake and tsunami risk
A damage register, delivery pipeline and feedback from the recovery effort that can be converted into changed rules
Where else it has been used: Post-disaster reconstruction programmes often travel as templates, but the Chilean lesson is the feedback loop: speed and volume matter, while durable preparedness depends on turning delivery failures and field experience into enforceable standards before the next shock.
Limits: The housing and progress figures are government-reported and do not independently verify build quality, equitable access, long-term habitability or whether risk was reduced for every community. A programme marked 99.93% complete is not evidence that every livelihood, public facility or ecosystem recovered, and the final homes were still in execution at the time of the 2020 report.
๐จ๐ดColombiasince 2022still operatingbears on Institutional responsiveness
In 2024 Colombia's Registradurรญa deployed version 2.0 of its digital ID app and sent 2,642,775 polycarbonate digital identity cards to offices nationwide while continuing a pilot to refine digital services.
The national identity authority coupled a secure physical credential with a mobile app, biometric records and staged pilots, using real election and citizen-service needs to test new functions before widening the digital service.
Preconditions
A trusted civil registry and biometric infrastructure that can issue and verify one identity
A public operator able to maintain both a physical credential and a mobile service during transition
Pilot users, service partners and support channels that can expose problems before a national feature becomes mandatory
Where else it has been used: Digital identity upgrades travel when a country has a reliable foundational register, but the Colombian case underlines the need to measure activation, assisted access and privacy alongside issuance.
Limits: Cards sent and produced measure administrative throughput, not possession, activation, usage or protection against fraud. The Registradurรญa describes the app as still under pilot review and does not provide an independent assessment of accessibility, privacy, exclusion or whether digital services reduce the burden of proving identity. A production count is evidence of a national rollout, not proof of a complete digital identity system.
๐จ๐ดColombiasince 2011still operatingbears on Institutional responsiveness
In 2024 Colombia's Unidad para las Vรญctimas ordered 307,619 administrative indemnifications benefiting 285,537 victims, approved 79 collective reparation plans and closed three plans for communities affected by the armed conflict.
307,619 indemnifications, administrative indemnifications ordered, 2024. 285,537 victims, victims benefiting from administrative indemnifications, 2024. Unidad para las Vรญctimas, 2024 accountability report, retrieved 2026-08-31.
The move, and what it needed
Colombia built a standing administrative route from a national victims register to individual payments, collective plans and territorial support, allowing the state to process a very large caseload while adding differentiated treatment for communities, ethnic groups and vulnerable people.
Preconditions
A legal entitlement and registry that make victims visible to a national administrative system
Budget, local offices and fiduciary processes able to deliver payments and plans over many years
Coordination with municipalities, community representatives and other social services so compensation is not the only response
Where else it has been used: Post-conflict reparations systems travel as legal and administrative ideas, but Colombia shows the scale challenge: a durable process can institutionalise recognition while still leaving a long queue and unresolved security conditions.
Limits: Administrative orders and collective-plan closures show the reach of a reparations bureaucracy, not whether victims received money promptly, were made whole, felt safe or regained livelihoods. The Unit reports its own outputs and the eligible population is much larger than the annual caseload. A payment or plan milestone cannot resolve the justice, security and territorial conditions that produced the harm.
๐จ๐ดColombiasince 2011still operatingbears on Disaster preparedness and recovery
After the 2010โ11 La Niรฑa floods, Colombia's Fondo Adaptaciรณn had advanced resettlement, reconstruction and improvement for more than 39,000 households by its 2021โ23 evaluation, presenting the work as reducing climate exposure and fiscal risk.
After a nationwide flood emergency, Colombia created a dedicated recovery fund that combined housing reconstruction with resettlement and risk-reduction works, using the recovery pipeline to move some households away from recurring hazards rather than simply replacing what was lost in place.
Preconditions
A national disaster fund with authority to coordinate housing, infrastructure and land decisions across departments
Damage and exposure information that can identify safer sites and prioritise households
Long-term finance and local implementation capacity beyond the first emergency-relief phase
Where else it has been used: Recovery funds and safer-resettlement programmes travel after major floods, but the Colombian case shows that recovery becomes preparedness only when the post-disaster pipeline changes exposure and land-use decisions rather than rebuilding the same vulnerability.
Limits: The evaluation's household count and exposure claims are public-administration figures, not an independent test of construction quality, long-term safety or whether the households remain out of harm's way. It combines resettlement, reconstruction and improvement across different projects and years, so it cannot be read as a single completion rate. Reduced exposure also depends on land-use control, maintenance and future climate conditions outside this record.
๐ฏ๐ตJapansince 2016still operatingbears on Institutional responsiveness
By July 2024 Japan reported 93.08 million My Number Cards held, about 75% of the population; 73.71 million had valid health-insurance-card registrations in June 2024, while 566 businesses used the public-key identity infrastructure.
Japan built one nationally recognised credential and progressively connected it to health insurance, public accounts, online procedures and private identity verification, using legal changes and a central digital agency to extend the rail while responding to implementation problems.
Preconditions
A national resident-number and civil-registration system that can anchor one credential
A central authority able to align ministries, municipalities, hospitals, banks and software providers
Correction, support and privacy mechanisms that can handle errors as the identity rail expands
Where else it has been used: Foundational digital identity can travel, but Japan's experience shows the difficult second phase: linking a widely issued card to many services while retaining a way to correct records and support people who cannot rely on an app.
Limits: Card holdings and registrations measure administrative reach, not active use, accuracy, privacy, accessibility or whether services became easier. The Digital Agency aggregates government and survey data, and the health-insurance and private-sector figures cover different dates and definitions. A high stock of cards can coexist with misregistrations, support burdens and people who cannot use a digital credential.
๐ฏ๐ตJapansince 2011still operatingbears on Disaster preparedness and recovery
After the 2011 earthquake and tsunami, Japan completed the planned 29,654 public-housing units by December 2020, completed 18,226 higher-ground residential-land units, opened about 570 kilometres of reconstruction roads by December 2021 and restored all damaged railways by March 2020.
29,654 housing units, public-housing units completed in the reconstruction plan, 2020-12. 570 kilometres, reconstruction roads opened to traffic, 2021-12. Reconstruction Agency of Japan, progress to date, retrieved 2026-08-31.
The move, and what it needed
Japan set a ten-year reconstruction frame, created a dedicated agency and combined higher-ground relocation, public housing, roads, rail restoration and livelihood support, reviewing the programme as the ordinary disaster areas moved from emergency recovery toward long-term revitalisation.
Preconditions
A national fiscal and administrative framework able to coordinate prefectures and municipalities after a catastrophic shock
Land-use, housing and infrastructure plans that can move people and assets out of hazard zones
A staged review process that separates completed physical works from the longer social and nuclear-recovery problems
Where else it has been used: Large post-disaster reconstruction agencies are a common response, but Tลhoku demonstrates why the physical delivery ledger must remain beside population, livelihood and social-recovery measures rather than being treated as the whole recovery.
Limits: Completion of housing, roads and railways shows delivery against a public programme, not whether communities regained population, livelihoods, trust or long-term safety. The agency notes that nuclear-disaster areas and support for evacuees continue, and the figures mix planned units, roads and different completion dates. A completed asset is not the same as a recovered place.
๐ต๐ชPerusince 2005still operatingbears on Institutional responsiveness
Peru's Juntos conditional cash-transfer programme, created in 2005 and placed under MIDIS in 2012, operated in 1,885 districts and benefited 793,623 households by August 2024 through SISFOH-targeted support.
Peru linked regular cash transfers to health and education conditions and used the intergovernmental SISFOH targeting system to identify poor households, giving a national programme a way to reach dispersed communities while retaining a welfare-policy feedback and evaluation function.
Preconditions
A household targeting registry that national and local institutions can maintain and update
Payment rails and local schools and health posts able to verify conditions without making access impossible
A ministry with the budget and monitoring capacity to adjust coverage as poverty and shocks change
Where else it has been used: Conditional transfers travel widely, but Juntos shows the dependency often hidden by the beneficiary number: targeting, local service quality and reliable payment systems decide whether a national register becomes a capability or just a list.
Limits: Household coverage shows administrative reach, not whether transfers arrive on time, improve health or schooling, reduce poverty durably or exclude eligible families. MIDIS reports the programme totals and the targeting system, while causal effects require evaluations rather than beneficiary counts. District presence can also mask uneven access across a mountainous and remote country.
๐จ๐ทCosta Ricasince 2005still operatingbears on Institutional responsiveness
Costa Rica's MICITT reported in October 2023 that more than 60 public institutions used certified digital signatures across more than 100 citizen services, with almost 75,000 certificates issued during that year.
Costa Rica combined a law granting electronic documents legal force with a national certification hierarchy, MICITT oversight and Banco Central support, then let agencies reuse the credential for signatures and online procedures instead of inventing separate trust systems.
Preconditions
A legal framework that makes a digital signature equivalent to a handwritten one
A certificate authority and identity-enrolment process that citizens and institutions can access
A central coordinator able to publish standards and help agencies migrate real procedures
Where else it has been used: Digital-signature systems travel more easily than full digital government because they start with a legal trust decision. Costa Rica shows that certificates become useful only when agencies convert them into complete services rather than a token login.
Limits: Institution and service counts show adoption of a legal and technical rail, not completion rates, user equity, fraud resistance or savings. Certificates issued are not the same as active users, and the adoption figures are reported by the coordinating ministry rather than independently audited across agencies.
๐จ๐ทCosta Ricasince 2010still operatingbears on Institutional responsiveness
Costa Rica's IMAS reported that its national Childcare and Early Development benefit reached 25,704 children in 2024 with an investment of โก13.2 billion, while interinstitutional referrals used SINIRUBE to match eligible families to available places.
Costa Rica linked IMAS, PANI, CEN-CINAI, local authorities and private providers through the REDCUDI network, using the SINIRUBE social registry to refer eligible families into available childcare capacity rather than asking each institution to target alone.
Preconditions
A shared beneficiary registry with rules for privacy, eligibility and referral
Multiple public and private providers that can accept common standards and funding
A financing mechanism able to sustain places across agencies and localities
Where else it has been used: Integrated childcare networks travel as a coordination problem, not just a subsidy. Costa Rica's model highlights the registry and referral layer that lets scarce places move toward eligible families while leaving delivery plural.
Limits: Beneficiary and spending totals show administrative reach, not child development, quality, waiting lists or whether the service enables parents to work. The programme combines several providers and modalities, so the national total can hide very different experiences and does not independently audit outcomes.
๐ช๐จEcuadorsince 2000still operatingbears on Institutional responsiveness
Ecuador adopted the US dollar as legal tender in January 2000 during a banking and currency crisis; a Central Bank review reports inflation falling from 96.1% in 2000 to below 8% in 2003 while the new regime removed the exchange-rate tool.
Facing a collapsing currency and banking system, Ecuador made a legally binding switch to the US dollar, converted prices and contracts, and rebuilt financial supervision around a regime that citizens could not devalue by political decree.
Preconditions
Political authority to make a fast, irreversible monetary commitment during a crisis
Access to enough foreign currency and a banking system that can clear dollar payments
Fiscal and financial reforms that prevent the new anchor from being undermined by domestic insolvency
Where else it has been used: Dollarization travels as a crisis commitment, not a generic stability policy. Ecuador's experience makes the trade-off explicit: credibility can be imported, but monetary flexibility and lender-of-last-resort capacity cannot.
Limits: The fall in inflation is an outcome of a regime change during a crisis, not a clean causal estimate of dollarization alone; oil prices, banking repair and fiscal policy also moved. The arrangement stabilises the nominal anchor but gives Ecuador less room to respond to external shocks through its own exchange rate or money supply.
๐ช๐จEcuadorsince 2016delivered and closedbears on Disaster preparedness and recovery
After Ecuador's 7.8-magnitude earthquake in April 2016, the national emergency committee reported 17 shelters and 59 temporary shelters active, assisting 25,931 people while mobile health units and security forces were deployed across affected provinces.
Ecuador used a national COE structure with provincial mesas to combine emergency logistics, health, security, road repair and social assistance, while publishing situation reports that gave agencies a common operating picture during a fast-moving shock.
Preconditions
A statutory emergency-management authority with a chain of command across provinces
Pre-existing military, police, health and municipal logistics that can be pooled quickly
A reporting cell able to update needs, shelters and service restoration as conditions change
Where else it has been used: Emergency coordination structures travel when they are rehearsed before the shock. Ecuador's report shows the value of a common operating picture, but the response layer still has to connect to a multi-year recovery authority.
Limits: This is an early response snapshot rather than a completed reconstruction account. Shelter and deployment counts show coordination under pressure, not whether displaced households returned safely, infrastructure was rebuilt resiliently or support reached every affected community.
๐ต๐ฆPanamasince 2009still operatingbears on Institutional responsiveness
Panama's 120 a los 65 programme paid a non-contributory pension to an average of about 123,000 older people per payment during July 2019โJune 2024, with B/.890.835 million programmed over the period and a national beneficiary register used to target eligibility.
Panama created a tax-funded pension for older people outside formal contributory schemes, then used the Registro Nacional de Beneficiarios and scheduled payments through banks and hard-to-reach-area operations to make eligibility portable across provinces and comarcas.
Preconditions
A national identity and beneficiary register that can verify age, residency and pension status
A budget commitment that survives changes in the contributory pension system
Payment logistics and local access points able to reach people outside the banking network
Where else it has been used: Non-contributory pensions travel where formal employment leaves older people uncovered. Panama's case shows that the registry and payment route are as important as the transfer promise, especially in remote and Indigenous territories.
Limits: Beneficiary and budget figures show administrative reach, not adequacy, take-up, payment regularity or reductions in poverty and health insecurity. The average covers several years and the register can miss older people in remote areas; the programme's own report is not a causal evaluation.
๐ต๐ฆPanamasince 2023operating below its peakbears on Disaster preparedness and recovery
During the extreme 2023โ24 drought, the Panama Canal Authority adjusted draft and booking rules to preserve water for the canal and the lakes supplying more than half of Panama's population, reducing daily transits from the normal 36 to 22 in December 2023 while keeping the waterway operating.
22 transits per day, daily transits during the december drought restriction, 2023-12. 36 transits per day, normal daily transit capacity, normal conditions. Autoridad del Canal de Panamรก, drought transit advisory, retrieved 2026-08-31.
The move, and what it needed
The canal authority used hydrological forecasts, vessel booking quotas and draft restrictions to ration a shared water system, protecting drinking-water reserves while preserving a predictable minimum service for maritime customers.
Preconditions
Real-time water and traffic data with an authority empowered to change operating rules
A single institution able to trade off shipping revenue against urban water security
Customers and pilots who can plan around transparent, staged restrictions rather than abrupt closure
Where else it has been used: Water-stressed infrastructure needs an operating playbook, not only new capacity. Panama's drought response shows how transparent rationing and forecasts can preserve a service while a longer watershed solution is built.
Limits: The reduction is a documented loss of peak transit capacity during one drought, not evidence that the watershed problem is solved or that every user receives equal service. The canal authority sets the rules and reports the trade-offs; the record does not quantify the effect on households, ecosystems or shipping costs, and later rainfall can restore capacity without removing the underlying vulnerability.
๐ง๐ดBoliviasince 2009still operatingbears on Institutional responsiveness
Bolivia's Bono Juana Azurduy opened its 2024 payment round through 47 authorised financial institutions in all 340 municipalities and six Indigenous territories, tying the transfer to prenatal, delivery and early-childhood health checks; the programme reported 1,525,566 pregnant women and 1,783,936 children reached between 2009 and 2023.
Bolivia linked a modest cash incentive to a sequence of prenatal, delivery and child-health checks, then used a nationwide bank network and mobile intersectoral brigades to move the programme beyond urban clinics and into Indigenous territories.
Preconditions
A beneficiary and health-record system able to verify visits without excluding people with incomplete documentation
Payment institutions and mobile teams that can reach remote and Indigenous communities
Primary-care facilities with enough staff and supplies for the conditions to mean more than a paperwork exercise
Where else it has been used: Conditional health transfers travel best when the payment rail and the service network are designed together. Bolivia's case makes the last-mile condition visible: national coverage is a logistics problem as much as a budget line.
Limits: Municipal and financial-channel coverage shows administrative reach, not whether every eligible mother is paid on time, attends the required visits or experiences better maternal and child outcomes. The ministry reports the programme's own totals, while the record does not establish the causal effect of the transfer or the quality of the health facilities where conditions are verified.
๐ง๐ดBoliviasince 2007still operatingbears on Institutional responsiveness
Bolivia's Renta Dignidad, created by Law 3791 in 2007 and paid since 2008, reached 1,241,440 beneficiaries in 2024 with Bs5.289 billion disbursed through the national social-security administration.
Bolivia converted a universal old-age entitlement into a predictable non-contributory payment, administered through the long-term social-security system and financed as a standing national obligation rather than a short project.
Preconditions
A population register and identity process that can establish age and prevent duplicate payments
A fiscal transfer and payment network that reaches older people outside contributory employment
Rules that protect payment continuity while the financing base and population structure change
Where else it has been used: Social pensions travel where informal work leaves many people outside contributory schemes. Bolivia shows the reach a standing entitlement can create, while leaving the financing and adequacy questions explicit.
Limits: Beneficiary and disbursement totals show the scale of a transfer system, not whether the benefit is adequate, paid reliably in every remote area or sufficient to reduce old-age poverty. The report is an administrative account and does not provide a counterfactual estimate of welfare effects or the programme's fiscal resilience under demographic change.
๐ต๐พParaguaysince 2005still operatingbears on Institutional responsiveness
Paraguay's Tekoporรฃ Mbarete social-protection programme protected more than 175,000 families nationwide in 2024, paid monthly after a 2023 redesign and reported 95% bankarisation plus georeferencing of 11,316 participating homes.
Paraguay combined a recurring transfer with family guides, monthly payment and a georeferenced beneficiary register, then pushed bank access into a programme that had previously paid less frequently and reached families unevenly.
Preconditions
A social registry and field network able to identify, revisit and update vulnerable households
Payment rails that can reach rural and Indigenous families without making bank access a new exclusion test
Health, education and social-work services capable of turning a transfer into sustained household support
Where else it has been used: Social-protection systems travel when the registry and payment rail travel with the benefit. Paraguay's redesign shows how frequency, field accompaniment and financial inclusion can be treated as one delivery problem.
Limits: Family coverage and payment-channel penetration show administrative reach, not whether transfers arrive on time, meet household needs or improve health, schooling and income. The report is the government's own account and does not provide a counterfactual impact estimate; georeferencing also cannot by itself reveal who remains excluded from the registry.
๐ต๐พParaguaysince 2021still operatingbears on Institutional responsiveness
Paraguay's tax administration approved 651,894,711 electronic tax documents during 2024 and reported 6,659 electronic billers using the SIFEN system, combining mandatory onboarding for larger firms with the free e-Kuatia'i route for small taxpayers.
Paraguay made the tax document a machine-readable event: SIFEN set a national standard, validated documents centrally and combined a mandatory calendar for larger taxpayers with a free tool and electronic signature path for smaller ones.
Preconditions
A tax authority with legal power to set the document standard and phase in compliance
Reliable identity, signature and validation services that firms can integrate at low cost
A support and exception process for small businesses with limited connectivity or software capacity
Where else it has been used: Electronic invoicing travels as a state-business coordination project, not merely an IT purchase. Paraguay's combination of a central validation rail and a free small-firm route is the part worth comparing.
Limits: Document and taxpayer counts show system throughput and adoption, not the revenue effect, compliance cost, uptime or whether informal firms can participate without disproportionate burden. The tax authority reports the implementation totals, and the staged mandate means the 2024 base was not yet a census of all taxpayers.
๐ฆ๐ทArgentinasince 2009still operatingbears on Institutional responsiveness
Argentina's Asignaciรณn Universal por Hijo (AUH) reached 4,083,987 children and adolescents in June 2024, with 2,324,716 adult titulars receiving the benefit through the national social-security administration and its health, vaccination and education conditions.
Argentina made children the legal beneficiaries of a national cash transfer, used ANSES identity and household records to deliver it at scale and held back a share until families documented health, vaccination and education conditions.
Preconditions
A national identity and social-security database that can link children, adult titulars and payment accounts
Schools and health services able to verify conditions without turning documentation into a barrier
A durable budget and appeal process that can keep the entitlement responsive through inflation and labour-market shocks
Where else it has been used: Child benefits travel when the register, payment system and service verification are treated as one capability. Argentina's AUH illustrates both the reach of that architecture and the need to measure exclusion and real purchasing power separately.
Limits: Beneficiary counts show the reach of a transfer system, not whether payment is adequate, households complete the required checks or children experience better outcomes. ANSES reports the administrative totals, while the record does not establish a causal effect or reveal how many eligible families remain outside the registry.
๐ฆ๐ทArgentinasince 2010still operatingbears on Institutional responsiveness
Argentina's SUBE system distributed more than five million cards and recorded about 6.7 million new cards in 2024, while expanding the network to more than 30,000 recharge points and making SUBE Digital available in every locality.
Argentina built one interoperable fare and data layer across jurisdictions, then used the transaction record to target subsidies, integrate transfers and add a mobile payment option without abandoning the physical card for people who need it.
Preconditions
A national standard that transport operators and provincial and municipal systems can adopt
Retail and digital channels that keep cards and recharging usable outside major cities
A public data and subsidy authority able to reconcile transactions with operator payments
Where else it has been used: A common fare rail travels when it solves a state problem as well as a passenger problem. SUBE's value is the combination of inclusion, operational data and subsidy control, not the card alone.
Limits: Cards and recharge points measure access to the payment rail, not travel-time improvements, fare affordability, service reliability or the accuracy of subsidies. The ministry's figures describe distribution and availability, while the record does not independently audit usage, system uptime or the effect of later open-payment reforms.
๐ต๐ญPhilippinessince 2018still operatingbears on Institutional responsiveness
The Philippine Statistics Authority reported 90,017,565 people registered to PhilSys by 19 September 2024, or 97.8% of the year's 92 million registration target, with the ID designed for government and private-sector transactions.
The Philippines made one permanent identifier the base for public and private transactions, combined PSA registration with mobile and local-government channels and added physical, paper and digital credentials so the system could reach people with different levels of connectivity.
Preconditions
A civil-registration authority able to reconcile demographic and biometric records at national scale
Registration sites and mobile teams that can reach islands, remote barangays and young children
Clear legal rules and relying-party integration so the credential is accepted beyond the issuing agency
Where else it has been used: Foundational IDs travel when enrolment, authentication and service integration are funded as one system. PhilSys shows how quickly registration can scale, while delivery and acceptance remain separate tests.
Limits: Registration against a target shows the reach of an identity enrolment campaign, not whether records are accurate, cards arrive promptly, agencies actually accept the credential or people receive better services. The PSA reports the milestone and target; this record does not independently assess exclusion, privacy or authentication performance.
๐ต๐ญPhilippinessince 2008still operatingbears on Institutional responsiveness
The Philippines' 4Ps conditional cash-transfer programme reached 4,064,354 households by 30 June 2024 across 41,746 barangays, 148 cities and 1,485 municipalities, with beneficiaries linked to health, education and nutrition conditions.
The Philippines combined a national household-targeting system with conditional grants, local verification and regular programme monitoring, allowing a central social-protection rule to operate through thousands of barangays and municipal service networks.
Preconditions
A household registry that can be updated as poverty, migration and family composition change
Schools, health facilities and local governments able to verify conditions without making remote families carry the full administrative burden
Payment and grievance channels that let beneficiaries correct records and challenge missed or suspended grants
Where else it has been used: Conditional transfers travel through local service capacity. The Philippine case is useful because its national scale makes the registry, verification and grievance machinery visible rather than treating the cash grant as the whole programme.
Limits: Household and barangay coverage show administrative reach, not whether payments arrive on time, children attend school, health outcomes improve or eligible families are excluded. The implementation report is the programme's own account and does not replace an independent causal evaluation or measure the quality of local services that conditions depend on.
๐ต๐ญPhilippinessince 2013still operatingbears on Disaster preparedness and recovery
After Typhoon Yolanda, the Philippines' recovery programme had provided shelter assistance to about 1.14 million families by September 2016; the National Housing Authority had completed 29,661 permanent units while 102,240 more were under construction across six regions, 14 provinces and 115 cities and municipalities.
The Philippines combined emergency shelter assistance with a multi-agency recovery plan, permanent housing and local-government coordination, while documenting the land, permitting and procurement bottlenecks that slowed the transition from relief to safer settlement.
Preconditions
A recovery authority able to coordinate national agencies, local governments, donors and affected communities
Pre-agreed land, titling and no-build-zone rules that can be activated after a disaster
Monitoring that tracks completed and occupied units, not just funds released or houses contracted
Where else it has been used: Disaster recovery travels as a governance sequence: cash and shelter first, then land, housing and livelihoods. Yolanda is valuable because its implementation record shows both the reach and the friction in that sequence.
Limits: The figures are a 2016 implementation snapshot, not a final account of long-term recovery, land tenure, livelihoods, relocation quality or resilience to the next storm. The same source records major housing delays caused by no-build zones, land conversion, titling and local approvals; shelter counts therefore should not be read as proof that recovery was complete or equitable.
๐ฎ๐ฉIndonesiasince 2014still operatingbears on Institutional responsiveness
Indonesia's Jaminan Kesehatan Nasional (JKN) health-insurance system had 275,413,133 registered participants, 97.50% of the population, by 31 July 2024; 219,263,970 participants were active.
Indonesia combined a national entitlement with a single administrative pool, subsidised poor households and required employers and local governments to enrol their populations, making portability across the archipelago possible.
Preconditions
A population register and membership identifier that can deduplicate people across schemes
A financing rule that lets the state subsidise households unable to pay
Claims and provider data capable of monitoring both coverage and active status
Where else it has been used: Near-universal insurance travels poorly when enrolment is easier than keeping contributions current. Indonesia's separate registered and active counts make that implementation gap visible.
Limits: Registration is not the same as timely access, quality of treatment or financial protection: the same report shows 56.15 million participants were inactive, including people in arrears or undergoing eligibility verification. The participant totals are compiled by the social-security council from BPJS and civil-registration data and are not an independent audit of claims or health outcomes.
๐ฎ๐ฉIndonesiasince 2014still operatingbears on Disaster preparedness and recovery
Indonesia's disaster agency runs InaRISK as an open, integrated risk-management platform; its 2024 national dashboard covered about 7,100 kecamatan and combined BNPB, geospatial, statistical, health and education data for exposure and risk mapping.
BNPB put hazard, exposure, vulnerability and capacity layers behind one public map, giving national and local planners a shared geography for deciding where preparedness resources should go.
Preconditions
Common administrative geographies and data standards across ministries
A technical agency able to maintain geospatial layers between disasters
Local authorities able to turn a risk map into drills, land-use rules and response plans
Where else it has been used: Risk maps travel when they are treated as operating infrastructure rather than one-off reports. InaRISK is a useful delivery case while leaving the harder questionโwhether the map changes readinessโopen.
Limits: A mapped risk and exposure layer is an input to preparedness, not evidence that warnings arrive, evacuation works or recovery is fast. The dashboard combines modelled hazards and administrative datasets that can change with new boundaries and inputs; it does not provide a comparable outcome measure of lives, assets or services protected.
๐ป๐ณVietnamsince 2021still operatingbears on Institutional responsiveness
Vietnam's Ministry of Public Security reported more than 70.2 million electronic-identification accounts issued by January 2024, with VNeID designed to replace paper documents and authenticate public and private services against the national population database.
Vietnam made one population-data-backed identity the default credential for administrative services, then connected health, driving, social-security and financial attributes to it instead of asking each ministry to issue another account.
Preconditions
A population database with a stable identifier and identity-proofing process
Legal authority to make one credential valid across ministries and service providers
Interoperability and privacy controls for the sensitive records attached to the identity
Where else it has been used: A national identity app travels only when service owners actually retire their parallel logins. Vietnam's account count shows the rollout; adoption and safeguards remain separate tests.
Limits: Accounts issued are not the same as active use, successful service completion or public trust in a single identity provider. The number is reported by the ministry that runs the system and the article also records unresolved rules for organisation identities, authentication levels and how widely the service can be distributed.
๐ป๐ณVietnamsince 1995still operatingbears on Institutional responsiveness
Vietnam Social Security reported 95.523 million people enrolled in health insurance in 2024, covering 94.29% of the population and approaching the country's 95% universal-coverage target for 2025.
95,523,000 people, health-insurance participants, 2024. 94.29 % of population, population covered by health insurance, 2024. Vietnam Social Security, coverage progress report, retrieved 2026-08-31.
The move, and what it needed
Vietnam merged the administrative systems for social and health insurance, used compulsory and subsidised categories to widen the pool and kept a long-running public target visible as coverage moved from a minority programme toward universality.
Preconditions
A national social-security agency with local offices and a common beneficiary record
Public subsidies for people whose income makes contributions unaffordable
A provider and claims system capable of serving a rapidly expanding insured population
Where else it has been used: Coverage expansion is easier to copy than the administrative merger behind it. Vietnam's long series is useful because it shows the institutional work between a legal entitlement and near-universal enrolment.
Limits: Enrolment and coverage do not show whether people can obtain timely, affordable and high-quality care, or whether minority and remote communities receive the same service. The series is reported by the agency administering the scheme, and the population denominator and programme rules have changed over the three decades in the source.
๐ป๐ณVietnamsince 2024still operatingbears on Disaster preparedness and recovery
After Typhoon Yagi struck northern Vietnam, the government reported on 16 September 2024 that power had returned to 92.5% of customers in affected localities and that 8,386 of 9,235 damaged mobile base stations had been restored.
Vietnam used central emergency direction, utility and telecom repair teams, and a shared damage report to restore lifeline networks while the storm response was still unfolding, keeping power and communications as parallel recovery priorities.
Preconditions
An emergency command chain able to direct utilities across provinces
Asset inventories and field reporting that identify damaged substations and base stations quickly
Repair crews, spare parts and access routes that can operate while floods and landslides continue
Where else it has been used: Recovery capacity is visible in the speed and completeness of restoration, not only in relief spending. Yagi provides a dated operational checkpoint while leaving long-term adaptation open.
Limits: This is an early restoration snapshot, not a final account of deaths avoided, housing and livelihood recovery, or resilience before the next storm. The ministry report covers affected northern localities and restoration status, while the figures do not independently verify the quality or durability of repaired power and communications assets.
๐ฒ๐พMalaysiasince 2021still operatingbears on Institutional responsiveness
Malaysia's national COVID-19 immunisation programme had administered 57,119,777 doses by 31 December 2021, while the MySejahtera application was used to collect 974,134 adverse-event responses for vaccine safety monitoring.
57,119,777 doses, covid-19 vaccine doses administered, 2021-12-31. 974,134 responses, adverse-event responses submitted through mysejahtera, 2021-12-31. Malaysia Ministry of Health, Annual Report 2021, retrieved 2026-08-31.
The move, and what it needed
Malaysia paired a central vaccination command with a single citizen-facing app for registration, appointments, certificates and safety reporting, then used public clinics and contracted providers to move millions of doses through one operational workflow.
Preconditions
A national health identity and digital appointment channel
Cold-chain logistics and a network of public and private vaccination sites
A safety-monitoring process that can connect self-reports to vaccine records
Where else it has been used: A crisis app is useful when it is attached to the delivery chain, not just a status screen. Malaysia's dose and reporting counts show both throughput and the limits of self-reported monitoring.
Limits: Doses and safety-report counts show the ability to mobilise a national vaccination operation, not coverage equity, completed-course protection, clinical outcomes or whether every adverse event was reported. The annual report is produced by the ministry running the programme and the MySejahtera responses are passive self-reports, so neither number is an independent measure of safety or effectiveness.
๐ฒ๐พMalaysiasince 2023still operatingbears on Disaster preparedness and recovery
Malaysia's National Disaster Management Agency (NADMA), water and meteorological agencies and telecommunications companies sent more than five million flood-warning SMS messages during the 2023/24 monsoon season and upgraded 442 transmission-tower sites to keep warnings and communications functioning.
NADMA joined hydrology, weather, communications regulation and commercial telcos into one warning workflow, using the mobile network's national reach while requiring operators to harden the physical sites that warnings depend on.
Preconditions
A single emergency agency able to trigger messages across competing networks
Hydrological and weather data tied to geographies that operators can target
Network resilience work funded before the monsoon rather than after an outage
Where else it has been used: Public warnings travel when the alert authority, forecast data and last-mile carrier are joined. Malaysia's system shows the coordination architecture; warning quality still needs independent testing.
Limits: Message volume and tower upgrades show an alert channel and network hardening, not whether warnings arrived before local flooding, were understood, or changed evacuation behaviour. NADMA reports the output itself, and the source does not provide false-alarm rates, geographic reach or outcomes for people who received the messages.
๐น๐ญThailandsince 2002still operatingbears on Institutional responsiveness
Thailand's National Health Security Office has sustained the tax-funded Universal Coverage Scheme since 2002; in fiscal 2023 it covered more than 47.2 million people and contracted 17,247 public and private health units to provide care.
Thailand created a statutory national purchaser, financed coverage from the public budget and repeatedly revised the benefits package and access rules instead of treating the 2002 launch as a one-off reform. A common fund could contract a mixed provider network and use the national ID rather than a paper card to carry eligibility.
Preconditions
A legal entitlement and budget process that survives changes of government
A purchaser with data and bargaining capacity across public and private providers
A population register and administrative claims system that can verify eligibility
Where else it has been used: Thailand's purchaser model is studied across low- and middle-income countries. The transferable move is durable pooled financing plus an agency that can learn the benefit package, not the exact 30-baht co-payment.
Limits: Beneficiary and provider counts show institutional reach, not equal quality, timeliness, financial protection in every case or the health outcomes attributable to the scheme. NHSO is the programme manager and reports its own administrative totals; the provider count includes units with different capacities and does not show how many patients each served. The record also covers one of three tax-funded schemes, not every route to Thai health coverage.
๐น๐ญThailandsince 2022still operatingbears on Institutional responsiveness
Thailand's Digital ID Framework aligned government and private identity providers around common standards in its 2022โ2024 first phase; the Electronic Transactions Development Agency reports at least 113 million users across the major ThaiD, Tang Rat, Paotang and NDID platforms and services.
ETDA acted as a coordinating standards body while the interior ministry, a state bank and financial-sector network each kept a user-facing identity channel. A common framework allowed those channels to connect to government and private services without forcing the country to wait for one monolithic identity provider.
Preconditions
A population-data authority able to anchor identity assurance
A standards body with convening power across state agencies, banks and platforms
Security, consent and recovery rules for people who lose a device or need assisted access
Where else it has been used: Federated identity ecosystems travel more easily than a single national login, but only if the trust and liability rules are explicit. Thailand's overlapping platforms make the scale visible and the deduplication problem visible too.
Limits: The 113-million figure is an aggregate of platform-reported users and can include overlapping accounts, so it is not a count of unique people or proof that all services accept the same credential. The ETDA review describes framework progress and platform stocks, not authentication success, exclusion, security incidents or the cost of integration. The framework's second phase was still expanding its target groups and service coverage when the source was published.
๐น๐ญThailandsince 2025still operatingbears on Disaster preparedness and recovery
Thailand launched T-Alert in September 2025 as a nationwide cell-broadcast warning system: it covers eight hazard categories, was built with the disaster agency, digital ministry, telecom regulator and AIS, True and NT, and followed three staged tests in 2025 before use in real floods and other incidents.
Thailand joined a disaster-warning centre to the telecom regulator, digital ministry and all three major mobile operators, using cell broadcast so the authority can target a geography without a subscriber list. Staged tests at building, district and provincial levels preceded wider use, while hazard-specific thresholds and field data determine when to alert.
Preconditions
A legally recognised authority able to issue a national alert through every operator
A common cell-broadcast gateway and compatible handset standards
Hazard data and local incident workflows that can turn a forecast into a bounded target area
Where else it has been used: Cell broadcast is spreading as a complement to sirens and SMS. Thailand's staged rollout shows the coordination needed to make a multi-operator warning channel real, while its handset limits keep redundancy essential.
Limits: The hazard count and test phases show a national warning architecture, not warning lead time, receipt by every at-risk person, false-alarm rates or evacuation outcomes. Cell broadcast depends on compatible 4G/5G devices and current operating systems, while older phones require another channel. The government reports the rollout and examples of use; independent tests of alert comprehension and last-mile action are still needed.
๐ช๐ธSpainsince 2014still operatingbears on Institutional responsiveness
Spain built Cl@ve as a common public-sector platform for identification, authentication and electronic signatures; the Spanish Tax Agency reported 20.867 million taxpayer identifiers enrolled in Cl@ve Mรณvil by 31 December 2024, including 12.323 million citizens active in the app.
Spain created one public-sector identity layer that agencies could embed in their own procedures, then added a mobile channel while retaining other assurance routes. A central platform reduced repeated account creation and let a high-volume agency such as the Tax Agency test and scale the identity flow.
Preconditions
A stable taxpayer and population registry that can anchor identity assurance
A central operator able to publish standards and support many agencies
Fallback credentials and assisted routes for people who cannot use a smartphone
Where else it has been used: Shared authentication rails travel when agencies agree on assurance and liability. Spain's separate identifier and active-user counts are a useful reminder that reach is not the same as unique people or successful service journeys.
Limits: Enrollment and active-app counts show a reusable identity rail, not whether every public service is reachable, whether people without smartphones can use it or whether authentication is secure against all fraud. The Tax Agency reports its Cl@ve Mรณvil figures and counts taxpayer identifiers separately from citizens, so the two numbers must not be added into a population total. The platform is shared across the public administration, but the record does not isolate its causal effect on service completion.
๐ช๐ธSpainsince 2021still operatingbears on Disaster preparedness and recovery
After the Cumbre Vieja eruption began on 19 September 2021, Spain coordinated a stateโCanary Islandsโisland recovery commission; by February 2024 the state reported โฌ929 million mobilised and 650 CO2 monitoring devices installed, allowing more than 220 families to return safely to homes in affected areas.
929,000,000 euros, state resources mobilised for la palma recovery, 2024-02. 650 devices, co2 monitoring devices installed in affected areas, 2024-02. Government of Spain, La Moncloa recovery commission report, retrieved 2026-08-31.
The move, and what it needed
Spain created a mixed commission linking the central government, the autonomous region, island government and municipalities, then paired large-scale aid with a locally targeted monitoring system. The CO2 programme turned a continuing hazard into a measured re-entry protocol rather than treating the end of the eruption as the end of the emergency.
Preconditions
Clear authority to coordinate national funds with regional and municipal reconstruction decisions
Scientific monitoring that can translate a changing hazard into household-level safety rules
A recovery budget that remains available after media attention and the acute emergency decline
Where else it has been used: Post-disaster recovery travels through coordination and monitoring more than through a single grant. La Palma shows how a small island can use national fiscal capacity while keeping local authorities in the operational loop.
Limits: Mobilised resources and monitoring devices measure administrative response and a specific safety intervention, not the completeness of reconstruction, the speed and fairness of compensation or long-term livelihood recovery. The โฌ929 million is a government total against its valuation of damage and includes different instruments over several years. The return of more than 220 families to monitored homes is an operational output, not proof that every affected zone was safe.
๐ต๐ฑPolandsince 2015still operatingbears on Institutional responsiveness
Poland made the mDowรณd digital identity document in the mObywatel app legally equivalent to the physical identity card inside Poland; the Ministry of Digital Affairs reports that eight million people use the document, which can also support PESEL blocking and other public-service interactions.
8,000,000 people, people using the mdowรณd digital identity document, 2026-08. 1 nationally recognised digital identity document, legal status of mdowรณd inside poland, 2026-08. Polish Ministry of Digital Affairs, mObywatel usage update, retrieved 2026-08-31.
The move, and what it needed
Poland put a verifiable identity document and several high-value public functions in one mobile wallet, then gave the digital document explicit legal standing so offices, banks and other authorised institutions could accept it. A central app reduced the need for each service to build its own mobile credential.
Preconditions
A trusted national population register and physical identity issuance system
A state operator able to sign and revoke verifiable mobile documents
A legal rule that tells frontline institutions when a digital document is equivalent to paper
Where else it has been used: Digital wallets travel when legal recognition follows technical deployment. Poland's adoption shows the pull of a useful document, while the single-app design keeps recovery and exclusion as live governance questions.
Limits: The user count measures adoption of a government wallet, not successful completion of public services, equal access for people without compatible phones or the privacy and fraud risks of concentrating documents in one app. The ministry reports a rounded current total and does not state how often each user presents the credential. Legal equivalence applies in Poland and does not automatically make the document a cross-border EU credential.
๐ต๐ฑPolandsince 2016still operatingbears on Institutional responsiveness
Poland's ZUS-administered child-benefit system has paid support since 2016 and moved applications and payments online; by November 2023, the government reported nearly seven million children reached and almost PLN255 billion transferred, before the monthly benefit rose from 500 to 800 zลoty in January 2024.
Poland turned a broad entitlement into a single national payment workflow, then shifted administration to ZUS and required online applications through ZUS, Emp@tia or bank channels. Automatic uprating from 500 to 800 zลoty preserved continuity for existing families without a new eligibility decision.
Preconditions
A population and social-insurance register that can match children to guardians
A payment agency with nationwide bank-transfer and fraud-control capacity
Assisted application routes for households that cannot navigate online forms
Where else it has been used: Large child-benefit systems travel when payment infrastructure is simpler than case-by-case welfare administration. Poland's digital-only application rule shows the efficiency gained and the inclusion risk that must be managed alongside it.
Limits: The cumulative transfer and child counts show administrative reach, not poverty reduction, fertility effects, take-up among eligible families or the programme's fiscal sustainability. The government reports rounded totals and the period includes a change from 500+ to 800+, so the numbers do not describe one constant benefit. Payments are made to bank accounts, leaving delivery quality dependent on digital access and accurate beneficiary records.
๐ต๐ฑPolandsince 2018still operatingbears on Disaster preparedness and recovery
Poland's Government Security Centre launched Alert RCB as an opt-out-free emergency SMS channel in 2018; in its first year it activated the system 23 times and, working with mobile operators, sent about 200 million warning messages to phones in threatened areas.
RCB receives hazard information from crisis services and meteorological agencies, defines a threatened geography and legally requires every mobile operator to broadcast the same short warning without prior subscription. The channel is deliberately simple and complements local sirens and emergency services.
Preconditions
A central crisis authority able to trigger messages across all mobile operators
A geographic targeting system at least as precise as the powiat
Operator capacity and redundant channels for outages, roaming and people outside coverage
Where else it has been used: Mass SMS warnings travel when the state can use operators as a coordinated last-mile network. Poland's early data also makes the latency and duplication limits explicit rather than hiding them behind a reach percentage.
Limits: Messages and activations measure the scale of a warning channel, not delivery time, comprehension, false-alarm rates or whether people changed behaviour. The bulletin notes that large-area alerts could take two to three hours to distribute and that phones outside network coverage might not receive them. Roaming can also produce duplicate messages, so SMS volume is not a unique-person count.
๐ธ๐ชSwedensince 2003still operatingbears on Institutional responsiveness
Sweden's bank-issued BankID became a shared identity layer for public and private services: in 2024 it handled 7.6 billion identification or signature transactions for 8.6 million users across 7,500 connected services, with 99.99% availability.
Major banks formed a common issuer and technical standard, then public authorities and firms adopted the same authentication and signing flow instead of each building a separate credential. That let one already-identified customer base become national digital infrastructure without a single government agency issuing every credential.
Preconditions
Banks with near-universal customer relationships and reliable know-your-customer records
A stable personal identity number that public and private services can use as the joining key
Public-sector willingness to accept a privately operated identity at a common assurance level
Where else it has been used: BankID travels as a federated-identity pattern: the reusable move is a shared trust and liability framework, not the mobile app alone, and countries without a similarly concentrated banking or identity base will need a different issuer coalition.
Limits: BankID is a private bank consortium rather than a state identity card, and access requires a Swedish personal identity number, an issuing bank and the ability to use the app without assistance. Transaction volume and availability show a functioning identity rail, not equal access, fraud resistance, or the quality of the public services reached through it.
๐ธ๐ชSwedensince 2003still operatingbears on Institutional responsiveness
Sweden's regions and municipalities run the 1177 national health hub through Inera, combining a common website, logged-in care services and the nationwide nurse helpline; 1177.se recorded just over 260 million visits in 2021.
Regional health systems kept operational responsibility but agreed on one national entry point, shared content standards and a common identity-backed service layer. The arrangement turns a fragmented provider landscape into a recognisable front door without removing regional control over care.
Preconditions
A durable inter-regional body able to maintain common content, interfaces and support
A national identity and consent model for exposing records and care transactions
Regional funding and governance that can keep local services connected to the shared platform
Where else it has been used: 1177 travels as a federated public-service portal: a common national front door can coexist with decentralised delivery when the shared layer is narrow, maintained and trusted.
Limits: The 2021 web total includes an exceptional pandemic information load. The visit figure is on page 33 of the report, and the report does not count helpline calls. Regions still run their own advice operations and decide which logged-in services and records are visible, so national branding does not mean identical service quality or outcomes everywhere.
๐ธ๐ชSwedensince 1986still operatingbears on Disaster preparedness and recovery
Sweden's Viktigt meddelande till allmรคnheten (VMA) system combines radio, television, mobile messages, apps and outdoor sirens; SOS Alarm's 2025 report records 37 VMA alerts issued in 2024, alongside the four scheduled annual siren tests.
Authorised authorities can request one centrally coordinated warning, while SOS Alarm and broadcasters distribute it through several channels and the siren network is exercised on a fixed public schedule. Redundancy makes the warning less dependent on any one medium and gives residents a familiar routine for checking what a signal means.
Preconditions
A legally defined authority and a single operational gateway for issuing warnings
Agreements with broadcasters, mobile operators and municipalities that keep channels interoperable
Regular public tests and clear instructions that convert a signal into protective action
Where else it has been used: VMA travels as a layered-warning pattern: countries can add mobile and app channels without discarding sirens and broadcast, but only if the trigger authority and message vocabulary stay unambiguous.
Limits: Alert count measures the activation of a warning chain, not the share of people who received, understood or acted on a message. Outdoor sirens do not cover every location, mobile delivery depends on network and handset conditions, and scheduled tests are evidence of maintenance rather than proof of performance in a major disaster.
๐ฎ๐ชIrelandsince 2015still operatingbears on Institutional responsiveness
Ireland built MyGovID as a reusable online identity for welfare, tax, transport, education and childcare services; the 2025 Digital Decade country report records more than 2.7 million verified accounts in December 2024, close to 65% of adults.
Ireland joined identity verification run by the Department of Social Protection to a common sign-on adopted by other departments, so each service could reuse one assurance process rather than issue its own credentials. The account layer makes cross-department integration visible to citizens while leaving service ownership distributed.
Preconditions
A national identity-verification process that can be reused by multiple departments
A stable personal service number and legal basis for linking benefits and tax services
A central product owner with authority to set authentication and accessibility standards
Where else it has been used: MyGovID travels as a government-wide identity pattern: a single sign-on only joins services when the back-end data, assurance rules and fallback routes are coordinated as carefully as the login screen.
Limits: Verified accounts are a reach measure, not proof that every service is available or easy to use. The account requires identity-verification steps and a Public Services Card or equivalent route, and the adult-share denominator excludes children and people who remain offline; one account can also be used for many services without showing the quality of those transactions.
๐ฎ๐ชIrelandsince 2020delivered and closedbears on Disaster preparedness and recovery
Ireland's HSE vaccination programme built a national booking, supply and reporting operation during COVID-19; by 21 July 2021 its dashboard recorded 5,344,686 administered doses, including 2,854,040 first doses.
Ireland combined HSE clinics, general practitioners, pharmacies, a central registration process and national data reporting under one prioritisation strategy. The shared dashboard let planners see supply and administration together while local providers supplied the last mile.
Preconditions
A health service able to allocate scarce doses through a published national priority order
Cold-chain logistics and a common reporting system linking hospitals, GPs, pharmacies and clinics
Trusted clinical advice that can change eligibility and dosing rules without fragmenting delivery
Where else it has been used: The vaccination pattern travels as a federated emergency operation: a central queue and data standard can coordinate many providers, but the last-mile workforce and supply contracts determine whether the plan becomes injections.
Limits: The July 2021 dashboard is an interim administrative snapshot and warns that recorded totals lag some GP vaccinations. Dose counts measure throughput, not equitable access, protection, adverse events or the speed of the programme's later phases; the campaign also depended on global supply and changing clinical guidance. HSE has since withdrawn its dashboard archive, so the record cites a Wayback Machine snapshot (10 December 2023) of the original PDF.
๐ฎ๐ชIrelandsince 2009still operatingbears on Institutional responsiveness
Ireland created NAMA in response to the property and banking crisis, acquired distressed loans and worked them down through asset sales and development; by the end of 2023 it had generated โฌ47.7 billion in cash and transferred more than โฌ4.25 billion to the Exchequer while preparing an orderly wind-down.
47,700,000,000 euros, cash generated by nama since inception, 2023-12. 4,250,000,000 euros (over), cumulative transfers to the exchequer, 2023-12. National Asset Management Agency, 2023 year-end review, retrieved 2026-08-31.
The move, and what it needed
Ireland isolated distressed property loans in a specialist agency with a statutory mandate, state-backed financing and commercial asset-management skills. Separating workout decisions from the surviving banks created a long time horizon for sales and development while making the agency's cash and wind-down obligations reportable.
Preconditions
Legislation that gives a crisis agency clear powers to acquire, restructure and dispose of assets
A credible funding and valuation framework that can survive political and market pressure
Transparent reporting and a sunset or wind-down plan so an emergency institution does not become permanent
Where else it has been used: NAMA travels as a crisis-workout pattern: a temporary balance-sheet institution can create room for recovery, but its public return must be read alongside the losses and guarantees that made the intervention possible.
Limits: Cash generated and Exchequer transfers are financial outputs, not a counterfactual estimate of how the crisis would have unfolded without NAMA. Asset sales depend on market conditions, the agency's mandate was exceptional and temporary, and the figures do not capture distributional effects, losses borne elsewhere in the banking system or the full public cost of the rescue.
๐ฎ๐ชIrelandsince 2017still operatingbears on Disaster preparedness and recovery
Ireland's national meteorological service uses a standard YellowโOrangeโRed warning system and briefs the National Emergency Coordination Group so government departments and agencies can coordinate ahead of severe weather; warnings are issued up to 48 hours before anticipated conditions.
Met รireann translates forecast uncertainty into a small, nationally understood set of action levels, then places the forecast inside a whole-of-government coordination forum. That joins technical information to a standing decision process instead of leaving each department to interpret raw weather data independently.
Preconditions
A national meteorological service with authority and credibility to issue one public scale
A cross-government coordination group that can turn warnings into transport, health and local-response actions
Clear thresholds and public guidance that link each colour to a precaution rather than a vague alert
Where else it has been used: The warning pattern travels when a forecast scale is paired with a coordination forum: colour labels alone do not create preparedness unless agencies know who convenes, decides and communicates next.
Limits: Lead time and colour levels describe the warning protocol, not forecast accuracy, public comprehension or whether local authorities reduced harm. The system is hazard-specific and weather-dependent, and a warning is only one input into the National Emergency Coordination Group's operational decisions.
๐ง๐ทBrazilsince 2023still operatingbears on Institutional responsiveness
Brazil's Congress and president established a permanent annual valorizaรงรฃo do salรกrio mรญnimo in Law 14.663/2023, setting the national floor at R$1,320 from May 2023 and linking later adjustments to INPC inflation plus real GDP growth from two years earlier.
Congress codified an INPC-plus-lagged-GDP rule, while the Executive issues the annual decree that sets the floor. Linking a visible wage rule to regularly published national indicators makes annual change routine and contestable instead of an ad hoc negotiation.
Preconditions
Reliable, timely inflation and national-accounts statistics that can anchor the adjustment formula
National labour-law and payroll enforcement that reaches formal employers and protects the statutory floor
Budget coordination because pensions, social benefits and public-sector pay are indexed to the minimum wage
Where else it has been used: Wage indexation travels as a rule-design pattern, not a universal number: countries can combine inflation and growth, but must decide how to protect purchasing power without amplifying fiscal or employment shocks.
Limits: A statutory floor is not the same as wages actually paid or compliance with labour law. The formula does not show real purchasing power, employment, informality, regional cost differences, enforcement or causal effects, and later fiscal-cap legislation altered the practical ceiling on increases. This record documents a national rule for routine adjustment rather than an outcome evaluation.
๐ณ๐ฌNigeriasince 2021operating below its peakbears on Institutional responsiveness
By June 2024, the Central Bank of Nigeria reported 28.46 million eNaira customers onboarded but only 197,041 active wallets, showing that the national CBDC rollout had not converted broad registration into sustained use.
The CBN responded to digital-payment change by launching a sovereign wallet and connecting it to banks, payment providers and identity checks. The gap between customers onboarded and wallets active shows the second task of institutional change: converting a national launch into a habit that survives competing channels.
Preconditions
A central bank with legal authority to issue and regulate a digital form of national currency
Interoperability with banks, payment-service providers and identity databases rather than a standalone app
A use case such as reliable transfers or public payments that gives households and merchants a reason to return
Where else it has been used: CBDC rollouts travel as an adoption problem: technical issuance can be national on day one, while active use depends on incentives, reliability and integration with the payment ecosystem.
Limits: The Central Bank's administrative counts distinguish onboarding from activity but do not explain why wallets became inactive, measure transaction quality or compare the CBDC with Nigeria's much larger private payment rails. The report is the regulator's own account, and the active-wallet figure is a snapshot rather than a causal evaluation of the rollout.
๐ช๐นEthiopiasince 2005still operatingbears on Disaster preparedness and recovery
Ethiopia's government-run Productive Safety Net Programme reached about 8 million people, while 6.8 million received short-term employment in public-works activities that support household income and community resilience.
8,000,000 people, people benefiting from the productive safety net programme, 2026-02-28. 6,800,000 people, people receiving short-term public-works employment, 2026-02-28. World Bank Group, Ethiopia country programme overview, retrieved 2026-08-31.
The move, and what it needed
Ethiopia made drought support predictable through a national safety-net platform that combines transfers for labour-poor households with public works and can scale assistance when shocks arrive. The programme links immediate consumption protection to land, water and community assets rather than treating relief as a one-off distribution.
Preconditions
A household and geographic targeting system that can identify chronic food insecurity before a crisis peaks
Local administrative and public-works capacity able to deliver cash, food or employment across remote districts
Early-warning and financing arrangements that release additional support when drought or other shocks exceed normal programme capacity
Where else it has been used: Shock-responsive safety nets travel when predictable core transfers and scalable emergency finance share one delivery system; a relief fund without that backbone cannot expand quickly when conditions change.
Limits: Beneficiary and public-works counts show administrative reach, not whether transfers arrive on time, food security improves, assets survive drought or households graduate from support. The World Bank summary describes a government programme supported by development partners and does not replace an independent impact evaluation or a complete account of conflict-affected regions.
The Dominican Republic expanded its National 9-1-1 Emergency and Security System to 27 provinces by July 2026, covering 94.64% of the population, or more than 9.6 million people.
The Dominican Republic joined police, health, fire and public-works responders behind one emergency number and expanded the dispatch network province by province. A common protocol and radio system let separate agencies act on the same incident instead of asking residents to coordinate them themselves.
Preconditions
A national dispatch centre with authority to route incidents across police, health, fire and transport agencies
Interoperable communications, location data and response protocols that work across provinces
A financing and expansion plan that adds ambulances, staff and local coverage as the service footprint grows
Where else it has been used: Emergency systems travel when a single public interface is backed by interoperable responders and a measured expansion path; the phone number alone is not the capability.
Limits: Population coverage and provincial integration show reach, not call-answering speed, dispatch accuracy, survival, crime reduction or equitable service quality. The figures come from the presidency's account of the system, and 5.36% of the population remained outside the stated coverage at the reported date.
๐ญ๐ณHondurassince 2015still operatingbears on Disaster preparedness and recovery
Honduras's Sistema Nacional de Emergencias 911 coordinated and attended 762,601 of 788,573 emergency tickets in 2024, reporting 96.71% efficacy through five regional coordination centres.
96.71 % of emergency tickets coordinated and attended, emergency-ticket coordination and attendance efficacy, 2024. 762,601 tickets, emergency tickets coordinated and attended, 2024. Sistema Nacional de Emergencias 911 de Honduras, Memoria Anual 2024, retrieved 2026-08-31.
The move, and what it needed
Honduras put emergency intake, dispatch and inter-agency coordination behind one national 911 service, then operated regional coordination centres to route incidents to the relevant responders. A shared ticket and protocol gives separate agencies a common incident record instead of leaving callers to coordinate them.
Preconditions
A national authority able to route incidents across police, health, fire and other response agencies
Interoperable communications, location data and protocols that work between regional centres
Operating budgets and performance reporting that keep centres staffed and reveal where tickets are lost or delayed
Where else it has been used: Emergency coordination travels when a common intake record is backed by interoperable responders and transparent throughput measures; the number alone cannot substitute for the dispatch network behind it.
Limits: The annual report's ticket counts and efficacy percentage show administrative throughput, not response time, dispatch accuracy, survival, crime reduction or equitable service quality. The figures are the emergency system's own operational accounting, and ticket counts can include calls requiring different agencies or multiple actions; the report does not independently audit the denominator or outcomes.
๐จ๐บCubasince 2021delivered and closedbears on Disaster preparedness and recovery
Cuba's national COVID-19 vaccination campaign had given at least one vaccine dose to 94% of the country's population by 31 December 2021, according to the Pan American Health Organization.
Cuba treated COVID vaccination as a national public-health delivery operation, using its existing primary-care and community health network to move doses from central procurement and production into local clinics. The campaign's reach depended on routine registration and trusted local health workers rather than a stand-alone emergency site.
Preconditions
A nationwide primary-care network with local staff and population registers
Domestic or secured vaccine supply plus cold-chain and scheduling capacity
Public-health surveillance and communication that can identify priority groups and monitor uptake
Where else it has been used: High-coverage emergency vaccination travels when routine primary care, a population register and a dependable supply chain already exist; a campaign launch without those delivery rails cannot reach most households quickly.
Limits: The PAHO coverage figure shows reach of at least one dose, not completion of a primary series, booster uptake, protection against infection, mortality avoided, equity across provinces or the campaign's logistical cost. It is based on reported national data, and the profile does not independently audit Cuba's dose records or distinguish domestically produced vaccines from other supplies.
๐ธ๐ปEl Salvadorsince 2021operating below its peakbears on Institutional responsiveness
El Salvador's 2025 reform made private-sector Bitcoin acceptance voluntary, limited the law to private participation and repealed Articles 4, 8 and 9 of the 2021 Bitcoin Law.
El Salvador launched a national Bitcoin mandate and later used legislation to narrow it: the 2025 reform removed mandatory acceptance, repealed three core articles and confined the legal framework to voluntary private participation. The case makes institutional responsiveness visible as the ability to unwind a national experiment when its obligations no longer fit.
Preconditions
A legislature able to amend the legal basis of a national digital-payment programme
Operational monitoring and fiscal or financing feedback that can expose a gap between launch and sustained use
A transition plan that removes public obligations while leaving private actors a voluntary route to continue
Where else it has been used: Policy reversals travel when governments design an exit path before a national experiment becomes permanent infrastructure; without legal clarity, a rollback can leave users, merchants and public agencies with stranded obligations.
Limits: The amendment records a legal rollback, not actual Chivo use, fiscal cost, user adoption, merchant behaviour or whether voluntary acceptance changed after the reform. The comparison with the original law describes the statutory sectors affected; it does not measure how widely the 2021 mandate was followed. The legislative text also cannot by itself establish implementation timing or operational outcomes.
๐ญ๐นHaitisince 2011operating below its peakbears on Disaster preparedness and recovery
WHO and UNICEF estimated Haiti's national DTP3 routine-immunization coverage at 51% in 2022, down from 68% in 2011, while documenting vaccine and syringe stockouts and weak administrative reporting in the national programme.
Haiti's Expanded Programme on Immunization lost routine coverage as stockouts, insecurity and weak reporting interrupted the chain from national procurement to local clinics. The decline shows how a public-health capability can erode when replenishment, supervision and data feedback fail together, even while the programme remains formally in place.
Preconditions
Reliable vaccine and syringe procurement with buffer stocks that can reach every department
A functioning cold chain and clinic network able to offer repeated doses rather than one campaign visit
Accurate child registers and district reporting that expose missed cohorts quickly enough for outreach and corrective supply
Where else it has been used: Immunization systems can recover through catch-up campaigns and community outreach, but the Haiti case warns that a formal national programme is not durable without stock management, trusted local access and data that decision-makers believe.
Limits: The 51% figure is a WHO/UNICEF estimate with low confidence and is not a direct census of vaccinated children: the profile records poor administrative reporting, national stockouts and disagreement from Haiti's Ministry of Health. The estimate does not show which districts lost coverage, whether children completed doses later, or the causal contribution of insecurity, financing, supply or clinic access; it is evidence of a reported national delivery failure, not a clean impact estimate.
๐จ๐ทCosta Ricasince 2021still operatingbears on Export diversification
Costa Rica's export agency PROCOMER reported that 40 SMEs graduated from its 16-week Ramp Up 2024 programme, delivered with INA and the Sistema de Banca para el Desarrollo, bringing the number of firms benefiting since the programme began in 2021 to 175.
PROCOMER, the national training institute and a development bank combine practical training in finance, strategy, governance and sustainability with export-market connections. The package builds the organisational routines and compliance readiness that small firms need before an overseas buyer or distributor can become repeat business.
Preconditions
An export agency that can identify firms, broker market access and coordinate services rather than offering disconnected workshops
SMEs with a product, management time and enough operational capacity to act on finance, quality and sustainability work
Training, development finance and buyer or standards networks that remain available after graduation
Where else it has been used: Export-readiness programmes travel when capability-building is tied to real market links and finance; a course certificate without buyer validation, working capital or standards support rarely diversifies exports by itself.
Limits: Graduation and reach counts show programme delivery, not verified export growth, new markets, product diversification, export survival or causal impact. PROCOMER describes the firms as strengthening export capability; the source does not establish that all graduates exported or that the programme changed Costa Rica's aggregate concentration.
๐ซ๐ทFrancesince 1975still operatingbears on Long-term unemployment share
France's national statistics office Insee reported that people unemployed for at least one year represented 23.3% of all unemployed people in 2024, while the long-term unemployment rate was 1.7% of the active population, down from 3.1% in 2015.
23.3 % of unemployed, long-term unemployed as share of all unemployed, 2024. 1.7 % of active population, long-term unemployment rate, 2024. Insee, Evolution du chรดmage / Employment Survey 2024, retrieved 2026-08-31.
The move, and what it needed
France's employment system combines unemployment insurance, employment-service counselling and targeted training or hiring support, helping some jobseekers move through the labour market before unemployment becomes persistent. The falling long-term rate is an outcome to investigate, not proof that one programme caused it.
Preconditions
A labour-force survey and employment-service records that distinguish duration consistently
Income support and active labour-market services that preserve contact with jobseekers while they search
Employers, training providers and regional intermediaries able to convert counselling into credible vacancies or skill transitions
Where else it has been used: Long-term-unemployment reductions travel when income support, matching and skills services are connected to real employer demand; a falling rate can otherwise reflect exits from the labour force rather than successful re-employment.
Limits: The figures describe labour-market duration, not the causes of unemployment, the quality of re-employment or whether training and employment services produced the decline. Insee's household-survey definitions and population scope matter, and a low long-term share can coexist with high short-term churn or discouraged workers outside the unemployment measure.
๐ต๐นPortugalsince 2014still operatingbears on Institutional responsiveness
Portugal's digital-government platform reported 4.4 million active Chave Mรณvel Digital credentials and 141 million authentications by 23 October 2025, with 490 entities using Autenticaรงรฃo.gov.
4,400,000 active digital identities, active chave mรณvel digital credentials, 2025-10-23. 141,000,000 authentications, authentications through autenticaรงรฃo.gov, 2025-10-23. Portugal Digital / ARTE, Identidade Digital statistics, retrieved 2026-08-31.
The move, and what it needed
Portugal made one state-certified mobile credential usable for authentication and signatures across public services and participating private services, so each new service can reuse an existing identity rather than build a separate account system.
Preconditions
A trusted civil-identity register and a legal basis for remote signatures
Interoperable authentication APIs and agencies willing to adopt one national gateway
Registration, recovery and offline support for people who cannot rely on a smartphone
Where else it has been used: A shared digital identity travels when legal recognition, cross-agency interoperability and assisted access travel with the credential; a login count alone can hide exclusion or weak service completion.
Limits: These are administrative adoption and usage totals, not measures of service quality, inclusion, authentication success, fraud or trust. An active credential is not necessarily a unique active person, and the government platform reports the totals itself; the figures do not show whether citizens can complete the services they authenticate for.
๐ต๐นPortugalsince 2005still operatingbears on Institutional responsiveness
Portugal's IRN reports 369,447 societies created through Empresa na Hora over its first 20 years; the service is available at more than 200 posts and averages 41 minutes from formation to registration.
A single state counter combines name approval, incorporation and registration, passes the record between public services electronically and later extends the same workflow online, converting a multi-agency legal process into one accountable transaction.
Preconditions
A shared company register and legal authority for one service to complete several acts
Interoperable registries and pre-approved names that remove sequential approvals
A fallback for complex companies and foreign founders whose cases do not fit the standard route
Where else it has been used: One-stop registries travel when back-office data sharing is real and the simple route is measured end to end; a fast counter that leaves tax, licensing and beneficial-ownership steps separate only moves the queue.
Limits: The throughput and elapsed-time figures measure a standard incorporation route, not the full time or cost of becoming operational, survival, access for less standard firms or the quality of subsequent regulation. IRN reports its own service and cumulative count, and the average can hide variation by company type, name choice and whether applicants arrive prepared.
๐ต๐นPortugalsince 2018still operatingbears on Disaster preparedness and recovery
Portugal's integrated rural-fire system reported 85 of 97 national-action projects in progress and 48 percent cumulative execution in 2024; โฌ638 million was spent that year, with 55 percent on prevention, even as 137,651 hectares burned.
After the 2017 fires, Portugal replaced a mainly suppression-led model with a cross-government system linking prevention, landscape management, preparedness, suppression and recovery, coordinated by AGIF and reported against a national action programme.
Preconditions
A permanent coordinating agency with authority to aggregate eleven ministries and dozens of entities
Shared risk data and annual reporting that make prevention visible beside emergency response
Municipalities, landowners and emergency services able to turn national priorities into local work
Where else it has been used: Integrated fire governance travels when prevention budgets, operational command and landowner incentives share one accountable plan; extra firefighting capacity alone does not change the hazard.
Limits: Projects in progress and spending are capacity measures, not proof that communities are safe; the 2024 burned-area and mortality figures show the residual risk plainly. Annual fire outcomes also depend on weather, land use and ignition behaviour, so the record cannot identify the causal contribution of coordination, prevention or suppression spending.
๐ต๐นPortugalsince 2006still operatingbears on Institutional responsiveness
AMA's 2022 activity report records 48 Simplex administrative-simplification measures, 24 PRR and modernisation milestones executed, and 62 innovation, participation and simplification activities across the agency's portfolio.
Portugal pairs a rolling political simplification agenda with a public-sector innovation lab: Simplex names and tracks changes, while LabX researches user needs, co-creates prototypes and tests service redesign before agencies adopt it.
Preconditions
A centre of government able to convene ministries around cross-cutting administrative changes
A safe experimental space with permission to test services with citizens and officials
Owners and monitoring that carry a successful prototype into law, procurement and routine operations
Where else it has been used: A simplification-and-experimentation loop travels when pilots have a route into the rulebook and service budget; a lab without an adoption owner becomes a workshop catalogue.
Limits: Counting measures and workshops shows an administrative improvement pipeline, not reduced burden, faster case resolution or better user outcomes. AMA reports its own activity, and combining the recurring Simplex programme with LabX experiments hides which initiatives reached routine service delivery and which remained pilots. AMA has since been renamed ARTE and no longer hosts the report, so the record cites a Wayback Machine snapshot (6 May 2024) of the original PDF.
๐ต๐นPortugalsince 1965still operatingbears on Disaster preparedness and recovery
Portugal's DGS reported 86 percent immunisation of eligible infants in the first RSV campaign, 98โ99 percent first-year coverage for the universal childhood schedule, and nearly four million seasonal vaccines administered in 2024โ2025.
A national schedule, procurement and registry make routine childhood vaccination a standing service, while seasonal campaigns add new products, broaden delivery through pharmacies and publish coverage by age and municipality so the system can adapt to a changing respiratory threat.
Preconditions
A trusted national immunisation schedule and supply chain
Primary-care, hospital and pharmacy channels able to share records and deliver at scale
Local surveillance and public communication that can identify pockets of low coverage
Where else it has been used: Vaccination systems travel when routine infrastructure can absorb a new campaign quickly and coverage is visible below the national average; a one-off mass purchase without records and delivery channels fades when the emergency does.
Limits: Coverage is not the same as immunity, health impact or equal access, and the rounded seasonal-dose total is not a count of unique people. These figures mainly cover Portugal Continental and the DGS reports the programme's administrative data; high routine coverage can coexist with gaps in older groups, migrants or particular municipalities.
๐ต๐นPortugalsince 2020dismantledbears on Disaster preparedness and recovery
Portugal launched the voluntary STAYAWAY COVID exposure-notification app on 1 September 2020; the European Commission's case study records more than 3.2 million downloads and says the app was later suspended.
Portugal combined a national health authority with research and technology partners to deploy an anonymous Bluetooth exposure-notification app quickly, using voluntary adoption rather than a centralised movement database.
Preconditions
A trusted health authority able to validate positive-test codes
Interoperable laboratory and notification systems that make an alert actionable
Public trust in privacy safeguards and a clear endpoint when the emergency passes
Where else it has been used: Digital emergency tools travel when they connect to routine health workflows and retain legitimacy after the crisis; a download surge without operational integration is easy to dismantle.
Limits: Downloads do not show active use, the share of infected people entering test codes, alerts delivered or epidemiological effect. The voluntary, privacy-preserving design depended on test-result integration and public trust, and suspension after the emergency means the adoption peak was not a durable surveillance capability.
๐ฎ๐ณIndiasince 2021delivered and closedbears on Disaster preparedness and recovery
India used CoWIN to register and scale COVID-19 vaccination nationally, recording more than 2.2067 billion vaccinations and 1.1093 billion registrations by 21 November 2023.
The Union government put registration, appointment, inventory-linked certification and national reporting on one platform, allowing states and vaccination sites to operate against a common record during a fast-moving shock.
Preconditions
A national identity and mobile-number layer capable of handling very large registration loads
Vaccination sites and state health teams connected to a shared reporting protocol
A central operations team able to change workflows as supply and eligibility rules shifted
Where else it has been used: Crisis platforms travel when a common data model links local delivery to central visibility; registrations and doses do not by themselves establish that the system reached the hardest-to-serve groups.
Limits: The counts are platform records, not unique people or a measure of protection, equity, adverse events or health outcomes. The ministry's explainer reports the totals itself and does not provide reconciliation rules, regional access gaps or the counterfactual contribution of CoWIN versus the underlying health and logistics system; the emergency programme has also ended.
South Africa operated Gauteng's e-toll system from December 2013 until the government withdrew the toll declaration effective 11 April 2024, ending urban e-toll charging after roughly 10.4 years.
After motorists resisted the user-pay system and the funding model stalled, a joint transport, finance and provincial process replaced toll revenue with an alternative debt and maintenance arrangement and repurposed the gantries.
Preconditions
A legal authority able to withdraw a toll declaration while keeping the upgraded roads in the national network
A cross-government committee able to negotiate debt, maintenance and replacement funding
Enough public pressure and political attention to reopen a scheme that had already been implemented
Where else it has been used: Responsiveness travels when institutions can change a live programme without abandoning its underlying asset; delayed reversal still leaves a long tail of debt and trust costs.
Limits: The duration is calculated from the ministry's stated December 2013 start and 11 April 2024 end rather than published as a single statistic. Cancellation shows that government eventually changed a rule after resistance and a debt impasse, but it does not show how quickly the response was reached, the road programme's net value or the fiscal cost of replacing toll revenue.
๐ฒ๐ฝMexicosince 2023still operatingbears on Institutional responsiveness
Mexico's medicines and health-products regulator COFEPRIS completed 509,921 procedures entirely online with traceability by 5 September 2024, replacing paper hand-offs for a large share of its regulated users.
The regulator converted submission, tracking and completion into one traceable digital workflow, so applicants could see a case move without physically carrying documents between offices.
Preconditions
A regulator with authority to make its own procedural systems mandatory
Document and identity standards that can be checked electronically
A case-management system able to expose status without exposing protected data
Where else it has been used: Digital regulatory workflows travel when traceability is designed into the case record; moving a form online without a visible status trail leaves the old bottleneck intact.
Limits: COFEPRIS publishes the number of completed digital procedures but not a before-and-after median processing time, rejection rate or compliance outcome. It is a regulator's own operational tally, and the article does not establish that every procedure type or every applicant could use the digital path.
๐ฒ๐ฝMexicosince 2024delivered and closedbears on Disaster preparedness and recovery
Mexico's national civil-protection system coordinated a countrywide earthquake exercise in September 2024 in which 9,727,000 people participated, with alerts carried over 14,490 public loudspeakers and a multi-agency task force of 1.9 million personnel.
The national protection agency set one scenario, aligned federal, state and municipal responders, pushed the same alert through public infrastructure and used the exercise to expose protocol gaps before a real shock.
Preconditions
A legal civil-protection system linking three government levels
An alert network and emergency command structure that can be tested at once
Schools, workplaces and households willing to participate in a common scenario
Where else it has been used: National drills travel when rehearsal is treated as an operational test rather than a ceremony; participation counts need follow-up on the failures found and fixed.
Limits: A drill records mobilisation and protocol testing, not performance in a real disaster. Participation is self-reported by agencies and organisations, and the loudspeaker count covers only listed jurisdictions; the exercise used a hypothetical earthquake and cannot establish casualty reduction or recovery speed.
๐ฒ๐ฝMexicosince 2014dismantledbears on Institutional responsiveness
Mexico abolished the 2014 decree creating PROSPERA on 31 May 2019 and ordered its people, assets and budget to transfer to the new Benito Juรกrez scholarship coordination within 90 days.
The executive used a new coordinating body and a single transition clause to move PROSPERA's staff, resources and education components into a different delivery model rather than preserve the old agency by inertia.
Preconditions
Central budget and administrative authority over the programme's people and funds
A replacement institution ready to receive the transferred functions
A legal instrument that can abrogate the old mandate and specify a deadline
Where else it has been used: Programme replacement travels when a government can name the receiving institution and fund the handover; a deadline on paper is not evidence that service continuity survived it.
Limits: The decree proves a formal policy and organisational change, not that the 90-day transfer happened on time or that beneficiaries experienced a seamless handover. It also does not compare PROSPERA's outcomes with the replacement scholarships, and the transition combined a programme redesign with a change in political administration.
๐น๐ทTurkeysince 2023still operatingbears on Disaster preparedness and recovery
After the February 2023 earthquakes, Tรผrkiye's AFAD established 350 tent cities providing shelter for 2.5 million people, moved about 700,000 people into 414 container cities before the first winter and delivered 46,000 permanent homes within the first year.
AFAD combined a national incident command system with temporary settlements, cash support and a hand-off to permanent housing, keeping shelter, health, education and debris removal in one recovery operation.
Preconditions
A statutory national disaster agency with authority to coordinate ministries and provinces
Pre-existing logistics, procurement and temporary-settlement templates
Land, contractors and financing available for a second phase of permanent housing
Where else it has been used: Recovery systems travel when temporary shelter and permanent rebuilding are designed as one pipeline; a large first response can still leave a long tail if the hand-off is not funded.
Limits: These are government-reported outputs after a catastrophe, so they demonstrate response and recovery logistics more directly than ex ante preparedness. Shelter and home counts do not show safety, quality, displacement duration or whether the response reached every affected household; recovery was still continuing when the report was published.
๐ซ๐ฎFinlandsince 1991still operatingbears on Disaster preparedness and recovery
Finland requires a shelter in every building above a size threshold, nationwide since 1991, and by 2023 had about 50,500 shelters with places for about 4.8 million people, in a country of about 5.6 million.
The state put the cost of protection on construction itself: the Rescue Act obliges owners of large residential and workplace buildings to build a shelter, which peacetime uses as storage, car parks or sports halls on condition it can be cleared within 72 hours.
Preconditions
A threat perception shared across parties since the Second World War
A building code and permit inspection that enforce the duty before construction
Housing built mostly as concrete multi-storey blocks
Where else it has been used: Switzerland runs the same building duty. Germany and Sweden let theirs lapse after 1990 and are now rebuilding from a smaller stock.
Limits: Both numbers are rounded by the ministry, which says the decades-long history makes precise register data hard to compile. Places are unevenly spread: the same release puts Helsinki at 34 percent more places than residents and South Ostrobothnia at places for 48 percent of its population, so the national figure overstates rural protection. A place is a built room, not a tested capability; the ministry's later condition review found most shelters in good order but some equipment needing maintenance. Shelters protect against attack and hazardous releases and say little about recovery from floods or pandemics, which the indicator also asks about. 1991 is when the duty became nationwide; designated municipalities built shelters for decades before.
๐ฉ๐ชGermanysince 1965dismantledbears on Disaster preparedness and recovery
Germany held 1,967 public shelters with about 1.6 million places in 2008, decided with the states to abandon them, and decommissioned them until March 2022, leaving 579 shelters with 477,593 places, none of them functional or operational.
The federal government and the states read the end of the Cold War as the end of the threat, stopped maintenance and budgeted to decommission and release the shelters, until the interior ministry froze the process by decree three weeks after Russia invaded Ukraine.
Preconditions
A peace dividend read as permanent across parties
Shelters owned federally but sited in municipalities with other uses for the land
No statutory floor on protection places that a budget line could not override
Where else it has been used: Sweden and the Baltic states ran down civil defence in the same years and began rebuilding after 2014; Finland and Switzerland kept their building duty and are the comparators the German debate now cites.
Limits: The 2008 figure is rounded by the government and covers only the western states and Berlin: shelters of the former GDR never entered the federal concept, so the stock was never national. Shelter places are an old measure of protection, and the government now argues bunkers cannot meet modern warning times and plans to rely on basements, car parks and metro stations, so the loss of places overstates the loss of preparedness if that concept is delivered. It had not been when this answer was given. The started year is the 1965 Shelter Construction Act; the federal funding programme for public shelters ran from 1968 to 1996. The source is a PDF; the numbers are in the answers to questions one and seven.
๐จ๐ฆCanadasince 1998still operatingbears on Institutional responsiveness
In December 1993 the Chief Actuary projected that the pay-as-you-go contribution rate would reach 14.2% by 2030 and that the fund would run out by 2015. The federal and provincial governments agreed a reform package in 1997. It took effect on 1 January 1998, raised the combined contribution rate in steps to 9.9% by 2003, replaced pay-as-you-go financing with steady-state funding and invested net cash flows through the CPP Investment Board. The 2024 valuation puts the minimum sustaining rate at 9.19% from 2034, below the legislated 9.9%, with assets of $651 billion.
The Chief Actuary's statutory triennial reports put a dated number on the funding gap. Federal and provincial finance ministers then negotiated one package that raised contributions, trimmed benefit growth and moved new money into market investment under an arm's-length board, and wrote into the statute that any future benefit increase must be fully funded.
Preconditions
An actuarial office whose projections both orders of government accept as the shared baseline
A joint amendment rule that forces federal and provincial governments into one negotiated package
Enough time before the projected exhaustion date to phase in a contribution rise
Where else it has been used: The 1997 package required any future benefit increase to be fully funded. The additional CPP in force since 1 January 2019 is fully funded, and the 32nd report sets its own minimum contribution rates on that basis.
Limits: The construct is speed, and this reform was not fast. Four years passed between the December 1993 warning and the amendments taking effect in January 1998, and the rate took until 2003 to reach 9.9%. Whether that counts as responsive depends on a comparison the record cannot supply: it shows a rule changed in time, before the projected 2015 exhaustion, and not that it changed quickly. The headline numbers are a legislated rate and a projection, not an observed outcome. The evidence that the reform held, a 9.19% minimum rate against 9.9%, is the Chief Actuary's model under his own assumptions. That office sits inside OSFI, so this is a federal body assessing a federal-provincial plan. The fix was paid for mostly by workers and employers through a contribution rate up about 70% from 1997, plus slower benefit growth, which a reader may count as a cost of the delay as much as a sign of capability. Quebec runs its own plan, so the CPP covers the rest of Canada. The amendment formula requires the agreement of provinces holding most of the population, and that slows any reform as well as stabilising it.
๐จ๐ณChinasince 2020dismantledbears on Disaster preparedness and recovery
China's State Council joint prevention and control mechanism ran a national COVID-19 containment regime from January 2020, under which the mainland had reported 4,634 COVID-19 deaths when vaccination began on 15 December 2020; it dismantled the regime by circular on 7 December 2022, and the National Health Commission then reported 59,938 COVID-related deaths in medical institutions between 8 December 2022 and 12 January 2023.
The State Council joint prevention and control mechanism used mass testing, health codes and area lockdowns to hold infections near zero for nearly three years, then withdrew the whole apparatus by one circular within weeks, before raising vaccination coverage among the oldest residents. The capacity to suppress was built; the capacity to exit was not prepared in parallel.
Preconditions
A command structure able to impose testing, movement codes and lockdowns uniformly across provinces
A population and economy able to absorb repeated closures, which wore out by late 2022
An exit plan, above all high vaccination of the oldest cohorts and hospital surge capacity, which was missing when the regime was lifted
Where else it has been used: Other zero-COVID jurisdictions (New Zealand, Australia, Taiwan) exited after vaccinating their elderly first; the Chinese exit is the case where suppression ended before that precondition was met.
Limits: Both numbers are the government's own and both are made under counting rules that understate deaths: the 59,938 covers only deaths inside medical institutions, and independent estimates put excess deaths far above the official count, so the official figure is a floor, not the toll. The two metrics cover different periods, populations at risk and counting rules (a cumulative count over 11 months against a five-week in-hospital count), so their ratio is not a mortality comparison. The SCIO text carrying 4,634 is an advocacy piece republished by the State Council Information Office, not a statistical release. The record does not show the costs of containment itself (lockdowns, lost output, mental-health and access-to-care harms), which were part of why it ended, nor that the dismantling followed Omicron's transmissibility rather than a planned exit. What it shows about preparedness is the exit gap: the State Council portal reported on 30 November 2022 that only 65.8 percent of people aged 80 and over had completed primary vaccination (english.www.gov.cn/statecouncil/ministries/202211/30/content_WS6386aabec6d0a757729e3d55.html), a week before the regime was lifted, and 90.1 percent of the reported deaths were aged 65 and above. The construct fit is contestable: the record evidences a national shock-containment capacity and its abrupt removal, not a standing disaster-recovery system. No official series of excess mortality for 2022-2023 has been published that would let the reversal be sized properly.
๐ณ๐ฌNigeriasince 2014delivered and closedbears on Disaster preparedness and recovery
After an infected traveller brought Ebola virus into Lagos on 20 July 2014, Nigeria's government set up an emergency operations centre with WHO country office support, traced contacts in Lagos and Port Harcourt, and held the outbreak to 19 cases and 7 deaths before WHO declared the country free of transmission on 20 October 2014.
The Federal Ministry of Health and Lagos State stood up an emergency operations centre within days of confirmation, repurposed the polio programme's staff and GPS tracking for contact tracing, and funded isolation and treatment centres quickly.
Preconditions
An existing surveillance and field workforce, here the polio eradication infrastructure, that could be redirected overnight
A domestic laboratory able to confirm the virus within days
Political authority to release emergency funds and coordinate federal and state agencies under one command
Where else it has been used: Nigeria's emergency operations centre model was formalised in the Nigeria Centre for Disease Control and cited by WHO as the preparedness standard for at-risk countries; the transfer depends on having a pre-existing field programme to repurpose.
Limits: This is one outbreak seeded by a single imported chain in two cities, so it shows a response to one shock and not a standing national capacity. The page is a WHO narrative, not a dataset, and WHO and partners supported the response, which also leaned on emergency-operations staff, GPS tools and infrastructure and technologies in place for polio eradication, as the WHO page puts it. Nothing here shows Nigeria's performance in later shocks such as Lassa fever, cholera, floods or COVID-19, and a counterfactual of how far the virus would have spread is not measured.
๐ณ๐ฎNicaraguasince 2000still operatingbears on Disaster preparedness and recovery
Nicaragua's national disaster system, SINAPRED, evacuated 160,597 people and sheltered 59,964 ahead of hurricanes Eta and Iota in November 2020, when about 3 million people were exposed, and 21 lives were reported lost.
Nicaragua's government built a statutory disaster system under Law 337 that runs a national operations committee and a technical arm (INETER), and used it in 2020 to move about 160,000 people out of the path of two major hurricanes within two weeks.
Preconditions
A statutory chain of command from national committee to municipal and community brigades
Forecasting and early warning that give days of lead time before landfall
A centralised state able to order and enforce evacuation, which in Nicaragua's case comes with weak independent oversight of its reported results
Where else it has been used: The same sequence of statutory system, community brigades and pre-landfall evacuation underlies Cuba's hurricane record; where the state is less centralised, the evacuation step depends more on trust in warnings than on command.
Limits: The World Bank document is an appraisal-stage project information note written days after landfall (dated 23 November 2020), and its evacuation, shelter and death figures are the Nicaraguan government's own preliminary counts relayed by the Bank, not an independent verification; the Nicaraguan government's casualty reporting has been disputed in other emergencies, notably COVID-19. The figures show people moved and lives reported lost, not how many people at risk were not reached, the conditions in shelters during a pandemic, or the speed and quality of recovery: the same document puts damage and losses at USD 738.6 million, about 6.2 percent of GDP. Low deaths also depend on where the storms struck (the sparsely populated North Caribbean Coast) and on forecast lead time from regional meteorological centres, so the count is not a clean measure of SINAPRED alone. The 1994-2013 average of 160 hydrometeorological deaths a year is a different set of hazards and is context, not a counterfactual. SINAPRED's own site did not open from here (connection timeout), so the government primary source could not be checked. One hurricane season is a single test.
๐ญ๐นHaitisince 2013operating below its peakbears on Disaster preparedness and recovery
After Haiti's Ministry of Public Health (MSPP) and water directorate DINEPA launched the National Plan for the Elimination of Cholera in 2013, the country reported no confirmed cholera case from January 2019 to January 2022, but transmission returned in October 2022 and the MSPP epidemiology directorate had reported 76,675 suspected cases in all 10 departments by 18 December 2023.
Haiti's MSPP and DINEPA pursued elimination through case surveillance, case-area rapid response teams and targeted vaccination, with partner financing, and drove confirmed transmission to zero. When insecurity, displacement and water-system collapse returned in 2022, the response chain broke and cholera spread back to every department.
Preconditions
Laboratory-confirmed surveillance with field teams able to reach every reported case within days
Safe water and sanitation coverage that removes the transmission route, not only emergency treatment
Enough security and sustained financing to keep response teams and supplies moving after donor attention fades
Where else it has been used: Case-area targeted response has been copied in other cholera-endemic settings, and Haiti shows it can push transmission to zero without fixing the water system, which leaves the result reversible.
Limits: The two numbers are in different units: a three-year run with zero confirmed cases is set against a cumulative count of suspected cases, so the record shows a loss of the elimination outcome, not a measured fall in a capability. Part of the apparent elimination may be under-detection; WHO reports national figures and does not audit them. Suspected-case counts depend on clinical definitions and reporting under gang violence that cut access to clinics and labs, so the 76,675 figure is likely an undercount and not comparable with the 2010 to 2019 series. The re-emergence was driven mainly by insecurity, displacement and the collapse of water supply, which the elimination plan's water and sanitation targets never fixed; it is weak evidence about the surveillance and rapid-response capacity itself. Much of the response (rapid response teams, oral vaccine, water investments) was financed and technically run with PAHO, UNICEF, CDC and NGOs, so attribution to Haitian institutions alone is partial. The started year, 2013, is the year the plan was announced and funding requested, since the PAHO page is a funding call, and not a dated start of delivery; control activities had run since the October 2010 outbreak. The secondMetric comes from a separate WHO page.
๐จ๐บCubasince 1997still operatingbears on Disaster preparedness and recovery
Cuba's Civil Defence system and the provincial and municipal Defence Councils it activates evacuated and protected more than 735,000 people across eastern Cuba ahead of Hurricane Melissa, which struck as a Category 3 storm on 29 October 2025, according to UN OCHA.
Cuba's Civil Defence, under the armed forces ministry, activates provincial and municipal Defence Councils on a meteorological alert, and they move residents of low-lying and fragile housing to shelters or relatives' homes and secure crops and assets before landfall. The phased alert system set in law lets local authorities order evacuation without waiting for national instruction.
Preconditions
A trusted national meteorological service whose forecasts trigger pre-agreed alert phases
Local government and mass organisations that know every household and can order and organise movement block by block
Annual drills and a population that expects to evacuate, so compliance is routine rather than negotiated
Where else it has been used: Caribbean and Central American governments have studied the model through UNDP and regional disaster agencies; the early-warning and phased-alert parts travel, while the compulsory, block-level mobilisation depends on Cuba's political organisation.
Limits: The 735,000 figure is the Cuban authorities' count relayed by OCHA, not independently verified, and it counts people moved, not lives saved: no official source located today states the death toll avoided, and the claim does not assert one.The system is military-led and rests on compulsory mobilisation through a one-party state's local structures, which may not transfer. Preparation is the strong half; recovery is weak: two weeks after landfall 53,000 people remained evacuated, nearly 150,000 homes were affected and nearly 60 per cent of Santiago de Cuba province had no electricity. The started year is the decree that set the current system of measures; the Civil Defence itself dates from the 1960s, which no official page opened today states with a year. The page is a UN news story, not a dataset, and a single event is one observation.
๐จ๐บCubasince 2021operating below its peakbears on Institutional responsiveness
On 1 January 2021 Cuba's government replaced its dual-currency system with a single exchange rate of 24 pesos per dollar for the whole economy. By 1 October 2026 the Central Bank of Cuba published three official dollar rates, of 24, 120 and 689 pesos.
The Council of State unified the currency by decree on a fixed date at a fixed rate. When reserves and supply did not support 24 pesos per dollar, the Central Bank layered new official segments, a 120 rate for individuals and then a daily floating rate for exporters, on top of the old rate rather than moving it.
Preconditions
Hard-currency reserves able to defend a fixed rate, which Cuba lacked
A state wholesale sector whose prices and contracts were written at 24 and would be repriced by any move
A tolerated informal currency market that kept setting the price the official rate ignored
Where else it has been used: Venezuela and Argentina have run the same layered-rate sequence after a unification; in each, the new segments were added as exceptions before the old rate was abandoned.
Limits: The API fields are named tasaOficial, tasaPublica and tasaEspecial. We map them to Segments I, II and III because the bank's rate page shows the same 689.00 figure under its three-segment tabs. The API does not state that mapping, so it is our reading. The API history starts in December 2025, at 410 for the floating segment, so the official series does not document when the single rate first split. A 120-peso rate for individuals existed by 2025, but this record cannot date it from an official source. The unification's other half, withdrawing the convertible peso (CUC), stands and is not counted as lost. A hostile reviewer would say the record shows a monetary outcome, not the speed of rule-making. The case for filing it here is that the state held a fixed rate while the informal market moved away from it, then added segments instead of adjusting the one rate. The record does not show the informal rate, inflation or shortages. The Gaceta's website lists Decreto-Ley 17/2020 as 'Derogada' without saying what replaced it. The MEP page reprints a state-media article and sometimes fails to load.
๐จ๐ญSwitzerlandsince 2020delivered and closedbears on Institutional responsiveness
The Swiss Federal Council adopted an emergency ordinance on 25 March 2020 and, through 125 banks including PostFinance under federal guarantee, granted 137,870 Covid-19 credits worth CHF 16.9 billion between 26 March and 31 July 2020, with 76,034 applications submitted in the first eight days.
The Federal Council wrote an emergency ordinance in 12 days that let firms self-declare eligibility for a bank loan of up to 10 percent of turnover, which recognised guarantee cooperatives backed 100 percent with federal money, so the banks' existing client relationships and digital channels did the distribution within hours of a request.
Preconditions
A constitutional emergency power that lets the executive legislate within days, with parliament ratifying afterwards
Existing guarantee cooperatives and a dense banking network that already knew every small firm's turnover
A one-form online application and acceptance of fraud risk, policed after payout rather than before
Where else it has been used: Germany, Austria, France and Italy ran state-guaranteed loan schemes at the same time; the Federal Council's own comparison credits the Swiss speed to the digital application step and the public-private set-up rather than to the size of the guarantee.
Limits: The source is the Federal Council evaluating its own programme, and the figures come from the programme's own platform (covid19.easygov.swiss). The construct is the speed at which rules and programmes change; the metric counts credits granted, which measures uptake, and the speed is carried by the dates (ordinance prepared in 12 days, adopted 25 March, credits from 26 March) rather than by a timestamp series comparable across countries. The speed was bought by suspending the normal process: the ordinance rested on the Federal Council's emergency power under Art. 185(3) of the Constitution, and credits were granted on a one-sided self-declaration with only rudimentary checks. The same report puts the upper bound of abusive applications at 5.5 percent of credits if every pending proceeding ended in conviction, with 416 convictions (0.3 percent) by mid-June 2023, and 51 percent of the credit volume (CHF 8.6 billion) still outstanding then, so the final loss is not known; the 0.37 percent loss rate is as of mid-2023. Delivery ran through private banks, so the speed is shared credit between the state and the banking network, and the number does not show how many firms survived because of the credit. The PDF's own prose rounds the first-eight-day figure to 'rund 76 000 gewรคhrten' credits, while the 76,034 is stated as applications submitted ('Antrรคge ... eingereicht'). The source is a 108-page PDF; the figures sit in the summary, Table 1 and section 3.3.
๐จ๐ญSwitzerlandsince 1963still operatingbears on Disaster preparedness and recovery
Under the shelter-building duty introduced with Switzerland's 1963 civil protection legislation, the Federal Office for Civil Protection reported about nine million shelter places in some 370,000 private and public shelters in 2023, a national coverage rate above 100 percent of the resident population.
Federal law obliges anyone building housing in a commune short of places to build, equip and maintain a shelter, or pay a replacement contribution (CHF 1,400 per place today) that communes spend on public shelters; cantons keep an assignment plan that allocates every resident a place.
Preconditions
Sixty years of building permits passing through one federal rule, so the stock accumulates with the housing stock
Communes and cantons with the registers to assign residents to places and update the plan
A political consensus, from the Cold War on, that civil defence is a cost property owners carry
Where else it has been used: Finland runs a similar duty and keeps it operating; Germany designated public shelters and then dropped them after 2007, which is the reversal this record should be read against.
Limits: The construct asks for demonstrated capacity to prepare for and recover from major shocks; shelters are preparation for armed conflict and have never been used for that purpose, so the record shows a stock built, not a capacity demonstrated. Both figures are rounded ('rund', 'ca.') on an information page, not a statistical table, and the page is from August 2023; the July 2026 page repeats the 370,000 shelters but not the place count, and describes the goal of one place per resident as 'nahezu erreicht', nearly reached rather than exceeded. 'Over 100 percent' is a national aggregate: the publisher itself says inner cities and peripheral regions still lack places and growing areas need more, so coverage where people live is lower than the headline. The basis is the permanent resident population, so tourists and short-stay residents are not counted. Most places are in private basements used for storage in peacetime and need preparation time before occupancy, and much of the stock dates from the Cold War. The 1963 start is the civil protection act; the separate building-measures act and the construction of the stock followed over decades, and the cost was borne largely by property owners rather than the state.
๐ป๐ชVenezuelasince 2021delivered and closedbears on Disaster preparedness and recovery
Venezuela's health authorities reported to WHO that 22,157,232 people, 78 per 100 population, had received at least one COVID-19 vaccine dose by 29 December 2023, from a first dose given on 22 February 2021, with 14,287,370 (50 per 100) completing the primary series.
Venezuela's health ministry ran a mass COVID-19 campaign through its public clinic network and mass vaccination points, and reached a majority of the reported population with a first dose within three years.
Preconditions
A public clinic and community network able to run mass vaccination points nationwide
Population registers good enough to count who received a second dose, which the drop-off suggests were weak
A supply of doses large enough for a national campaign, however obtained
Where else it has been used: Cuba (cub-covid-vaccination) ran a campaign on domestic vaccines with a much higher completion rate; the Venezuelan case shows the first-dose reach a centralised network can deliver and the follow-up it cannot.
Limits: The figures are Venezuela's own reports to WHO (DATA_SOURCE 'REPORTING'), with no independent coverage survey behind them, from a government whose routine immunization data WHO and UNICEF grade as lacking accepted empirical support (see ven-pai). The per-100 rates divide by a population estimate that is highly uncertain after the emigration of millions of Venezuelans, so the 78 and 50 could be off in either direction. The 28-point gap between first dose and completed series, and boosters at 2 per 100, show a campaign that stalled after first doses; the record is a partial delivery, not full protection. The VACCINES_USED field is empty, so the record cannot show which products were used or whether each held WHO emergency use listing, and the source does not show where the doses came from, so the record evidences distribution only. Vaccination is a response to one shock, not proof of general preparedness, and the file was last updated in December 2023. The source is a CSV with no prose; the numbers are read from one row.
๐ฆ๐บAustraliasince 2020delivered and closedbears on Institutional responsiveness
Australia's Treasury designed, Parliament legislated and the tax office began paying the JobKeeper wage subsidy just over five weeks after its announcement on 30 March 2020, and in its first phase it supported on average around 925,000 businesses and 3.6 million individuals each month before ending on 28 March 2021.
Treasury designed a flat-rate wage subsidy and the Australian Taxation Office paid it through the single-touch payroll and business registration systems it already ran, so Parliament could legislate eligibility and money reached employers within about five weeks.
Preconditions
A tax office already holding near-real-time payroll and turnover data for most employers
A Parliament able to pass emergency legislation within days with bipartisan support
Fiscal room to accept large deadweight in exchange for speed
Where else it has been used: Germany's Kurzarbeit, the UK furlough scheme and New Zealand's wage subsidy did the same job in the same months; the schemes that already existed in law, like Kurzarbeit, paid sooner, and the new ones depended on how much payroll data the tax authority already held.
Limits: Treasury is evaluating its own programme, and the five-week figure is a statement in prose, not a timestamped series. Speed was bought by routing payment through existing tax-office systems and a flat $1,500 fortnightly rate, which shipped known errors: the same report finds around $8.9 billion went to businesses whose turnover rose, and the 2023 Independent Evaluation of the JobKeeper Payment (https://treasury.gov.au/sites/default/files/2023-10/p2023-455038.pdf) records that around 2 million short-term casual and temporary-visa employees were excluded, that the first-phase cost estimate of $130 billion, which a reporting error had inflated, was revised down to $70 billion, and that JobKeeper was announced 'later than job retention schemes in comparable countries'. The total cost was $88.8 billion. One emergency response shows the state can change programmes quickly under extreme pressure; it does not show routine responsiveness of rules in normal times, which is what the indicator's construct mostly asks about. Coverage counts are monthly averages from tax-office payment data, not unique people.
Announced on 25 March 2020 and paying from April, the Dominican government's Quรฉdate en Casa transfer was built on the existing SIUBEN registry and PROSOLI payment system and, by the president's report to the Senate of 10 July 2020, reached more than 1.6 million households, 780,000 of them new to social assistance.
The social policy cabinet (GCPS) chaired by the Vice-Presidency raised the transfer for 811,003 households already on PROSOLI and added new households classified as poor or vulnerable in SIUBEN, while ADESS partnered with banks and card networks and signed on more than 2,000 new retailers to pay them.
Preconditions
A proxy-means social registry (SIUBEN) with recent survey data to target new households without new fieldwork
A working electronic payment administrator (ADESS) and a retailer network that could be extended
Emergency fiscal room
Where else it has been used: Most Latin American governments scaled transfers through their social registries in 2020; the countries with registries covering a larger share of the population needed fewer on-demand fixes.
Limits: The headline count is the government's own figure, relayed by the World Bank, and the same footnote says different sources and timelines report different figures: the Vice-Presidency announced 899,354 existing PROSOLI households plus 600,646 temporary ones on 2 April 2020. The report was written in 2020 and its spending figures through December are projections. Speed came from infrastructure built long before: SIUBEN identified the households and PROSOLI and ADESS paid them, so the record shows a registry being used under pressure rather than a rule changed from nothing, and SIUBEN covered only about 60 percent of the population, which is why on-demand applications (about one million claims, 70,000 admitted, per the Vice-President) were needed. The programme was temporary by design and was extended by decree to December 2020 and then phased out; nothing here shows the responsiveness carried over to later shocks. Coverage is households paid, not poverty averted.
๐ช๐ชEstoniasince 2021still operatingbears on Disaster preparedness and recovery
After COVID-19 exposed supply gaps, Estonia set up a state stockpiling company in 2021 that holds food, healthcare, veterinary and fuel stocks. By the National Audit Office's January 2026 review, its food stocks covered 10% of the population for 30 days, against a government target of 14 days for the whole population, and just over 90% of the required pharmaceutical stock was in place.
The government turned its fuel-reserve agency (OSPA) into a single stockpiling company under the economy ministry and gave it food, medicines, protective equipment and veterinary stocks. Responsibility had been spread across ministries, which COVID-19 showed left the country buying supplies in the middle of an emergency.
Preconditions
An existing fuel-stock agency with procurement, rotation and storage contracts to extend
A domestic food market large enough to hold delegated stocks, which the agency says is the binding limit here
Ministries that agree in advance on who releases a stock and how it reaches people, which is the step still missing for food
Where else it has been used: The 2020 announcement names Finland's centralised security-of-supply model as the example. Estonia copied the single agency but not yet the scale, because its market is too small to hold Finnish-style delegated stocks.
Limits: This is evidence that capacity was built and also that it falls short. The audit itself says it could not judge whether the food stocks are sufficient. Our arithmetic, not the auditor's: 10% of the population for 30 days is about three person-days per head, roughly a fifth of the 14-day target. Only the summary is public, because the full report is restricted for five years, so the underlying volumes and methods cannot be checked. About two-thirds of the food stock is flour, and the contract partner holds half of that as grain, which is raw material. The stocks are delegated, held by contracted firms rather than the state. The agency set its own lower target, the ministries accepted it, and the government target has not been formally changed, so it is unclear which target the system answers to. Release has been rehearsed only for pharmaceuticals, which have been drawn three times. The release of food, veterinary products and protective equipment has not been agreed or practised: the veterinary release exercise was postponed from autumn 2025 to March 2026, food release is first to be practised in the Ilves 2026 national exercise, and some release guidance is due only by the end of 2027. Nothing here shows performance in a real shock. The status is operating rather than eroded because no peak was lost: the stock never reached its target. The founding year rests on the agency's own site and a 2020 ministry press release.
๐ฌ๐นGuatemalasince 2020delivered and closedbears on Institutional responsiveness
Guatemala's Ministry of Social Development (MIDES) made the first Bono Familia payments in May 2020, two months after the pandemic emergency was declared. By December 2020 the programme had reached 2.7 million beneficiaries, 21 times the 128,036 beneficiaries of the regular Bono Social cash transfer in 2019.
A law created a Q6 billion emergency fund and MIDES, which had no social registry, used electricity distributors' customer records to identify households, telecom operators to register and notify them, and banks' virtual accounts and ATMs to pay them. Its caseload grew 21 times in eight months.
Preconditions
Near-universal electricity metering, so a utility customer list could stand in for a social registry
A banking network willing to open virtual accounts at scale and pay out through ATMs and retail points
A congressional appropriation passed within weeks of the emergency declaration
Where else it has been used: Utility and telecom data used as a stand-in registry appeared across Latin America in 2020. The usual lesson is that the stand-in is retired after the shock and the regular programme returns to its small base; the brochure still calls for a national social registry, which suggests the stand-in was not institutionalised here.
Limits: The publishers are Guatemala's development partners (the World Bank lent to its 2020 crisis response) and this is a lessons-learned brochure, not an independent evaluation, and no Contralorรญa audit was consulted. Beneficiaries are registrations tied mostly to electricity meters, not verified unique people or households. Eligibility ran through electricity consumption, so households without power came in only through special cases capped at 10% (143,014 special cases). The 21-times figure compares a three-payment emergency transfer with a conditional programme of a different design, so it measures scale-up, not like-for-like coverage. The poverty effect is a World Bank simulation. The responsiveness did not last. The same document says cash transfer programmes shrank between 2012 and 2021, except in 2020. Bono Social reached only 152,702 households in 2018 at 0.05% of GDP, and the brochure lists open challenges in carrying the payment model into regular programmes. One episode of fast emergency scale-up does not show that rules and programmes in general change quickly.
๐ซ๐ทFrancesince 2005operating below its peakbears on Disaster preparedness and recovery
France's health ministry and its stockpile agency built a strategic pandemic stock of about 1.7 billion FFP2 and surgical masks by 2009, but by the start of the covid-19 crisis it held only about 100 million valid masks, according to the Senate commission of inquiry.
The health ministry and EPRUS, later Santรฉ publique France, built a national stock of nearly two billion masks after the H5N1 scare. The directorate-general for health then let expiring and non-conforming lots go without replacement. It ordered 613 million surgical masks destroyed in 2018 and did not renew them, so the stock was close to empty when covid-19 arrived.
Preconditions
A dedicated stockpile agency with a budget line that survives years without a crisis
A replacement schedule tied to expiry dates and to changes in product standards, not to annual budget pressure
Political ownership of a stock target, so that a quiet downgrade from 'stock' to 'need' is visible and has to be defended
Where else it has been used: Most OECD countries rebuilt their medical stockpiles after 2020. The French case shows that procurement is not the hard part. The hard part is keeping the stock in date, and now the crisis-era stock faces its own expiry in 2026.
Limits: The status is the contestable part. The collapse from 2009 to 2019 is well documented, but the state rebuilt the stock during and after the crisis: Santรฉ publique France's January 2024 inventory, quoted in a 2025 Senate bill, counts 2.1 billion masks, more than the 2009 figure. The case for 'eroded' rests on two things. About 706 million of those masks had already expired, which leaves roughly 1.39 billion in date, a subtraction we made and the source does not publish. The same text says every mask bought during the crisis expires in 2026. The 2009 figure is itself overstated as capability, because the 613 million surgical masks bought in 2005 and 2006 later failed a 2014 standard. Measured in conforming masks, today's stock may exceed the 2009 one, so a reviewer could fairly argue for 'operating'. The Cour des comptes chapter on mask stock management (RALFSS, May 2025) did not open from this machine, because the host refused the connection, so its figures are cited only through the Senate text. The Senate figures were compiled by the commission from Santรฉ publique France data and are given in round numbers. Mask counts measure one input to pandemic readiness. They say nothing about distribution capacity, domestic production or the rest of the stock: drugs, gowns, gloves.
The UAE's health authorities ran a national COVID-19 vaccination campaign from late 2020 that WHO records as having given at least one dose to 9,991,089 people, a first-dose coverage of 100 per 100 people on WHO's capped measure, and 24,922,054 doses in total by 6 July 2022.
Federal and emirate health authorities registered vaccines early, supplied them free to citizens and residents through public and private clinics, drive-throughs and mass centres, and tied access to daily life to vaccination status through the AlHosn app.
Preconditions
A small, concentrated population with nearly universal digital identity and smartphone use
Fiscal room to buy doses early from several suppliers, including through a clinical-trial partnership
Authority to make vaccination status a condition of entry to workplaces and public venues
Where else it has been used: Fast campaigns in small, wealthy states such as Israel and the UAE were widely cited in 2021. The digital-pass part travelled more readily than the procurement part, which depended on early bilateral deals.
Limits: The figures are the UAE's own reports to WHO, not an independent count, and the last update is July 2022. The per-100 share is capped at 100 by WHO and the UAE's population denominator is unusually uncertain because most residents are foreign nationals on visas, so a 100 percent first-dose coverage claim cannot be taken literally; the absolute counts are safer, and they may include doses given to short-stay visitors. The start-date evidence is the Abu Dhabi emirate's media office describing its own emirate, which also says vaccines reached frontline workers from September 2020, so 2020 is the start of use, not of a public national campaign on a fixed date; the federal health ministry's site did not open from here. Coverage is an output of a shock response. It does not show deaths averted, cold-chain performance, equity across nationalities and labour camps, or preparedness for the next shock.
๐ต๐ฆPanamasince 2020delivered and closedbears on Institutional responsiveness
Panama's government created the Plan Panamรก Solidario digital voucher during the 2020 COVID-19 closures, reaching 90,492 people in its first month (April 2020), 458,662 in May and a peak of 949,343 in November 2020, then tapered it as employment returned and ended it in 2023.
The national government, through MIDES and the Autoridad de Innovaciรณn Gubernamental, built a digital voucher tied to the national identity number and paid merchants directly, so it could switch on a nationwide transfer within weeks of the closures and switch it off gradually as people went back to work.
Preconditions
A national identity number usable as the key for eligibility and payment
A digital-government agency able to build and run a registration and recharge platform at national scale
A merchant network willing to accept vouchers and a treasury able to pay them promptly
Where else it has been used: Pandemic voucher schemes ran across Latin America in 2020; Panama's shows that an exit rule (re-screening through a social file as employment returned) is part of a responsive programme, not an afterthought.
Limits: The figures are the ministry's own management report at the end of a government's term, sourced to the digital-government authority's platform, and are not audited. Monthly beneficiary counts show how fast the programme scaled, not whether it reached the right people: MIDES itself introduced the Ficha Social later to screen out people who did not need it, which implies early targeting was loose. The table counts people receiving a recharge in a month, and the report's own headline of about 1.3 million people ever covered is a different, cumulative measure. The voucher was a temporary emergency measure ended by design, so a later case with no comparable shock cannot be read from it. Speed came partly from an emergency mandate and from spending about B/.2.08 billion paid to merchants between 2020 and March 2024 (Cuadro No. 15), which says nothing about fiscal sustainability. The figures sit in a PDF table, and the server rejects requests without a browser-like user agent, so checking takes manual work.
๐น๐ญThailandsince 2020operating below its peakbears on Disaster preparedness and recovery
Thailand's State Audit Office found in 2024 that the public warning system, for which the state allocated THB 1,074.95 million in fiscal years 2020 to 2022, and whose telemetry the Department of Disaster Prevention and Mitigation publishes through its DPM Alert app, was not working as intended: 318 of 555 telemetry stations (57.30%) failed to measure continuously for more than two days, and 54 of 82 sampled warning towers and receivers (65.85%) were not ready to sound an alert, which the auditor attributed to damaged or deteriorated key parts.
Thailand's disaster department funded and installed warning towers, telemetry and a public app, but left maintenance after warranty, data integration and provincial alerting duties without owners, so the installed network decayed into equipment that was present but not ready to sound.
Preconditions
A maintenance budget and named custodians for every device after the supplier warranty ends
A routine national test that records which towers actually sound and feeds repairs
Provincial and local units obliged to issue and log alerts within the warning window, with someone checking they did
Where else it has been used: Siren and telemetry networks erode the same way elsewhere once the post-disaster budget passes. Thailand's audit is unusual for counting the decay device by device, which is what made the later shift to cell broadcast an evidence-based choice rather than a fashion.
Limits: The page is hosted on the State Audit Office's own site, but it is a newspaper clipping (Siang Samphan, dated 28 March 2024) of the office's release, not the full audit report; the report itself was not opened. The release names no agency; attributing the network to DPM rests on the DPM Alert app. Part of the telemetry was still under the supplier's defect warranty, so its 57.30% failure rate is underperformance of new equipment, not decay; the erosion is carried by the tower test. The device test covers 82 devices in seven provinces chosen by the auditors, not a national random sample, so 65.85% is not a national failure rate. The 'peak' in secondMetric is the installed size of the telemetry network, not an observed year in which all 555 stations worked: the audit does not say when, or whether, the network last ran at full function, so the erosion is from design capacity rather than a measured earlier high. The started year is the first fiscal year of the audited budget window. The audit tests readiness of equipment, not outcomes in a real disaster, and it predates the September 2025 T-Alert cell-broadcast rollout recorded separately as tha-t-alert, which may have replaced part of the function. The finding of THB 514 million in poorly spent funds is the auditor's judgement of value for money.
๐ฆ๐ทArgentinasince 2011still operatingbears on Disaster preparedness and recovery
The national water resources secretariat contracted the state technology firm INVAP in 2011 to build Argentine weather radars, and by December 2024 the network operated 14 radars feeding an operations centre at the national weather service, after 11 had covered 40 percent of the territory together with three INTA radars in 2023.
The water resources secretariat bought radars from INVAP, the state-owned engineering firm, instead of importing them, and later contracted INVAP to operate and maintain the network while the national weather service ran a central operations centre.
Preconditions
A domestic firm with prior radar and complex-systems experience (INVAP)
A ministry willing to fund a multi-stage build across changes of government
Provinces supplying sites, power and fibre under cooperation agreements
Where else it has been used: Most countries import weather radars; Argentina's choice to build them domestically is the unusual part, and the slow third stage shows the cost of depending on one national supplier and one budget line.
Limits: A radar count is an input to early warning, not demonstrated capacity to prepare for and recover from a shock, which is what the indicator asks for; nothing here shows that warnings reached people or reduced losses. Both figures are the government's own announcements, and the count is inconsistent across them: 11 radars plus a prototype in March 2023 and 14 in December 2024, so it is unclear whether the prototype is counted. The 2011 start is a contract signature, and the sources do not date the first operational radar. The March 2023 announcement announced a third stage of 10 more radars (RMA12 to RMA21) for 70 percent coverage, and by December 2024 only about three had been added. Later installations in 2025 were reported but not verified here at an official source. The 40 percent coverage figure includes three radars owned by INTA, outside the network.
๐ท๐ผRwandasince 2024delivered and closedbears on Disaster preparedness and recovery
Rwanda's Ministry of Health declared a Marburg virus disease outbreak on 27 September 2024 and declared it over on 20 December 2024. WHO reported 66 confirmed cases and 15 deaths, a case fatality ratio of 23 percent, with the last confirmed case on 30 October.
Rwanda's Ministry of Health and Rwanda Biomedical Centre ran an incident-management structure with WHO and partner surge support. They traced and tested contacts at scale, with 7,874 Marburg tests by 6 December. Cases were moved to a dedicated treatment centre with critical care, and the WHO 42-day rule was followed before the end was declared.
Preconditions
A national public health institute with in-country filovirus PCR capacity
An administrative system able to trace contacts quickly in a dense capital
Standing relationships with WHO, Africa CDC and US agencies that could deploy surge teams within days
Where else it has been used: The Ebola response model from West Africa and the DRC, adapted to a mostly urban, hospital-centred outbreak in a capital city.
Limits: This is one 12-week event, not a standing programme, and a single outcome cannot separate preparedness from luck or case mix. The low fatality ratio is partly a case-mix effect. Almost 80 percent of cases were health workers, from two health facilities in Kigali (WHO: almost 78 percent of confirmed cases; 46 percent of all cases aged 30 to 39), treated at a dedicated MVD treatment centre with WHO critical-care support. They are not comparable to rural community cases in earlier outbreaks. That most cases were health workers is itself evidence of an infection-control failure at the start: the virus spread inside hospitals before it was detected. WHO's page does not say how long transmission ran before detection. The response was not purely national. WHO provided surge clinical support, IOM supported points-of-entry capacity, and US CDC, USAID and Africa CDC supplied laboratory supplies and technical support, and the DON lists therapeutics and vaccine research as a response function, so the investigational products may have contributed in ways this record cannot separate. The DON does not give the historical fatality range, so the comparison with earlier outbreaks is not made from this source. The figures are WHO restating Ministry of Health reports, so the government is reporting on its own response.
๐ธ๐ฌSingaporesince 2020delivered and closedbears on Institutional responsiveness
Singapore's Ministry of Finance introduced the Jobs Support Scheme in the Unity Budget of 18 February 2020, raised and extended it through the Resilience, Solidarity and Fortitude Budgets, the fourth of them presented on 26 May 2020, and disbursed $22.6 billion of wage support from April to December 2020.
The Finance Ministry built one wage-subsidy scheme on the existing payroll and pension contribution records, paid it automatically without applications, and then changed its rate and duration in each supplementary budget as the lockdown was imposed and extended.
Preconditions
A payroll and social contribution register that already identifies every local employee and employer
Fiscal reserves that Parliament could draw on without borrowing
A cabinet and Parliament able to pass supplementary budgets within weeks
Where else it has been used: Similar automatic wage subsidies ran in other countries during the pandemic, including Australia's JobKeeper and Germany's Kurzarbeit, which already existed; the distinctive part here was paying without applications through existing contribution records.
Limits: Every figure is the Ministry of Finance assessing its own measures. Disbursement measures how fast money moved, not whether the rules changed for the right reasons or whether the jobs would have been lost without it; the ministry's own job-saving figure on the same page is a modelled estimate, and the record does not carry it. The speed rested on accumulated reserves that most of the benchmarked countries do not have, so the case shows what a cash-rich centre of government can do, not responsiveness under fiscal constraint. Four budgets in about 14 weeks counts revisions; it does not show that each revision was well targeted, and wage subsidies paid on every local employee reached firms that would have kept their workers anyway. The scheme was temporary by design and wound down after the crisis, so the record is a concluded delivery and not evidence of standing institutional agility.
๐ธ๐ฌSingaporesince 1976still operatingbears on Disaster preparedness and recovery
Singapore's drainage authorities, the Drainage Department and later PUB, expanded the national drainage system under a drainage master plan, with a development drainage code from 1976, and reduced flood-prone areas from about 3,200 hectares in the 1970s to less than 25 hectares by 2025, with about $2.5 billion spent on drainage upgrades since 2011.
A single drainage authority drew a national master plan, built and widened canals and drains against it, and wrote flood protection into the development code so that every new building carried its own platform level and runoff controls.
Preconditions
One authority with jurisdiction over drainage across the whole territory
Planning control strong enough to impose platform levels and detention requirements on every developer
A small, fully urbanised catchment where engineering can reach most flood-prone land
Where else it has been used: The source-control rule, detention tanks and rain gardens required on larger sites since 2014, is the part that travels to larger cities; the full reduction depended on a city-state's control of land use.
Limits: Flood-prone area is PUB's own definition, counted by the agency that is responsible for reducing it, and the method behind the count is not published with it. The 1970s baseline is a rounded single figure with no year. An intermediate official figure, 40 hectares in 2013 in the PUB annual report, is consistent with a falling series, but the curve cannot be rebuilt from one dataset. The drop combines drainage works with land reclamation, raised platform levels and redevelopment, so it does not isolate the drainage programme. Flash floods still occur at listed hotspots, and the number says nothing about recovery from a major shock, because Singapore has not had a catastrophic flood in the period to test it. Most of the reduction predates the climate projections the ministry now plans against, so it shows capacity built against past rainfall, not readiness for sea-level rise. The started year is when the drainage code began, not the first drainage works.
๐ณ๐ฎNicaraguasince 2015still operatingbears on Disaster preparedness and recovery
Nicaragua's Ministry of Finance joined the CCRIF regional catastrophe insurance pool in 2015, renewed every year to at least 2021, raised its cyclone cover ahead of the 2020 season, and received five payouts totalling US$32.25 million, including US$30.6 million after hurricanes Eta and Iota.
The Ministry of Finance bought parametric insurance from a regional risk pool, sized the cyclone cover each season on the national meteorological institute's forecast, and drew the payout within two weeks of a storm to bridge the gap before other funding arrived.
Preconditions
A regional pool that spreads risk across many small economies so premiums are affordable
Donor or development-bank finance for entry fees and premiums while the budget cannot carry them
A hazard agency whose forecasts the finance ministry trusts enough to change its cover
Where else it has been used: The regional pooling model also runs in Africa and the Pacific, where low premium budgets have repeatedly limited cover.
Limits: The World Bank is reporting on its own project, and the brief covers the years to project closure in 2021; nothing in it documents payouts or cover after that. Operating status after 2021 is unconfirmed by the opened source. The premiums were mostly not paid by Nicaragua: an IDA credit of US$12 million financed participation fees and premiums for seven years; CCRIF and its donors offered 50 percent premium discounts, and the share of the premium paid from the credit fell in the last two years, with the Ministry of Finance budgeting the balance. The record therefore evidences a financing arrangement the state used, not one it funded. US$30.6 million is liquidity, not recovery: the brief gives no damage total for Eta and Iota to set it against, and a parametric payout measures whether a model threshold was crossed, not how well the money was spent. The brief's comparison of fatalities with hurricane Felix in 2007 is not carried here. The record overlaps in time with the existing SINAPRED evacuation record and documents a different delivery, the fiscal side of preparedness.
๐ฎ๐ฑIsraelsince 2020delivered and closedbears on Disaster preparedness and recovery
Israel started COVID-19 vaccination in December 2020 through its four health funds and had fully vaccinated about 70 percent of adults by April 2021; at the end of April 2021, 55 percent of its population was vaccinated, when coverage in most OECD countries was between 11 percent and less than one percent.
70 % of adults (about), adult population fully vaccinated (two doses), 2021-04. 55 % of population, population vaccinated at the end of april 2021, 2021-04-30. OECD, Economic Surveys: Israel 2023, retrieved 2026-10-01.
The move, and what it needed
The government bought early supply with a data-sharing agreement. It then delivered the campaign through the four health funds, which used their electronic records to call people in by risk group and their employed nurses to vaccinate, and it paid the funds premiums for reaching coverage targets.
Preconditions
Near-universal enrolment in a few health funds with complete electronic records
A primary care nursing workforce the funds could redeploy at once
A government able to sign a fast, unconventional procurement deal
Limits: The 70 percent is an approximate figure that the OECD takes from the Bank of Israel, not an OECD count. The speed was bought in part by an early Pfizer/BioNTech supply deal paid for with weekly anonymised aggregate epidemiological data, and Israel has a young population, so the case says little about what a slower procurement position would have achieved. An OECD working paper says Israel later fell into the lower half of OECD vaccine coverage, and it puts ultra-Orthodox vaccination at about half the rate of the rest of the population. The 2023 survey also records that excess mortality was around the OECD average, so fast vaccination did not by itself produce better mortality outcomes. This is response to one shock, not evidence of preparation before it. Both figures are in PDFs.
๐ณ๐ฌNigeriasince 2017still operatingbears on Institutional responsiveness
Nigeria's National Social Safety Nets Coordinating Office built a National Social Register of 15,673,296 households and, during COVID-19, a new Rapid Response Register that captured 2.7 million households, while the project paid targeted cash transfers to 9,666,420 direct beneficiaries in 37 states between 2017 and its close in December 2022.
Nigeria's federal government set up NASSCO and the National Cash Transfer Office to build one national social register and an electronic payment rail, then reused that infrastructure during COVID-19 to stand up a rapid-response register from satellite poverty maps and phone data for urban households outside the original register.
Preconditions
A standing social register and payment system already in place before the shock, so a new caseload could be added rather than built from nothing
External finance and restitution funds ring-fenced for transfers, which kept payments flowing through federal budget stress
An identity layer (NIN or BVN) that later became a hard gate, and whose coverage gaps then slowed payments
Where else it has been used: Shock-responsive registers spread across Africa during COVID-19, Togo's Novissi being the best known. Nigeria's version shows that the register scales faster than the payment and identity-verification rails behind it.
Limits: The gap asks for the speed at which programmes change, and the source gives no elapsed time: it shows that a shock register was built during COVID-19, not how many weeks it took or how quickly payments followed. The figures come from the World Bank's own completion review of a project it financed (IDA credit of about US$491 million, plus about US$335.5 million from the Abacha restitution fund), not from a Nigerian statistical source, and the review prints the original NSR target as 4,000, apparently a unit error that it does not explain. Registration is not payment: only 53 percent of beneficiaries were in the bottom two quintiles against a target of 80, only 6 percent of households received top-up transfers, more than three-quarters of households joined too late for three years of support, and benefit levels were lower than designed. After the review closes, the World Bank's January 2025 status report for the follow-on scale-up (https://documents1.worldbank.org/curated/en/099012925120031079/pdf/P17693518b066e0ff1b62c194b8b78c78dd.pdf) records that transfers since March 2024 go only to NIN- or BVN-verified beneficiaries, that oversight passed for a period to an inter-ministerial panel, and that about 1.4 million of 5 million supported households had received second and third tranches, so the delivery system's continuity is not secure.
๐ธ๐ชSwedensince 1989dismantledbears on Disaster preparedness and recovery
Sweden's civil defence agency รCB began procuring long-term stocks of medical consumables for wartime care in 1989, and the national build-up by รCB and Socialstyrelsen was completed in 1995 to 1996; the stocks were run down from the late 1990s, รCB's stockpiles ended entirely in 2001, and by the start of the COVID-19 pandemic in 2020 Socialstyrelsen held no personal protective equipment in its emergency stockpile.
0 items, personal protective equipment held in socialstyrelsen's national emergency stockpile before the pandemic, 2020. 150,000 patients (peak; reduced to 5,000), patients covered by socialstyrelsen's emergency stockpile, peak versus after the 1996/1997 defence decision, 1996/1997. Coronakommissionen, Sverige under pandemin (SOU 2021:89), published by Sveriges riksdag, retrieved 2026-10-01.
The move, and what it needed
The government had รCB and later Socialstyrelsen build national stocks of medical consumables and equipment for wartime casualties, then, after the 1996 total defence decision, the agencies sold off and ran down the stocks and moved supply onto the regions' just-in-time procurement.
Preconditions
A threat assessment that justified paying to store goods that are mostly never used
A national agency with storage sites and a mandate to rotate perishable stock
A political settlement that would keep funding the stock once the war threat receded, which is what failed
Limits: SOU 2021:19 sizes the stockpile at its peak in patients covered, not in volume or value: stocks for 150,000 patients and at most about 200 consumable items in 1996/1997, reduced to 5,000 patients with a trauma-care focus after the 1996/1997 defence decision. The build-up was sized for one to two months of wartime surgical consumption. The 2020 PPE zero measures a category the 1990s wartime stock is not shown to have held. The zero is Socialstyrelsen's own answer to the commission, and the same report says PPE bought during the 2014 Ebola outbreak had been held and lent to the regions, so 'no PPE' describes what was on hand in early 2020, not a stock that never existed. The dismantling was a deliberate decision based on a reassessed wartime threat, with the regions expected to hold their own stocks; the record documents a capability given up by choice, not one that collapsed. The 1950s figure of about 124,000 emergency hospital beds in SOU 2021:19 covers a different, older programme and is not used here. Both inquiries are long HTML transcriptions of PDFs with hyphenation breaks, and Sweden has since begun rebuilding health-sector stockpiling, which this record does not evaluate.
๐ฌ๐งUnited Kingdomsince 2020delivered and closedbears on Disaster preparedness and recovery
The UK government set up the Vaccine Taskforce in April 2020, became the first country to authorise a COVID-19 vaccine and begin vaccination in December 2020, and by the end of October 2021 the NHS in England had given two doses to 85% of adults.
The government created a Vaccine Taskforce outside normal departmental lines, gave it fast-track approval and broad indemnities to sign advance contracts with several developers before any vaccine was proven, and had the NHS deliver doses mainly through GPs and community pharmacies.
Preconditions
A single national health service with a GP registry that could call people by priority group
Treasury willingness to fund at-risk purchases across several candidate vaccines
A domestic regulator able to authorise quickly and a domestic research base (Oxford) with a vaccine candidate
Limits: The figures cover England only, not the United Kingdom; health is devolved and Scotland, Wales and Northern Ireland ran their own rollouts. The rollout responded to a shock already under way, so it shows capacity to recover, not preparation: the UK's pre-pandemic preparedness and 2020 death toll are outside this number. Uptake was uneven, at 64% among 18 to 24 year olds and 48% to 49% among some ethnic minority groups, and around 3.7 million adults were still unvaccinated at the end of January 2022. About 4.7 million doses were wasted, including 1.9 million AstraZeneca doses written off. The deaths averted figure in the same report is a UKHSA model estimate and is not used here. The data come from NHS England's own administrative reporting, analysed by the NAO.
๐ฒ๐ฝMexicosince 1999dismantledbears on Disaster preparedness and recovery
Mexico's federal government ran FONDEN from 1999 as a pre-arranged national fund for post-disaster reconstruction, allocating a record US$833 million to it in 2011 and insuring it with catastrophe bonds, and abrogated the FONDEN trust by the decree of 6 November 2020, moving disaster spending onto ordinary budget lines.
The Finance Ministry and the Interior Ministry built a rules-based trust that set aside budget for reconstruction before disasters and reinsured it through catastrophe bonds. A new administration and Congress then folded the trust into ordinary budget lines through a decree converting it into a budgetary scheme, justified by transparency.
Preconditions
A trust structure that could be extinguished by decree
An administration that treated off-budget trusts as a corruption risk rather than as insurance
No statutory floor that protected the reserve once the trust was gone
Where else it has been used: FONDEN's design, an ex-ante disaster fund topped up by parametric catastrophe bonds, was promoted by the World Bank as a template.
Limits: The metric shows the fund at its peak, not its dismantling. The abolition rests on a second World Bank document with no number in it: the Mexico Inclusive and Sustainable Economic Growth DPL (P178224, https://documents1.worldbank.org/curated/en/556411654274244589/pdf/Mexico-Inclusive-and-Sustainable-Economic-Growth-Development-Policy-Loan.pdf) refers to "the abrogation of the Mexico's National Disaster Fund (FONDEN)" (page 31), and the government's own letter in its annex says that after "the Decree published on 6 November 2020, the financial instrument FONDEN is changing from a trust to a budgetary scheme". So disaster spending did not stop. What ended was the trust, its reserve-building and its inter-institutional rules, and the same DPL says the 2022 emergency budget covered only an average year's losses. The record has no post-abolition figure for reconstruction spending or speed, so the claim that capacity fell is the World Bank's reading, not a measurement. The 2011 allocation is one year inside a volatile series. The same review says spending exceeded FONDEN's resources in five of the years from 1999 to 2011 and needed transfers of over US$3.3 billion in 2010, so the fund was never fully self-sufficient. The review was co-authored with the Mexican government it describes.
๐น๐ทTurkeysince 2000still operatingbears on Disaster preparedness and recovery
After the 1999 Marmara earthquake, Turkey's government created the Turkish Catastrophe Insurance Pool (DASK) in 2000 to provide compulsory earthquake insurance, and coverage rose from 23 percent of dwellings as reported by the World Bank in January 2011 to 57.7 percent, or 11.6 million insured homes, by October 2025.
The government, with World Bank financial and technical assistance, set up a public catastrophe pool that sells statutory earthquake cover through around 30 private insurers, keeps a retention layer and passes excess risk to international reinsurers.
Preconditions
A catastrophic earthquake that made the fiscal cost of rebuilding private homes politically visible
A property registry able to identify eligible dwellings
Access to international reinsurance at actuarially priced rates
Limits: Insurance penetration measures financial preparation for rebuilding private homes, not the capacity to prepare for or recover from a shock. It says nothing about building safety. The scheme has been compulsory since 2000, yet after 25 years 42 percent of dwellings are still uninsured, falling to 45 percent coverage in the Black Sea region. Only legally built dwellings on registered land are eligible. The two figures come from different publishers and may use different denominators. The 2025 figure is DASK's own data reported through the government's annual programme, not an independent count. The record has no figure for claims paid or payout speed after the February 2023 earthquakes, which is the test that matters. Cover per policy is capped, so insured losses understate household losses. The World Bank, which co-designed the pool and seeded it with a US$100 million contingent loan, is also the source of the earlier figure.
๐ฆ๐บAustraliasince 2018still operatingbears on Disaster preparedness and recovery
The Australian Government paid $3.1 billion to state and territory governments through the Disaster Recovery Funding Arrangements between 1 June 2018 and 30 June 2022, and the arrangements were activated for 49 disaster events that occurred in 2021โ22.
The Commonwealth pre-commits by ministerial determination to reimburse a set share of state and territory relief and recovery spending in defined categories once an event is activated, so states can spend first and claim afterwards under an agreed assurance process.
Preconditions
A federal fiscal capacity large enough to absorb a bad disaster year
State governments able to design, run and acquit relief programmes
An assurance process both levels of government accept, which the audit found still lacked performance monitoring
Where else it has been used: Canada's DFAA is the older version of the same cost-sharing move. Both reached modernisation reviews that pushed toward funding mitigation, because a reimbursement scheme alone pays to rebuild what the next event can destroy again.
Limits: Dollars paid measure the federal share of state recovery spending, which tracks how much damage the hazards did as much as how well recovery was run: a bad flood year raises the number. The audit is an independent assurance review of administration, not an evaluation of recovery outcomes, and it found the agencies did not monitor or report DRFA performance, so no figure on speed or quality of recovery exists to set beside the spending. The 49 events is the ANAO's own headline count of disasters that occurred in 2021โ22 and activated the arrangements; it counts activations, not their size or how well each was handled. Delivery is by the states; the Commonwealth reimburses. The ANAO site sits behind a Cloudflare check that blocks scripted requests, and its headline counter renders as 0 without JavaScript; the numbers were read in a browser at the same URL, and 49 also appears in paragraph 2.16. DRFA replaced the earlier Natural Disaster Relief and Recovery Arrangements in 2018, so the start year is that of the current determination, not of federal recovery cost sharing.
๐จ๐ฆCanadasince 1970still operatingbears on Disaster preparedness and recovery
Public Safety Canada's Disaster Financial Assistance Arrangements paid $1.84 billion in contributions to provinces and territories between 2011โ12 and 2015โ16, almost 45 percent of everything paid since the programme began in 1970, while 53 disaster events were approved by the Governor in Council between 2011 and 2016.
Ottawa cost-shares provincial and territorial response and recovery spending on a sliding scale above a per-capita threshold, triggered by an order in council, so that a single large disaster does not fall on one province's budget.
Preconditions
Provinces and territories with their own disaster assistance programmes to reimburse
A federal audit and acquittal process that provinces accept, even when it takes years
Fiscal room at the federal level to absorb a spike like the 2013 Alberta floods
Where else it has been used: Australia runs the same move as the DRFA. The Canadian expert panel's 2023 report and the 2025 redesign moved money toward mitigation and earlier support for survivors, which is the usual lesson of a long-running reimbursement scheme.
Limits: This is the department evaluating its own programme. Payments follow disaster losses, so a high figure is partly a record of bad luck: the 2013โ14 year alone (about $1.02 billion) carries more than half the five-year total. The same evaluation shows the money arrives slowly: for the 53 events approved in 2011โ2016 the estimated federal share was $3.8 billion against $1.6 billion actually issued, because provinces have up to five years after approval to request final payment, and before 2008 there was no time limit at all. The figures are nine years old; a 2025 parliamentary briefing note on the Public Safety site puts the cumulative total since 1970 at $8.9 billion, and the programme was redesigned from April 2025, so this record documents the arrangement as it stood in the evaluation period. Provinces and territories design and run the assistance; DFAA reimburses part of it.
๐บ๐ธUnited Statessince 1968operating below its peakbears on Disaster preparedness and recovery
OpenFEMA's records of National Flood Insurance Program contracts in force fell from 4,667,875 on 31 December 2009 to 3,589,275 on 31 December 2025, a fall of about 23 percent in the federal flood insurance base over 16 years.
3,589,275 contracts, nfip policy contracts in force (openfema records active on the date), 2025-12-31. 4,667,875 contracts, nfip policy contracts in force at the first complete year-end in the dataset, 2009-12-31. FEMA, OpenFEMA NFIP Redacted Policies dataset (v3), retrieved 2026-10-02.
The move, and what it needed
Congress created a federal insurer of last resort for flood, tied to community floodplain management and to a purchase requirement for federally backed mortgages in high-risk zones, so that recovery after floods would be paid for by premiums rather than only by disaster aid.
Preconditions
Flood maps that identify who must buy and that stay current
Enforcement of the mandatory purchase requirement by lenders, which is where take-up leaks
Premiums that households will keep paying once the mortgage requirement no longer binds them
Where else it has been used: Public flood pools elsewhere (Flood Re in the United Kingdom, CCRIF for Caribbean governments, DASK for Turkish earthquakes) face the same tension: pricing risk honestly drives voluntary buyers out, and keeping them in requires subsidy or compulsion.
Limits: Both numbers are computed, not published headlines: they are the record counts the OpenFEMA API returns for policies whose effective date is on or before, and termination date after, each date, which is the method FEMA's own NFIP data FAQ gives. The 2009 figure comes from the same URL with the date changed to 2009-12-31. The count is of contracts, not policies: one condominium contract can cover many units, and FEMA's own 'NFIP by the Numbers' report for 31 August 2026 shows 3,388,273 contracts but 4,460,103 policies. The API count runs about 2 percent above FEMA's contracts-in-force report (3,589,275 against 3,512,555 for December 2025), so the level carries a small method error, though both ends use the same method. The dataset starts with policies effective in 2009, so the true earlier peak cannot be shown from it. The fall is not all lost protection: private flood insurance has grown, some properties left mapped flood zones, and Risk Rating 2.0 (2021) raised premiums by design to price risk. What the record shows is that the public programme's reach shrank, not that total insured flood exposure fell by the same amount.
๐ฐ๐ทSouth Koreasince 2015still operatingbears on Disaster preparedness and recovery
The Korea Meteorological Administration expanded the national seismic network from 195 stations when its earthquake early warning service started in January 2015 to 550 stations, and by May 2026 issued early warnings within 5 to 10 seconds of first detection.
The national weather agency densified the seismic network roughly threefold and tied automatic magnitude estimates to the national cell-broadcast emergency text system, so a warning reaches every phone in the affected area without a human approval step.
Preconditions
A national cell-broadcast channel already carried by all mobile operators
Budget for a decade of station build-out and telemetry
An agency trusted to issue automatic alerts that will sometimes be wrong
Where else it has been used: Japan's JMA system is the model; Taiwan, Mexico (SASMEX) and the United States (ShakeAlert) run variants. The Korean retelling stresses network density, because a small, densely populated peninsula has little lead time between fault and city.
Limits: This is the agency describing its own service in a government press item, in Korean. A station count is an input, and the 5 to 10 second issuance time is a stated operating standard, not a measured average over real events; the page gives no record of how warnings performed in actual earthquakes, how many were false or missed, or what share of people received the cell-broadcast text. The agency itself names the weakness: close to the epicentre, strong shaking arrives before the warning, which is why it added an on-site alert stage from 28 May 2026. Korea has had few damaging earthquakes since 2016, so the system has been tested rarely. Earlier reporting puts the 2015 issuance time at about 50 seconds, but that figure is not on this page and is not used here.
๐ฟ๐ฆSouth Africasince 2001operating below its peakbears on Disaster preparedness and recovery
The South African Weather Service reported that the availability of its weather radar infrastructure fell from 73.83 percent in 2021/22 to 52.23 percent in 2022/23 and 44.25 percent in 2023/24, against a 75 percent target, as load shedding, missing spares and ageing C-band radars took sites down.
A national weather service built and ran a countrywide radar network that underpins severe-weather warnings, then let it degrade when power supply, spares contracts and maintenance budgets failed together.
Preconditions
Reliable grid power or funded backup at remote radar sites
Maintenance contracts with the original manufacturer and a spares budget
Site security against theft and vandalism
Where else it has been used: The Thai warning-tower and German siren cases in this corpus show the same pattern: warning infrastructure decays quietly between disasters, because nobody measures it until it is needed.
Limits: The figure is in the PDF of SAWS's own annual report (printed page 70, PDF page 72), and the copy that opens is hosted by the Parliamentary Monitoring Group, a civil-society mirror of documents tabled in Parliament; the SAWS website did not serve the report to a script. 2021/22 is not shown to be the historical peak: the 2024/25 plan presented to Parliament gives 73 percent for 2020/21 as well, so the record shows a fall from a plateau, not from an all-time high. Availability is the share of time radars produced data, not forecast or warning quality, and the population covered by a working radar is not reported. The main cause named is national load shedding, an Eskom failure outside SAWS's control (see zaf-eskom-loadshedding), compounded by spares shortages and vandalism. SAWS says the ageing C-band radars will be dropped from the statistic from 2024/25, which will raise the reported figure without repairing them. The start year is SAWS's establishment under the South African Weather Service Act of 2001, which is not stated on the cited page.
๐ฐ๐ชKenyasince 2009still operatingbears on Disaster preparedness and recovery
Kenya's National Drought Management Authority expanded the shock-responsive Hunger Safety Net Programme from four to eight arid counties, enrolling 32,000 new households paid from the government's own budget by December 2024, while 100 percent of regular transfers in the original four counties were financed by the government.
The drought authority pre-registers and banks almost every household in the arid north, pays the poorest a regular transfer, and uses satellite vegetation triggers to extend payments to more registered households when drought hits, so money can move without a new targeting exercise.
Preconditions
A household registry and bank accounts set up before the shock
A trigger rule tied to an objective drought index
Contingent money ready to flow when the trigger fires, which is where Kenya's version fell short in 2021 to 2023
Where else it has been used: Ethiopia's PSNP and the Sahel adaptive safety nets use the same scalable design. The Kenyan retelling shows that the registry and trigger are the easy part: the contingent finance is what fails under fiscal stress.
Limits: The source is the lender's completion report on a project it financed, and the numbers are in the PDF (32,000 households on printed pages 22 and 35; four to eight counties on printed page 17; drought arrears on printed page 16). The headline number is regular coverage, not drought response: the report says that during the 2021 to 2023 drought the scalable emergency component ran a deficit, emergency transfers were underfinanced and delayed, and arrears were paid only later, and that planned scale-up payments in 2024/25 were not triggered. So the record shows the standing system, not a fast drought payout. HSNP covers arid and semi-arid counties, not the whole country, by design. The 2009 start year is the first HSNP phase, which ran as a donor-funded pilot before the government took it over, and is not stated in this report.
๐ง๐ทBrazilsince 2024delivered and closedbears on Disaster preparedness and recovery
After the floods that began in late April 2024 in Rio Grande do Sul, the federal government paid R$ 2.1 billion in Auxรญlio Reconstruรงรฃo to 429,000 families, according to the balance its extraordinary secretary for the reconstruction of Rio Grande do Sul presented to a Chamber of Deputies committee on 12 February 2025.
The federal government created a flat, one-off cash grant by provisional measure within two weeks of the disaster, paid through the existing social-payment rails to families on lists the affected municipalities certified, and accepted self-declaration to trade verification for speed.
Preconditions
A federal emergency-declaration process that names eligible municipalities
Payment rails and household registries that can reach families without a bank visit
Municipal civil-defence offices able to list affected households quickly
Where else it has been used: Flat post-disaster cash grants travel easily; the Brazilian case shows the bottleneck moving from paying to certifying, which is where the audit court found the delays.
Limits: The figure is the executive's own balance, stated by the federal secretary for the reconstruction of Rio Grande do Sul at a Chamber hearing and reported by the Chamber's news agency; it is not an audited count, and the Ministry of Integration pages that carried the running totals were behind a login wall during the 2026 electoral period, so the primary executive source could not be opened. The grant was eligibility-by-municipal-list plus self-declaration, so the count shows money reaching families, not need met: the Tribunal de Contas da Uniรฃo (Acรณrdรฃo 1297/2025, per its news page 'Tribunal analisa benefรญcios a desabrigados das enchentes') found 87,300 requests (11.97%) pending for more than 100 days at 13 September 2024 and one municipality where none of 1,722 requests had been approved, and ordered the ministry to clear the backlog. The budget was sized for an estimated 375,000 families. A cash transfer is response and early recovery, not reconstruction of housing or protective works, and says nothing about whether exposure fell.
๐จ๐ทCosta Ricasince 2017delivered and closedbears on Disaster preparedness and recovery
In the 17 months to September 2018, the Instituto Costarricense de Electricidad, under an agreement with the national emergency commission (CNE), completed 64 of the 65 priority reconstruction works after Hurricane Otto of November 2016, rebuilding 20 bridges and 211 km of rural roads for โก18,500 million.
The CNE, which holds the national emergency fund and the declared emergency plan, contracted the state electricity and telecom utility as a single builder for the priority works instead of tendering each bridge, and coordinated it with the public works ministry and five municipalities.
Preconditions
A standing emergency fund and a legal emergency declaration that unlocks direct contracting
A state enterprise with engineering crews and plant spread across the country
A short, costed list of priority works agreed in the emergency plan
Where else it has been used: Using a state utility as the reconstruction contractor recurs in Costa Rica's later emergencies; it travels only where such a utility exists and is trusted, and it concentrates risk if that enterprise is overstretched.
Limits: The source is the government's own completion announcement; the original presidencia.go.cr page now returns 404 and the text survives in ReliefWeb's republication, which names the Government of Costa Rica as source. Completion is measured against the government's own list of 65 priority works, not against total damage: the same government estimated recovery needs at about โก130,000 million, so โก18,500 million is a fraction of the recovery and the record shows the priority tranche only. The started year is computed from the release's '17 months' and is not stated directly. The remaining work, a bridge over the Zapote river, was delayed by expropriation. Nothing here shows whether the rebuilt bridges and roads were built to a higher standard or maintained; the Contralorรญa has reported maintenance gaps on emergency-fund projects, but its document server refused connections today and that finding is not relied on.
๐ฌ๐นGuatemalasince 2009delivered and closedbears on Disaster preparedness and recovery
On 21 April 2020, five days after its second World Bank disaster-risk loan with a Catastrophe Deferred Drawdown Option became effective, Guatemala's government decided to withdraw the entire US$200 million to meet the fiscal shock of COVID-19, and the loan was fully disbursed. A first such credit of US$85 million, approved in 2009, had been drawn in 2010 after the Pacaya eruption and Tropical Storm Agatha.
The finance ministry pre-negotiated a contingent credit line tied to a declared state of calamity, so that a declaration released money within days instead of waiting for a new loan through Congress, and layered parametric insurance and a contingent IDB line on top under a 2018 disaster risk financing strategy.
Preconditions
A legal state-of-calamity declaration that can serve as the drawdown trigger
Congressional approval of the credit before the shock, which is where Guatemala lost a year
A finance ministry with a disaster risk financing strategy and budget tags to track the spending
Where else it has been used: CAT DDOs are used across Latin America (Colombia, Peru, Costa Rica, Guatemala); Guatemala's sequence shows the instrument lapsing between uses unless it is renewed before the next disaster.
Limits: Both numbers are on page 2 of the 18-page IEG review PDF ('The loan was fully disbursed'; 'On 21 April 2020 (just five days after the project became effective), the Government decided to withdraw the entire loan'). The US$85 million first credit, its approval on 14 April 2009 (the started year) and its 2010 drawdown are from the World Bank's 2019 Program Document for the same operation (documents1.worldbank.org/curated/en/977151559008837975/pdf/Guatemala-Second-Disaster-Risk-Management-Development-Policy-Loan.pdf, PDF pages 25 and 48), not from the IEG review. The money is a World Bank loan: the Guatemalan delivery is arranging it, legislating it and drawing it, and the same documents show how fragile that was. There was no contingent instrument in place for the June 2018 Fuego eruption, when 'in the absence of any other financial instrument' Congress fast-tracked an unrelated budget-support loan (Program Document page 11); total resources available for disasters were under 0.1 percent of GDP in 2017 (page 25); the second credit took a year to become effective because of conflict between the executive and Congress; and the DRM law it supported was not adopted (IEG review page 17). The withdrawal shows liquidity arrived fast, not that it was spent well; the review notes the bank could no longer supervise once the line was drawn.
๐ต๐ชPerusince 2021delivered and closedbears on Disaster preparedness and recovery
From its first dose on 9 February 2021 to the end of 2023, Peru's national COVID-19 vaccination campaign administered 91.3 million doses and completed the primary series for 28.7 million people, 87 per 100 population, with 22.2 million boosted.
The Ministry of Health ran a national campaign through its own clinics plus mass sites and age-tiered scheduling, and tied access to indoor public spaces to a vaccination card from December 2021 to push the late adopters.
Preconditions
Advance purchase contracts and a central cold-chain warehouse (CENARES)
A national immunization register that can count doses by person
Regional health directorates able to staff mass sites and outreach
Where else it has been used: The same campaign model ran across the region; Peru's coverage outpaced most neighbours, but the gap between Lima and the Amazon regions is the part that recurs everywhere.
Limits: WHO publishes what Peru's Ministry of Health reports, so the count is the country's administrative record, not a serosurvey; the CSV is machine-readable and the Peru row carries DATE_UPDATED 2023-12-29. Doses and persons are cumulative and do not show timeliness: vaccination came after Peru had already recorded one of the world's highest COVID-19 death rates, so the campaign evidences recovery capacity, not preparedness that prevented the first waves. Coverage was uneven: the ministry's own 11 December 2021 release, when 70 percent of the target population had two doses, named Madre de Dios, Puno and Ucayali below 50 percent. In January 2023 the ministry reported 1,964,490 doses close to expiry in its central warehouse. Per-100 figures use WHO's population denominator, not the national target population.
๐ต๐พParaguaysince 2019delivered and closedbears on Disaster preparedness and recovery
By 29 May 2019, Paraguay's Secretarรญa de Emergencia Nacional had assisted 62,521 of the 65,000 families affected by the flash floods of March to May 2019 across 14 departments, delivering 1.5 million kg of food, 150,000 roofing sheets and 110,000 plywood boards.
The national emergency secretariat, holding its own emergency fund, procured food and building materials centrally and distributed them department by department against a household count, with shelters in Asunciรณn run alongside civil-society groups.
Preconditions
A national emergency fund that can be drawn without a new appropriation
Stockpiles or framework contracts for food and roofing material
Municipal and departmental lists of affected households
Where else it has been used: Relief-in-kind against a household count is the common Latin American model; it closes the response gap but leaves recovery and exposure to other agencies.
Limits: Both the numerator and the 65,000 denominator are the government's own counts, published by the agency that did the work, and nobody independent checked them; the text is read on ReliefWeb's republication because sen.gov.py did not answer a scripted request today. Assistance is relief in kind (food, props, sheets and boards for self-repair), not reconstruction, resettlement or risk reduction, and the release gives no timing, so it cannot show how fast families were reached. Some assistance in the Chaco was delivered with the First Lady's office. Repeated flooding of the same Asunciรณn riverside settlements shows the relief cycle does not reduce exposure.
๐ธ๐ปEl Salvadorsince 2021delivered and closedbears on Disaster preparedness and recovery
From its first dose on 17 February 2021 to the end of 2023, El Salvador's national COVID-19 vaccination campaign administered 11.5 million doses and completed the primary series for 4.4 million people, 68 per 100 population.
The Ministry of Health vaccinated through mass sites and the public clinic network after securing early supply, and published a running national count.
Preconditions
Early vaccine supply contracts or donations
A small territory and dense road network that make mass sites reachable
An immunization register that counts persons as well as doses
Where else it has been used: The mass-site model travelled across Central America; El Salvador's case shows that a fast start does not by itself carry coverage past two thirds.
Limits: WHO republishes the Ministry of Health's administrative count; the El Salvador row carries DATE_UPDATED 2023-12-29 and no independent coverage survey is cited. A third of the population never completed the primary series and only 29 per 100 received a booster, so the delivery is national in reach but partial; Costa Rica (86) and Peru (87) completed more in the same file. The cumulative count does not show the campaign's early speed, which is its usual selling point, and that speed is not evidenced here. The government's covid19.gob.sv dashboard did not answer a scripted request and the PAHO country profile returned 403 in an earlier pass, so no national publication was read. The issue-47 rejection of this lead was on publisher_metric; the WHO file is the new evidence that fixes it.
๐ฌ๐งUnited Kingdomsince 2020delivered and closedbears on Institutional responsiveness
The UK's Coronavirus Job Retention Scheme, which started in March 2020, had 8.9 million jobs on furlough at its peak on 8 May 2020 and supported 11.7 million jobs from 1.3 million employers before it closed on 30 September 2021.
The Treasury designed a wage subsidy covering 80 percent of a furloughed employee's pay and HMRC paid it to employers through the PAYE Real Time Information payroll system it already ran, so a new programme reached millions of jobs within weeks.
Preconditions
A tax authority already receiving each employer's payroll in real time
Borrowing capacity to fund an open-ended subsidy
Willingness to accept fraud and deadweight in exchange for speed
Where else it has been used: Germany and France extended short-time work schemes that already existed in law; the UK built one from nothing on top of its payroll data, as Australia did with JobKeeper.
Limits: HMRC is reporting on a scheme it ran, from claims data, and the counts are jobs (employments), not people: one person furloughed from two jobs counts twice. The release dates the scheme's start as March 2020 and its daily series from 23 March 2020; the announcement date (20 March 2020) and the opening of the claims portal in April 2020 are not on this page, so the speed reading rests on the peak falling on 8 May 2020, about seven weeks after the series begins. The release says nothing about cost, deadweight (jobs that would have survived anyway) or fraud and error, which HMRC and the National Audit Office reported separately, and it cannot show whether furlough preserved jobs or delayed their loss: 1.16 million jobs were still on furlough when it closed. A new scheme built in weeks under extreme pressure says little about how fast British rules change in normal times.
๐ซ๐ทFrancesince 2020delivered and closedbears on Institutional responsiveness
France scaled its short-time work scheme (activitรฉ partielle) from an industrial instrument into an economy-wide one in the spring of 2020, placing 8.4 million employees on it at the April 2020 peak and 11.3 million at some point between March 2020 and December 2021.
The state widened eligibility and raised the compensation rate of an existing short-time work scheme by decree, and the state and Unรฉdic paid employers for hours not worked, so firms could cut hours instead of jobs.
Preconditions
A short-time work scheme already in law with an employer claims channel
A jointly managed unemployment insurance fund able to share the cost
Fiscal room to fund roughly a billion euros a week at the peak
Where else it has been used: The German Kurzarbeit model, which France's scheme resembles, did the same work in 2008-2009 and again in 2020; the existing legal channel is what let both pay within weeks.
Limits: Unรฉdic is a private-law association run jointly by the employer and union confederations that manages unemployment insurance under a statutory mandate; it is not a statistics office, and this is a signed analytical study (Oana Calavrezo) in which it reports on a programme it co-financed at about one third: 14.4 billion euros in 2020 and 2021 (11.5 billion of spending and 2.9 billion of forgone receipts), more than half of its deficit in those years. The counts are Unรฉdic's own calculations on the payment agency's claims data (ASP, Extranet Activitรฉ partielle, data at 17 January 2022, de-duplicated), stated in the full report linked from the page; the December 2021 months were provisional at that date. The labour ministry's statistics service (DARES) publishes the official series but its pages refused automated access on the retrieval date. The scheme existed before the crisis; what changed was its rules and scale, which the page describes as a change of scale 'in a few days' in the spring of 2020 without dating each decree. The count is of employees placed on the scheme, including long-term partial activity (APLD), not of jobs saved, and 45 percent of those placed on it only used it in the first lockdown. Cost per job and deadweight are not shown.
The American Rescue Plan, passed in March 2021, expanded the Child Tax Credit and made it payable monthly, and on 15 July 2021 the IRS paid roughly 15 billion dollars to families of nearly 60 million children; Congress did not extend the expansion beyond 2021.
60,000,000 children (nearly), eligible children in families paid in the first monthly payment, 2021-07-15. 15,000,000,000 US dollars (roughly), paid in the first monthly payment, 2021-07-15. US Department of the Treasury, press release of 15 July 2021, retrieved 2026-10-02.
The move, and what it needed
Congress turned an annual tax credit into a monthly payment and the IRS paid it automatically from prior-year tax returns, reaching most families within four months without a new application system.
Preconditions
A tax authority holding recent returns and bank details for most families
A legislative majority willing to pass a large temporary expansion
A political settlement to make it permanent, which was missing
Where else it has been used: Child allowances paid monthly are permanent in Canada and much of Europe; the US case shows a temporary expansion can be delivered quickly and lost as quickly when its legal basis is time-limited.
Limits: The source is the Treasury announcing its own first payment, rounded ('nearly', 'roughly'), not an audited count. The four-month interval from the March 2021 law to the July payment is derived from the release's 'passed in March' and its date. The expansion was legislated for 2021 only and lapsed when Congress declined to extend it, so 'dismantled' records a decision not to continue rather than a repeal. The loss is documented by the Census Bureau, not by this source: in Poverty in the United States: 2022 (https://www.census.gov/library/publications/2023/demo/p60-280.html) the Supplemental Poverty Measure child poverty rate more than doubled, from 5.2 percent in 2021 to 12.4 percent in 2022, and other programmes also expired in that year, so not all of the rise is the credit. Families the IRS could not find through prior tax filings were missed by the automatic payments.
๐ฎ๐ณIndiasince 2020delivered and closedbears on Institutional responsiveness
India's Department of Food and Public Distribution, acting on the Garib Kalyan package announced in March 2020, allocated about 121 lakh tonnes of extra free foodgrains to States on 30 March 2020, and States distributed 37.5 lakh tonnes covering about 75 crore beneficiaries in each of April and May 2020.
The central government doubled the grain ration by order and pushed the extra stock from Food Corporation of India depots through the existing ration-shop network, so no new eligibility system had to be built.
Preconditions
Large public grain stocks already held in central depots
A national ration-card register and shop network covering about 80 crore people
States able to lift and distribute stock within weeks
Where else it has been used: In-kind rations through an existing public distribution system are hard to copy where no such stock and register exist; cash transfers did the same job elsewhere.
Limits: The figures are the ministry's own, compiled from what States reported to the Food Corporation of India, and are not an audit of what reached households; distribution counts at ration shops are known to include leakage. The scheme worked by adding 5 kg per person to the existing National Food Security Act ration lists, so people outside those lists, including many migrant workers stranded in the 2020 lockdown, were not covered by this channel. The extra entitlement was extended repeatedly and ended in December 2022, when the Cabinet instead made the regular NFSA ration free from January 2023. Speed is read from the release's dates (allocation on 30 March 2020, distribution from April) rather than a timestamped series.
The Central Bank of the UAE launched the Targeted Economic Support Scheme in March 2020, combining a zero-cost liquidity facility with loan deferrals and capital relief, and by 24 September 2020 more than 300,000 individuals, close to 10,000 SMEs and more than 1,500 private corporates had benefited from the facility.
The central bank lent to banks at zero cost against collateral and relaxed capital and liquidity buffers on condition that they defer repayments for affected borrowers, so relief reached households and firms through existing bank relationships.
Preconditions
A banking sector holding most household and SME credit
A central bank with authority to set conditional liquidity and capital relief
Fiscal and reserve strength under the currency peg
Where else it has been used: Most central banks ran deferral and liquidity schemes in 2020; the conditional zero-cost facility is the UAE's variant.
Limits: The figures are a central bank press release about its own scheme, rounded ('more than', 'close to'), with no breakdown of amounts or of who would have received relief anyway. The March 2020 start, the extension the following month and the run to 30 June 2022 are stated in the central bank's UAE MSMEs Business Survey Report 2020 (page 10, https://centralbank.ae/media/rtdlng34/cbuae-msme-survey-report-2020-en.pdf), not in this release, and the central bank's web pages sit behind a bot check while its PDFs open. Benefiting from a liquidity facility means a bank passed on a deferral or cheaper credit; it does not show firms or jobs saved. Banks surveyed in that report counted 7,751 MSME customers helped by TESS in 2020, a different measure from this one.
๐จ๐ณChinasince 2020delivered and closedbears on Institutional responsiveness
China's human resources ministry, finance ministry and tax administration let provinces exempt or halve employer pension, unemployment and work-injury contributions from February 2020 under a notice dated 20 February, and exemptions, reductions, deferrals and rate cuts together reduced contributions by 1.54 trillion yuan between February and the end of 2020.
Three central agencies used the Social Insurance Law's authority to suspend the employer share of contributions by notice, and the tax administration, which collects them, applied it from the next collection cycle.
Preconditions
Contributions collected centrally by the tax administration
Pension fund balances large enough to absorb a year of lost revenue
Central authority to change contribution rules by notice rather than law
Where else it has been used: Contribution holidays were used widely in 2020; China's was large because contributions are a large share of labour cost, and its cost was a fund deficit rather than a budget line.
Limits: The figure is a government figure republished by the NDRC Employment Department from China News Service (ไธญๅฝๆฐ้ป็ฝ), which the page names as its source; the page does not name the originating ministry or event, and the figure does not appear in the human resources ministry's 2020 statistical bulletin. It adds exemptions, reductions, deferrals and rate cuts together, and the rate cuts include reductions in force before the February 2020 notice, and a deferral is postponed rather than forgiven, so 1.54 trillion yuan overstates the relief. The rule itself is dated by the notice (Ren She Bu Fa 2020 No. 11, 2020-02-20, https://www.gov.cn/zhengce/zhengceku/2020-02/21/content_5481861.htm), which let provinces exempt small firms for up to five months and halve large firms' contributions for up to three. The cost landed on the funds: the same page reports 2020 revenue of 5.02 trillion yuan against spending of 5.75 trillion. This is relief for enterprises that pay contributions, not for informal workers.
๐ซ๐ฎFinlandsince 2020delivered and closedbears on Institutional responsiveness
Finland redirected Business Finland, its innovation funding agency, to pay emergency development grants to firms hit by the epidemic, starting from a 700 million euro allocation in the supplementary budget adopted on 26 March 2020, and by 2 September 2020 it had allocated about 912 million euros to 18,770 projects out of 28,715 that applied.
Rather than build a new aid channel, the government gave its innovation agency new grant instruments and money by supplementary budget, so applications could open within days through an existing online system.
Preconditions
An agency already running an online grant system that reached most SMEs
Parliament able to pass a supplementary budget within days
Tolerance for weaker screening than normal R&D funding
Where else it has been used: Several countries used existing enterprise agencies for 2020 grants; the Finnish case shows the trade between speed and targeting when the agency's purpose is innovation, not relief.
Limits: The ministry is reporting its own agency's grants, and an allocation is a decision, not a payment. The 26 March 2020 supplementary budget and the 700 million euro allocation come from a separate government release (https://valtioneuvosto.fi/en/-//1410877/valtion-rahoitusta-yrityksille-koronavirustilanteessa-vahvistetaan-yritystukiin-miljardi-euroa), not from the source page, and neither page explains how allocations reached 912 million euros against the initial 700 million. Using an R&D agency for emergency aid was fast but loosely targeted: the grants were for 'development' (79 million euros for preliminary analyses and 832 million for development funding), and later reports of recoveries and suspected misuse are not on either page. Neither page shows whether the firms survived or what share of the grants was later recovered.
๐บ๐พUruguaysince 2020delivered and closedbears on Institutional responsiveness
Uruguay's labour ministry created partial and flexible unemployment-subsidy modalities by resolution after the health emergency of 13 March 2020, and the social security bank (BPS) paid 12,773 beneficiaries under them in April 2020, while total unemployment-subsidy beneficiaries rose from 45,937 in February to a record 185,769 in May.
The labour ministry used its resolution power to open the unemployment subsidy to workers whose hours were cut rather than suspended, and BPS paid it through the claims channel employers already used.
Preconditions
An unemployment insurance law that let the ministry create special regimes by resolution
A single social security agency holding employer and wage records
Employers able to file partial suspensions online
Where else it has been used: Partial unemployment benefits are the short-time work idea under another name; Uruguay's version was built by resolution in weeks because the legal hook for special regimes already existed.
Limits: BPS is analysing a benefit it pays, from its own administrative records, and the numbers are inside a PDF (Cuadro 2 on page 6 and Cuadro 4 on page 13 of the file). Most of the surge went through the ordinary subsidy for total suspension, which already existed; the new modalities, set by labour ministry resolutions 143/20, 163/20, 233/20 and 1024/20 (partial) and 2225/20, 21/21 and 97/21 (flexible), peaked at 30,221 beneficiaries in June 2020 and fell to 10,549 by December 2021. The record evidences rules changed and paid within weeks, not jobs kept: a beneficiary count is a monthly stock, and the report does not estimate how many of those workers returned to their jobs. The end date of the special regime is not in this source.
๐จ๐ฑChilesince 2020delivered and closedbears on Institutional responsiveness
Chile's Congress created the Ingreso Familiar de Emergencia by Ley 21.230, published on 16 May 2020, and the Ministry of Social Development and Family issued payments to 3,419,351 households in 2020 and to 8,374,718 households in 2021, before the programme ended in 2021.
Congress legislated a time-limited cash benefit keyed to an existing social registry, then amended it as the pandemic went on, and the ministry paid through bank accounts and in-person collection to every registered household that met the income test.
Preconditions
A national social registry (Registro Social de Hogares) covering most households before the shock
Fiscal space and a legislature able to pass and amend a benefit law within weeks
Payment rails reaching households with and without bank accounts
Where else it has been used: Registry-keyed emergency transfers were the common Latin American response in 2020, from Guatemala's Bono Familia to Panama's Vale Digital. Chile's version differs in scale relative to its registry, and in being repeatedly widened by law rather than by decree.
Limits: The ministry is reporting on its own programme, and the counts sit in a six-page PDF monitoring sheet (component table on page 4); no Contralorรญa audit was consulted. The unit is households in the Registro Social de Hogares to which a payment was issued in the year, which is a registry unit and not a census household, and a household paid in several months counts once per year as far as the sheet shows. The sheet gives annual totals only, so it does not show how many weeks passed between the law and the first payment; the law's publication date (16 May 2020) comes from the Biblioteca del Congreso Nacional record of Ley 21.230 (https://www.bcn.cl/leychile/navegar?idNorma=1145400), which also shows Ley 21.251 amending it on 3 August 2020. Initial budget execution in 2021 was 2,024 percent of the budget law (page 2), which shows the scale of in-year change but also that the programme ran outside the normal budget process. The monthly ministry reports that would carry the week-by-week roll-out (desarrollosocialyfamilia.gob.cl/storage/docs/ife/) returned HTTP 500 on 2 October 2026. Spending of about 17.0 trillion pesos in 2021 says nothing about targeting errors or the poverty effect, and a transfer that ended in 2021 does not show that Chilean rules change quickly in general.
๐ฐ๐ชKenyasince 2020delivered and closedbears on Institutional responsiveness
The Central Bank of Kenya announced emergency measures on 18 and 24 March 2020 that let banks restructure loans performing on 2 March 2020, and by the end of February 2021, days before the measures expired on 2 March 2021, loans of Ksh 1.7 trillion, 57 percent of the banking sector's gross loans, had been restructured.
The central bank changed loan classification and provisioning rules by circular within a week of the first case, letting banks extend repayment, grant moratoria and waive fees on loans performing before the shock, and set a fixed expiry date after which standard classification resumed.
Preconditions
A central bank with direct supervisory authority over loan classification
Banks capitalised enough to carry restructured loans for a year
A dated cut-off (loans performing on 2 March 2020) that kept pre-existing bad loans out
Where else it has been used: Time-bound restructuring windows were used by many central banks in 2020; the fixed expiry and the published cumulative-versus-outstanding figures are the part that makes Kenya's version auditable.
Limits: The regulator is grading its own measure: the release calls it highly effective and gives no independent evaluation, and the figures are a one-page press release PDF, not a supervisory dataset. Restructuring is a forbearance rule, not new money, and restructured loans can hide losses: the share says nothing about how many restructured loans later defaulted, beyond the release's own statement that over 95 percent of the Ksh 569.3 billion still outstanding was being repaid on restructured terms. The cumulative 57 percent counts amounts, not borrowers, and includes loans restructured more than once if banks reported them so. The speed evidence is a separate CBK release: the Monetary Policy Committee statement of 29 July 2020 (https://www.centralbank.go.ke/uploads/mpc_press_release/603807985_MPC%20Press%20Release%20-%20Meeting%20of%20July%2029%202020.pdf) reports KES 844.4 billion, 29 percent of the loan book, restructured by the end of June 2020, the same release that reports the gross NPL ratio at 13.1 percent. One temporary financial-sector rule does not show how fast Kenyan public programmes change in general.
๐ญ๐ณHondurassince 2020operating below its peakbears on Institutional responsiveness
Honduras widened its national conditional cash transfer during COVID-19 and Hurricanes Eta and Iota, from 1,153,069 beneficiaries in 2019 to 1,528,030 in 2020, and paid emergency top-ups to 24,891 households outside the roster in late 2020, but after the 2022 reorganisation of the social protection sector payments paused, rural payments resumed only in December 2022 and urban payments only in August 2025, after a three-year suspension.
The social protection ministry used an existing conditional cash transfer and its registry to add vulnerable households and pay top-ups after COVID-19 and two hurricanes, financed by an additional World Bank credit; a new administration then moved the programme to a new ministry and new targeting rules, and payments stopped until the new architecture could pay.
Preconditions
A standing transfer programme with a beneficiary registry before the shock
External finance that could be added within a year
Implementing institutions that survive a change of government, which is the precondition that failed here
Where else it has been used: Honduras published a shock-response protocol for the transfer (PROMESA-PSA, Ministerial Agreement 159-SEDESOL-2025, 9 December 2025) built on the 2020 top-up. The lesson the ICR draws is that institutional reform can disrupt delivery more than the external shocks did.
Limits: The publisher financed the project (US$30 million, then a US$70 million additional financing in June 2021) and rates it Substantial on efficacy; the figures are in a PDF (beneficiary counts and the top-up figure on page 19, paragraphs 31 and 33; the payment pause on page 14, paragraph 14; the three-year urban suspension in paragraph 39). The ICR states explicitly that the 1,528,030 peak (cumulative enrollment) and the 922,832 final-year caseload are distinct measures that should not be compared directly, and attributes the lower figure to a compressed implementation window, not a reversal of reach; the Borrower's completion report summarised in the annex gives 961,339 beneficiaries at closing (page 58), 62 percent of target. The record therefore does not use the final-year caseload as its metric: the erosion it documents is the payment pause, which the ICR does state (metric, the three-year urban suspension, paragraph 39, PDF page 21): payments stopped during the 2022 transition to SEDESOL and the Red Solidaria, with rural payments resuming in December 2022 and urban payments after a three-year suspension. The emergency top-up was small (24,891 households) and the source gives no elapsed time for it beyond late 2020. The 2019 to 2020 rise mixes horizontal expansion with routine enrolment. A change of government that redesigns a programme is a political choice, not only a loss of capacity.
๐ท๐ผRwandasince 2021delivered and closedbears on Institutional responsiveness
After the cabinet ordered a lockdown on 14 July 2021, Rwanda's government released part of the national food reserves on 15 July, and between 18 and 30 July local governments delivered about 7,000 tonnes of food door to door to more than 327,000 families in Kigali and eight districts, including 103,000 families not on the initial list.
The cabinet's lockdown decision was paired the next day with a release from the national food reserve; sector and cell officials drew up household lists, opened a channel for unlisted families to claim by toll-free line, SMS and social media, and delivered door to door with youth volunteers.
Preconditions
A physical national strategic food reserve
Local administration down to the cell and village level able to list households within days
A claims channel so the initial list could be corrected during delivery
Where else it has been used: Lockdown food distribution was widespread in 2020 and 2021 and often failed on targeting; the claim-in channel that added roughly a third to the list is the transferable element.
Limits: The source is the implementing ministry's own news item, written as a success story with quotes from a beneficiary and the minister, and the figures are rounded ('more than 327 thousand', 'around seven thousand tones'); there is no audit, no list of families, and the page shows no publication date. It covers one 15-day lockdown in Kigali and eight of Rwanda's 30 districts, in kind rather than cash, and says nothing about who was missed, the quality of targeting or the cost. The four-day lag from the cabinet decision (14 July) to the start of deliveries (18 July) is the speed evidence, and it rests on the ministry's own dating. Earlier lockdown relief in 2020 is not covered here. One well-run emergency distribution, in a state with unusually dense local administration and mandatory community structures, does not show that Rwandan rules change quickly in general.
๐ฎ๐ฑIsraelsince 2020delivered and closedbears on Institutional responsiveness
After the Monetary Committee decided on 6 April 2020 to lend banks three-year fixed-rate money for credit to small and micro businesses, the Bank of Israel had NIS 1.2 billion outstanding under the programme by 30 April 2020 and NIS 28.1 billion by 31 March 2021.
The Monetary Committee created a dedicated funding line for banks, conditioned on lending to small and micro businesses, within weeks of the first lockdown, and reopened and extended it with cheaper terms as the crisis continued.
Preconditions
A central bank with legal room to lend to banks against a lending purpose
A concentrated banking system that can pass a funding line to small firms quickly
Policy rates near zero, so the subsidy had to come through term and price of funding
Where else it has been used: Funding-for-lending schemes of this shape ran in the United Kingdom from 2012 and at the European Central Bank as targeted longer-term refinancing operations; Israel's version was set up and drawn within one month.
Limits: The site answers scripted requests with a Radware bot-verification page; the table was read in a normal browser at the same URL, so a reviewer needs a browser to check it. The figures are the Bank reporting on its own programme. They are balances of central-bank loans to banks, not credit that reached firms and not a count of businesses: how much of the money became new small-business lending, rather than refinancing, is not shown. The programme was not continuous: the footnote says it ran first until the end of May 2020 and was renewed in August, and the balance sits at NIS 4.6 billion from May to July. Part of the March 2021 balance comes from a second pillar of four-year loans at minus 0.1 percent decided on 22 October 2020. The cited page's last row is 31 March 2021; the status rests on a later report in the same series (7 July 2021, boi.org.il/en/communication-and-publications/press-releases/monthly-report-on-programs-the-bank-of-israel-is-operating-in-the-financial-markets-in-view-of-the-coronavirus-crisis-12/), which records a Monetary Committee decision of 5 July 2021 to end the programme on 1 October 2021 or once NIS 40 billion had been extended, and a balance of NIS 37.2 billion on 30 June 2021. The same table shows the Bank's other rapid moves (repo from 15 March, dollar-shekel swaps from 18 March, government bond purchases from 23 March 2020, NIS 8.5 billion outstanding by 31 March), which are monetary operations rather than programme changes and are not the claim. The State Comptroller's 2021 COVID-19 special report found that financial regulators prepared rapidly but that timeframes for handling loan requests were not set, which concerns the separate state-guaranteed loan funds.
๐บ๐พUruguaysince 2015still operatingbears on Disaster preparedness and recovery
Uruguay's national housing directorate relocated 2,618 households from flood-prone or contaminated land between 2015 and 2020 under the Plan Nacional de Relocalizaciones and related agreements with departmental governments, against a target of 3,500 to 6,000 households by 2025.
The national housing directorate co-finances departmental relocation projects that move households from land that floods repeatedly into new or purchased homes, with multidisciplinary teams handling each move and the department responsible for keeping the vacated land unoccupied.
Preconditions
Flood-risk maps and settlement censuses precise enough to identify which households to move
Departmental governments that co-execute and police the vacated land
A national housing budget able to buy or build replacement homes year after year
Where else it has been used: Relocation out of flood plains is the slow, preventive counterpart to evacuation; it is listed in Uruguay's climate-adaptation pledge, which is how it came to be measured.
Limits: The number is in a six-page PDF: page 1 gives 'รltimo valor disponible 2.618 (2020)' and page 4 the version history (2,245 for 2019, 2,618 for 2020); pages 5 and 6 give the yearly totals 294, 442, 660, 438, 411 and 373 for 2015 to 2020, which sum to 2,618. The headline on page 1 ('A 2025 se han relocalizado al menos entre 3.500 y 6.000 hogares') is the 2025 target written as if achieved, and the sheet's own data show 2,618 by 2020, below the low end of that target; no later official value was found, so whether the target was met is not shown. The count mixes flood-prone and contaminated land and does not split them, and it includes relocations after specific floods ('Inundados 2015', 'Inundados 2017'), which are recovery rather than preparation. It counts households, not people, and says nothing about whether vacated land stayed empty, which the sheet itself names as a condition that depends on departmental enforcement. The plan predates 2015, but the sheet's series starts with the 2015 to 2019 housing plan, so 'started' is the first year counted, not the plan's creation. The sheet states the plan's objective as reversing social segregation and territorial fragmentation, so it is a settlement and housing policy as much as a risk-reduction one; it is filed here because the same sheet classes it as a climate-adaptation measure. Status 'operating' rests on the sheet naming the 2020 to 2024 five-year housing plan as the plan's current frame; no count after 2020 was found. The sheet is a government monitoring document for Uruguay's climate pledge, self-reported by the ministry that runs the programme.
๐ช๐นEthiopiasince 2024still operatingbears on Institutional responsiveness
The National Bank of Ethiopia floated the birr on 29 July 2024, and by 20 September the former official rate had moved from 57.8 to an indicative 119.2 birr per US dollar while the parallel-market spread over it fell from 96 percent to around 5 percent.
The central bank replaced an administered exchange rate with a market-determined one in a single directive, moved FX surrender from itself to commercial banks and lifted import bans and controls at the same time, while tight monetary policy held the new rate near the old parallel rate.
Preconditions
External financing (here the IMF arrangement) large enough to absorb the transition
A central bank able to tighten liquidity enough to stop the float turning into a spiral
Commercial banks able to quote and report FX rates daily
Where else it has been used: Egypt (2016) and Nigeria (2023 to 2024) unified rates under similar pressure; Cuba's 2021 unification (cub-ordenamiento) fragmented again, which is what the widening to 16 percent in October 2024 warned of here.
Limits: The numbers are in paragraph 6 of the staff report ('moving from 57.8 birr per US dollar on July 26 to an indicative rate of 119.2 on September 20 ... The parallel market spread to the indicative rate collapsed from 96 percent to around 5 percent over the same period'); the eLibrary HTML has no page numbers. The 5 is approximate as published. What responded fast was the rule change once decided, not the decision: the report estimates the overvaluation at 52 percent in 2022/23, so the distortion was years old, and the float was a prior action for the IMF arrangement approved on 29 July 2024, while Ethiopia was seeking a debt restructuring, with calls to Eurobond holders held from December 2023. Responsiveness here is a government acting under financing pressure and external conditionality, not on its own timetable. The same paragraph says FX liquidity stayed limited and the interbank market inactive, and the second review (Country Report 25/027) records the premium widening to a peak of 16 percent on 22 October 2024 before falling back to single digits. The IMF is a party to the programme and its staff assessment is not an independent evaluation; the costs of the depreciation (inflation pass-through, the burden on importers and households) are not shown by this number.
๐ง๐ดBoliviasince 2021delivered and closedbears on Disaster preparedness and recovery
Bolivia's health authorities reported to WHO that 7,505,389 people, 64 per 100 population, had received at least one COVID-19 vaccine dose by 29 December 2023, from a first dose given on 30 January 2021, with 6,307,628 (54 per 100) completing the primary series.
Bolivia's health ministry ran a national COVID-19 campaign through its routine immunisation network and mass vaccination points, and reached a majority of the population with a first dose within three years.
Preconditions
A routine immunisation programme with a cold chain reaching rural municipalities
A supply of doses large enough for a national campaign, however obtained
Registers good enough to count who received a second dose
Where else it has been used: Cuba (cub-covid-vaccination) reached far higher completion on domestic vaccines and Venezuela (ven-covid-vaccination) a similar partial reach; the Bolivian case sits between them, with a smaller gap between first dose and completion than Venezuela's.
Limits: The figures are Bolivia's own reports to WHO (DATA_SOURCE 'REPORTING'), with no independent coverage survey behind them. Nearly half the population never completed the primary series (54 per 100), first-dose reach was 64 per 100, and boosters reached 2,521,192 people, 22 per 100, so the record shows a partial delivery, well below the near-universal reach of the strongest campaigns; it is filed because the campaign did reach a majority nationally, on the same terms as ven-covid-vaccination (50 per 100 completed). The VACCINES_USED field is empty, so the record cannot show which products were used or where the doses came from, and evidences distribution only. The per-100 rates divide by WHO's population estimate. Vaccination is a response to one shock, not proof of general preparedness, and the file was last updated in December 2023. The source is a CSV with no prose; the numbers are read from one row. The WHO dashboard address covid19.who.int/who-data/vaccination-data.csv now returns an HTML page, so the cited URL is the blob the dashboard serves, the same one ven-covid-vaccination cites.
๐ต๐ฐPakistansince 2020delivered and closedbears on Institutional responsiveness
Pakistan's government launched Ehsaas Emergency Cash in early April 2020 on BISP's existing payment and registry systems and had paid US$464 million to 6.4 million beneficiaries by 30 April 2020, reaching 15 million beneficiaries in all.
6.4 million beneficiaries, beneficiaries paid emergency cash by 30 april 2020, within about four weeks of launch, 2020-04-30. 15 million beneficiaries, beneficiaries given temporary income support through ehsaas in 2020, 2020. World Bank, COVID-19 G2Px country brief: Pakistan, retrieved 2026-10-02.
The move, and what it needed
The federal government raised the eligibility cutoff on an existing proxy-means-test registry, added an SMS application channel, and pushed one-off payments through BISP's contracted payment providers, so a new caseload was paid without building a new delivery system.
Preconditions
A national socio-economic registry covering most households, with a proxy means score that could be re-cut
A biometric-verified payment network already contracted for a standing cash transfer
National ID and mobile coverage high enough for SMS self-registration
Limits: Both figures are government-reported disbursement counts passed through World Bank documents, not independent verification that the money reached eligible families; the brief does not report inclusion or exclusion errors. The two World Bank documents disagree on the launch month (early April in the brief, March in the PAD) and do not reconcile the 15 million with the 12 million originally targeted. The transfer was one-off and small, about US$71 per family, so it shows speed of reconfiguration, not depth of protection. The speed rested on assets built over a decade (the BISP payment network and the national socio-economic registry), so the record evidences a programme being changed fast, which is the construct, but credits the 2020 government with infrastructure its predecessors built. The 'four weeks' is our reading of the brief's dates, not a published duration.
๐ง๐ฉBangladeshsince 1973still operatingbears on Disaster preparedness and recovery
Before Cyclone Remal made landfall on 26 May 2024, Bangladesh's government-run cyclone early warning and evacuation system moved about 0.8 million people into 9,424 evacuation shelters across 19 districts. The UN-coordinated response plan records 16 deaths among 4.6 million people affected.
The government took over the cost of a Red Crescent volunteer warning network in 1973 and ran it jointly with the Bangladesh Red Crescent Society. Village volunteers relay meteorological department signals door to door and move people into shelters before landfall.
Preconditions
A meteorological service able to issue track forecasts days ahead of landfall
A shelter stock within walking distance of exposed coastal villages
A volunteer corps, trained and equipped, kept active in years without a major storm
Where else it has been used: The volunteer-relay model shaped community early warning in Odisha and Myanmar. Copies tend to keep the sirens and drop the standing volunteer corps, which is the part that turns a warning into an evacuation.
Limits: The evacuation and death figures are the government's own preliminary counts, repeated in a UN appeal document whose job is to raise funds. They are not an independent audit. The figures cover one cyclone, and Remal's death toll depended on its track, timing and surge as well as on preparedness. The plan credits the Cyclone Preparedness Programme with helping develop the warnings, but it does not attribute the evacuation to it alone: district administrations, shelters built under other programmes and anticipatory-action partners all took part. The plan states no baseline, so the record shows one storm's outcome and not a trend. The same document records 4.6 million people affected and 173,866 houses damaged or destroyed, so the system protected lives far more than it protected housing and livelihoods. The construct also covers recovery, which this number does not show. The figure sits inside a PDF.
๐บ๐ฆUkrainesince 2023still operatingbears on Disaster preparedness and recovery
Ukraine's government, through the Ministry for Development and the Diia digital platform, paid war-damage housing compensation to 231,681 families totalling UAH 119.1 billion by 10 July 2026. Of these, more than 153,000 families received repair payments, and more than 74,000 received certificates to buy replacement housing for homes that were destroyed.
The Rada created a statutory right to compensation for war-damaged housing. The ministry and the Ministry of Digital Transformation then ran claims, damage inspection and payment through the Diia app and the state register of damaged property, paying repair grants to bank accounts and issuing purchase certificates for destroyed homes.
Preconditions
A national digital identity and services app that most households already used
Property and damage registers that could be linked to verify claims
External concessional finance willing to fund payouts through government systems during a war
Limits: These are the ministry's own running totals, published as press releases and not as a statistical series, and nothing independent audits them. The page shows recovery for households whose claims were accepted, not the share of all damaged housing compensated. For destroyed homes, more than 161,000 applications produced certificates for about 74,000 families, and only about 33,000 had bought a home, so most of the destroyed-housing caseload is unresolved. The page ties the repair payments to the World Bank HOPE project and the own-plot rebuilding payments to the Council of Europe Development Bank HOME project, and does not say who funds the certificates for destroyed homes, so how much of the delivery rests on external money is not shown here. The record evidences recovery after a war shock and says nothing about preparedness before one. Amounts are nominal hryvnias. The figure is the ministry's results update of 10 July 2026.
๐ฒ๐ฆMoroccosince 2020still operatingbears on Disaster preparedness and recovery
Morocco's dual catastrophe insurance regime under Law 110-14 took effect in January 2020. It offered private cover to close to 9 million insured people and set up the FSEC, financed by a parafiscal levy, for the uninsured. After the September 2023 Al Haouz earthquake, the government deployed the FSEC for the first time, and it mobilised US$300 million to compensate uninsured victims.
The Ministry of Interior and the Ministry of Economy and Finance legislated a two-tier regime. Private insurers carry catastrophe cover for people who already hold policies. A public fund, paid for by a parafiscal levy and backed by sovereign reinsurance, gives basic compensation to the uninsured once the government declares an event a catastrophe.
Preconditions
An insurance market with a policy base wide enough to carry catastrophe cover, and a levy base to fund the public tier
An Interior ministry able to declare a catastrophe and register uninsured victims through local agents and an online portal
International reinsurance capacity and World Bank technical and financial backing for the design
Limits: Both sources are the World Bank reporting on reforms it financed: an FSEC programme-for-results loan and a Cat DDO whose prior actions were these reforms. The US$300 million is money mobilised, not money paid to households. The ICR says the compensation process was still under way in June 2024 and gives no count of victims compensated. Nothing here shows what share of the earthquake's losses the regime covered. The 9 million and 6 million figures are coverage estimates from a 2022 results brief, not counts of policies or payouts. The regime's one real test is a single earthquake. One of the Cat DDO's own result indicators was missed by a wide margin: Interior ministry officials trained on victim registration stood at 7.4 percent against a 90 percent target. The audit of the FSEC's internal controls had not been completed. The two World Bank documents disagree on the law's date: the ICR says Law 110-14 was promulgated in 2016, while the brief says it was adopted in 2018. Started is therefore the year the regime took effect. The headline figures are read from a 45-page PDF and a web page.
๐ช๐ฌEgyptsince 2015still operatingbears on Institutional responsiveness
Egypt's Ministry of Social Solidarity launched the Takaful and Karama cash transfers in early 2015 and expanded coverage from an initial 1.5 million to 4.675 million households (about 17 million people) by project close in 2024, revising the eligibility threshold and adding top-ups in response to COVID-19 and later inflation, while the cash transfer budget rose from EGP 3.5 billion in 2015 to EGP 41 billion in 2024.
The Ministry of Social Solidarity built a proxy-means-tested registry and a cash transfer programme with World Bank financing, then used the registry's threshold and benefit levels as the levers it moved when COVID-19, inflation and devaluation hit.
Preconditions
A national social registry, which reached 11.3 million registered households, to retarget against
A payment rail through post offices and later Meeza cards that could absorb more beneficiaries without a new system
Fiscal room created by the 2014 energy subsidy reform, which the programme was launched to offset
Limits: The construct is the speed of rule and programme change, and the headline number is coverage, which is scale rather than speed. The ICR describes the COVID-19 response (a revised proxy-means-test threshold, wider coverage, Meeza cards, top-ups) in prose without dates or a count of households added by the revision, so how fast the rules moved is not measured here. The 4.675 million includes legacy Daman social pension beneficiaries absorbed into the programme by consolidation, so part of the growth is relabelling rather than new reach. The budget figure is nominal Egyptian pounds across years of high inflation and two large devaluations, so the 12-fold rise overstates the real increase. Coverage still reaches only about 28 percent of the poorest quintile. The source is the World Bank's own completion report on a project it financed with US$900 million, read from a PDF, and the counts are administrative figures reported by the ministry.
๐ฎ๐ถIraqsince 2015still operatingbears on Disaster preparedness and recovery
From September 2015 Iraqi ministries, governorates and the Reconstruction Fund for Areas Affected by Terroristic Operations used two World Bank loans totalling US$750 million (IBRD-85200, US$350 million, and IBRD-87930, US$400 million), of which US$610 million was disbursed by 30 November 2024, to rebuild water, power, roads, bridges and railway in five governorates, reaching 5 million direct beneficiaries by March 2024, with the first loan's closing date pushed six years past the original June 2020.
The federal government channelled World Bank loans through line ministries, governorates and a dedicated reconstruction fund, rebuilding basic services in retaken cities sub-project by sub-project rather than through one master plan.
Preconditions
Military recovery of the territory before any rebuilding could start
Sovereign borrowing capacity and an international lender willing to extend the closing date repeatedly
A reconstruction fund and planning ministry able to register and approve hundreds of sub-projects
Limits: The figures come from a World Bank supervision report on its own loan, compiled from implementing agencies' reporting; procurement and M&E are rated Moderately Unsatisfactory in the same report, and the beneficiary count is a 'direct beneficiaries' definition that counts everyone in a served area, so it can overlap across water, power and road indicators. The project is still open, so nothing here is a completion figure, and the 72-month extension is our computation from the table, not a published delay figure. The shock is war, not a natural hazard, and the record evidences recovery, not preparation. The money is World Bank lending the government must repay, implemented by Iraqi agencies under Bank supervision, and the US$68 million repurposed to railway works in 2024 shows the scope moved with time. It covers five governorates, not the country, and the report gives no measure of whether rebuilt services were kept running after handover.
๐ฎ๐ทIransince 2010operating below its peakbears on Institutional responsiveness
Iran's government replaced universal fuel and bread subsidies with a universal cash transfer of 455,000 rials (about US$40) per person per month from late 2010, then left the transfer unindexed, so that by 2019 its purchasing power had fallen to less than US$6 per person before a top-up was added in late 2019.
The Iranian government raised domestic energy and bread prices in one move and paid the savings back as a flat universal cash transfer. It then left the transfer nominal, so inflation dissolved the benefit while energy prices drifted back toward subsidy.
Preconditions
A national population and household register able to pay nearly every resident within months
Large implicit energy subsidies whose removal freed fiscal room for a transfer
A fiscal position, later hit by sanctions and inflation, that made indexing the transfer unaffordable
Where else it has been used: Cited by the World Bank as a model for Angola's 2023 fuel subsidy reform (Policy Research Working Paper 10939). The retellings stress the 2010 compensation and rarely the later erosion.
Limits: The World Bank restates every figure here from secondary sources (Salehi-Isfahani et al. 2015, Enami and Lustig 2018, UNICEF 2019). It does not compute them, and no Iranian official series is cited. The two dollar figures are not comparable: US$40 converts at the 2010 rate, the 2019 figure is undated within the year and published only as 'less than US$6', and the rial's collapse after 2018 inflates the fall in dollar terms. The 'nearly in half' real decline from 2012 to 2016 is the cleaner measure, but it is published as prose and not as a number. The reversal is partial. The transfer still reached about 90 percent of households, and the government added a top-up in late 2019 and COVID-19 payments in 2020. The construct asks how fast rules change. This record shows a fast rule change in 2010 followed by about nine years in which the transfer was not changed as inflation rose. It is evidence of the second half, not a measure of speed. The figures are read from a PDF.
๐ฎ๐ทIransince 2002delivered and closedbears on Disaster preparedness and recovery
After the April 2002 earthquake in Kermanshah and the June 2002 earthquake in Qazvin, Zanjan and Hamedan, Iran's Ministry of Interior project unit and the Housing Foundation repaired or rebuilt about 24,000 damaged houses to seismic standards by June 2009, using government funds and a US$180 million World Bank loan, of which US$174.9 million was disbursed.
Iran's Ministry of Interior set up a project unit and handed housing reconstruction to the Housing Foundation, which tied reconstruction materials to compliance with seismic codes and trained local engineers and masons to enforce them.
Preconditions
A standing national housing body (the Housing Foundation) with rural reach and a reconstruction mandate
Government funds already committed before external finance arrived
Access to multilateral lending, which later sanctions closed off
Where else it has been used: The Housing Foundation reused the model of owner-driven rebuilding to code after Bam in 2003, with a World Bank loan approved in 2004.
Limits: The figure is self-reported by the Bank and the borrower's project unit in a completion report, and no independent audit of the houses is cited. The report contradicts itself: the results table gives 24,000 actual against a 24,200 target, the narrative says around 24,200, and a footnote gives an appraisal estimate of 29,232. The project financed materials (steel, cement and equipment) for reconstruction rather than building every house, so the count is of houses it supplied. About 59 percent of the 40,727 damaged houses are covered. The response was regional, in four provinces. It was not national, and it is not the Bam 2003 reconstruction that Iranians would name first. Schedule slipped: the loan closing moved from June 2007 to June 2009, and the indicator was achieved six months after its target date. The disaster insurance component was dropped from Bank financing at the government's request. The evidence of preparedness is limited to seismic-code training for 350 staff against a target of 200 and a public awareness campaign, with no later test against a subsequent earthquake. The figures are read from a PDF.
๐ฆ๐ดAngolasince 2015operating below its peakbears on Institutional responsiveness
Angola's government removed gasoline and diesel subsidies by raising retail prices at the end of 2015, but subsidies returned as costs rose, reaching 3.7 percent of GDP in 2023 from 1.5 percent in 2019. A second reform raised gasoline prices 87.5 percent in June 2023, but the kwanza's depreciation took the gasoline subsidy from 39 percent of the market price in June back to 65 percent by October 2023.
Angola's government raised administered fuel prices in discrete steps, in 2015 and again from June 2023, while leaving them fixed in kwanza between steps. Each devaluation or oil price rise then rebuilt the subsidy the step had removed.
Preconditions
Administered retail fuel prices set by decree with no automatic adjustment
A currency exposed to oil price shocks and sharp devaluation
Fuel subsidies acting as the de facto social spending before a cash transfer (Kwenda) could reach the poor
Where else it has been used: The 2023 reform drew on the 2010 Iranian and Indonesian lessons cited in the same paper: compensate before raising prices. Kwenda was meant to be that compensation, but the same paper says it has failed to reach enough people.
Limits: This is a World Bank working paper, which disclaims representing the Bank's view. Every subsidy share is the authors' own estimate from IGAPE and INE data, not a published government series. No figure is given for the 2015 removal itself: the source only says subsidies were removed and then reemerged. The 2019 to 2023 rise is in part an oil price and exchange rate effect, so it measures exposure as much as inaction. The October 2023 figure reflects a 40 percent depreciation in May and June 2023, which no price rule could have tracked. The government did act again: the gasoline rise in June 2023, a 1 June 2023 decree setting out a multiphase reform, and a diesel rise from 135 to 200 kwanzas in April 2024. Between the two steps retail prices remained constant until May 2023, and the June 2023 decree was the first since 2016. The figures are read from a PDF.
๐ฒ๐ฟMozambiquesince 1999still operatingbears on Disaster preparedness and recovery
Mozambique's National Disaster Management Institute, created in 1999 after a new national disaster policy, evacuated around 150,000 people with two days of warning in the 2007 floods, and by 2011 had ten days of warning for an event of similar scale and spent under US$1 million on evacuation, against about US$5 million in 2007.
The government replaced a relief department with a standing disaster agency, INGC, gave it a national emergency operations centre and a civil protection unit, and paired forecasting with district and community risk committees that carry warnings to people who then move.
Preconditions
A flood catastrophe in 2000 that made preparedness a political priority
Hydrological and meteorological forecasting with enough lead time to act on
Donor funding and an annual contingency budget line for evacuation
Limits: Every figure here is a round number in a GFDRR case study, a conference version that says it aims to learn from country experience rather than evaluate it and carries a disclaimer on accuracy; the evacuation counts and costs are INGC's own, with no underlying series published. 'Similar in scale' is the case study's judgement, not a measured comparison of hazards. The same document's comparison of the 2000 and 2013 floods (700 and 117 deaths) does not show a falling death rate once divided by the population affected (4.5 million and 478,892, our computation), and its own text gives 30 direct deaths for 2013, so mortality is not used as the metric. The record says nothing about the cyclone seasons after 2014. Much of the system was built with donor money.
๐ฒ๐ฌMadagascarsince 2016delivered and closedbears on Disaster preparedness and recovery
Madagascar's Fonds d'Intervention pour le Developpement turned its safety net project into a shock-response channel from December 2016, delivering crisis-response unconditional cash transfers to 622,961 people and crisis cash-for-work to 155,649 people after droughts in the South and the 2022 cyclones, by the project's close in October 2024.
FID, a government agency with field staff and a beneficiary registry from the regular safety net, repeatedly restructured the project to redirect funds to drought and cyclone victims, paying unconditional transfers and cash-for-work through the channels it already ran.
Preconditions
An implementing agency with a targeting method and payment channels already operating before the shock
A financier willing to restructure a project within weeks rather than design a new one
Enough security in affected districts to move cash, here secured by an agreement with the police
Limits: The figures are cumulative over 2016 to 2024 and come from the World Bank's own evaluation of a project it financed, which draws on the implementing agency's reporting; they count people reached, not whether households recovered, and say nothing about speed of delivery after each shock. Almost all the money was IDA credit and grants (US$305.9 million actual), so the record shows the state's agency delivering donor-financed transfers, not domestic fiscal capacity to respond. Targets were set and raised repeatedly in emergency conditions, and IEG notes that some reach targets were not scaled up as financing grew (its remark about targets set quickly and with limited clarity concerns the emergency nutrition targets), so exceeding them is weak evidence. The response was reactive, reallocating funds after each shock rather than drawing on a pre-arranged trigger, which is recovery more than preparedness. Security problems in the South forced FID to sign an agreement with the police to deliver cash. The project closed in October 2024, so the status is concluded.
๐ณ๐ตNepalsince 2016delivered and closedbears on Disaster preparedness and recovery
After the April 2015 Gorkha earthquake, Nepal's National Reconstruction Authority ran an owner-driven housing grant programme that, by 23 September 2021, had paid the third and final tranche of the reconstruction grant to 704,341 of 866,207 eligible households across 32 districts, with the World Bank-financed share revised to 309,335 houses.
Parliament created a time-limited National Reconstruction Authority that enrolled every damaged household, paid a flat grant in three tranches, and released each tranche only after engineers verified that the house met a seismic building code, so the payment schedule enforced the standard.
Preconditions
A damage census that enumerated eligible households district by district before money flowed
Concessional and pooled donor finance large enough to fund a uniform grant for most of the caseload
Enough trained masons and inspecting engineers in remote districts to verify construction stage by stage
Where else it has been used: The tranche-on-verification model is the owner-driven reconstruction approach also used after the 2005 Pakistan earthquake; Nepal's version is distinguished by its scale and by tying each tranche to a code-compliance inspection.
Limits: A grant tranche is a payment, not a finished house: in the programme's rules the third tranche is released on completion of construction up to the roof band, so 704,341 counts households verified to that stage, not occupied, completed homes. The figures are reported by the government agencies to the lender and published by the lender, which financed about $700 million of the programme and rates its own operation; the 309,335 count was itself revised in 2023 after the project clarified that its reported total had included 3,766 houses financed by a trust fund grant (TF0A4783). About 162,000 eligible households had not received the final tranche at the reporting date, and the World Bank's ICR for the MDTF Funding for Earthquake Housing Reconstruction project (P162067, TF0A4783) notes the average rebuilding cost was about three times the NPR 300,000 grant, and one third of households used loans to close the gap. The number shows recovery after one shock; it says nothing about preparedness before the next one, and the NRA was a temporary body dissolved in December 2021, with its residual work passed to the National Disaster Risk Reduction and Management Authority. The record documents the national programme's scale through a single donor's monitoring report; there is no independent national statistical series behind it.
๐ฟ๐ฒZambiasince 2024delivered and closedbears on Disaster preparedness and recovery
After the President declared the 2024 El Niรฑo drought a national disaster, Zambia's Ministry of Community Development and Social Services enumerated households in all 116 districts within four to eight weeks. From May 2024 to April 2025 it paid a drought emergency cash transfer to 1,631,007 households in 84 districts through the Social Cash Transfer, and paid a nationally funded price-shock transfer to 632,664 households in the other 32 districts.
Once the drought was declared, the Cabinet chose the existing Social Cash Transfer instead of a parallel humanitarian pipeline to deliver emergency aid. The Ministry of Community Development and Social Services then used enumerators already on retainer in every district and a simplified proxy-means test to enrol drought-hit households in weeks and pay them through the same digital rails.
Preconditions
A national cash transfer already running in all 116 districts, with a beneficiary information system and digital payments
Enumerators kept on retainer and a targeting questionnaire that could be adapted in weeks
Donor crisis financing available at short notice to fund the expansion while the treasury was in debt restructuring
Limits: The source is the World Bank's own completion report on a project it financed, written with the implementing ministry, and the household counts are programme administrative data, not an independent survey. The 1.63 million drought transfers were financed by IDA grants and a donor trust fund. Only the 632,664 price-shock transfers were paid from national funds, so the record shows government delivery capacity more than government financing. Payment timeliness fell to 17 percent in the first half of 2025, at the height of the drought, and averaged 29 percent over the project against a 90 percent target. Households counted as supported were not necessarily paid on time. The consumption and coping effects the report cites come from evaluations it says should be interpreted cautiously. The report's own footnote 51 puts enumeration at about eight weeks, against four weeks in the main text. The report dates the disaster declaration to 28 February 2024 in one place and 29 February in another. The record shows a response after a shock was declared. It does not show forecasting or contingent financing in place before the drought, and the report says the government began strengthening disaster risk financing, with contingent financing instruments, only from 2024, after the drought. The transfer ended in April 2025 and the caseload returned to its pre-drought level, so nothing here shows recovery outcomes.
๐ฌ๐ณGuineasince 2021delivered and closedbears on Disaster preparedness and recovery
Guinea's Ministry of Health announced an Ebola outbreak in N'Zรฉrรฉkorรฉ Prefecture on 14 February 2021 and declared it over on 19 June 2021. It activated national and district emergency committees, vaccinated 10,873 people and followed 1,031 of 1,110 contacts daily, and held the outbreak to 23 confirmed and probable cases and 12 deaths, five years after the 2014 to 2016 West African epidemic.
Guinea's Ministry of Health activated national and district emergency management committees within days of the announcement. It deployed case-search, contact-tracing, burial and community-engagement teams, set up PCR testing in N'Zรฉrรฉkorรฉ, and ran ring vaccination of contacts and frontline workers with WHO and partner support. It declared the outbreak over after the 42-day count.
Preconditions
Response structures, trained staff and community networks left in place from the 2014 to 2016 epidemic
A licensed Ebola vaccine and therapeutics that partners could ship within days
A laboratory able to confirm cases near the outbreak, here a PCR capacity stood up in N'Zรฉrรฉkorรฉ
Where else it has been used: The ring-vaccination and emergency-committee model from the 2018 to 2020 eastern DRC response, applied in a country that had lived through the 2014 to 2016 epidemic.
Limits: This is one outbreak in one prefecture, so it shows a response to one shock and not a standing national capacity, and a single outcome cannot separate preparedness from luck. Detection was slow: WHO reports the index case, a health worker, fell ill on 15 January and died on 28 January after visiting two health facilities and a traditional practitioner. She was buried without safe burial practices, and the outbreak was announced only on 14 February. Five cases were health workers, which points to infection-control failures. A second cluster of community deaths surfaced in late March after more than three weeks with no cases. One confirmed case was lost to follow-up from 1 April until 18 June, and 79 of the 1,110 contacts were not followed daily. WHO notes that only nine (39%) of the 23 cases were known contacts at detection, and that the alert system reported a suboptimal number of alerts. The response was not purely national: WHO reports the MoH acted together with WHO and other partners, and their contribution cannot be separated from that of the national institutions. The figures are WHO restating Ministry of Health reports, so the government is reporting on its own response. The page gives no figures for the 2014 to 2016 epidemic, so the record makes no quantitative comparison with it.
๐ฒ๐ผMalawisince 2019still operatingbears on Disaster preparedness and recovery
From 2019 the Government of Malawi identified every household it registered for shock-responsive cash transfers through its Unified Beneficiary Registry. It expanded the Social Cash Transfer Programme to give drought and COVID-19 aid to more than 300,000 rural households, and it drew US$30 million of pre-arranged contingent financing on 10 April 2020, three weeks after declaring a state of disaster.
Malawi's Ministry of Finance and its social protection agencies built one registry of rural households and set protocols that every shock-response transfer must use it. They paired the registry with financing pre-arranged before a disaster (a contingent credit line, a project contingency window and sovereign drought insurance), so cash could reach households without a new appeal each time.
Preconditions
A census-style household registry covering most of the rural population and kept current
Contingent financing agreed before the shock, which here depended on concessional lenders and an insurance pool
A legal trigger, a declared state of disaster under the 1991 Disaster Preparedness and Relief Act, that releases the funds
Limits: The source is the lender's self-evaluation of its own policy operation, and the 300,000 is a floor ('more than') summed over about four years and two kinds of shock. It is not a per-event count and has no denominator of affected households. The money behind the transfers was largely external: a World Bank project's contingency window, ARC sovereign drought insurance whose premiums donors financed in full (ICR Annex), and IDA's Cat DDO. The evidence is that the national system routed aid, not that Malawi financed it. The Cat DDO was triggered by a pandemic, not a natural hazard, and its US$30 million was budget support with no traceable link to households. The ICR says fewer than 10 percent of the households receiving transfers were registered before the protocols. It rates the operation moderately satisfactory and reports that resilient transport standards were never developed. The source is dated June 2023; the operation itself closed in 2022. Cyclone Freddy (February and March 2023) and the 2024 El Nino drought fell after the operation closed and are not assessed. 'Operating' assumes the registry and the scale-up mechanism are still in use, which this source does not confirm beyond 2023.
๐ฒ๐ฌMadagascarsince 2020delivered and closedbears on Institutional responsiveness
Madagascar's social protection agency FID designed the Tosika Fameno emergency cash transfer after the March 2020 urban lockdown, registered almost 200,000 households in about 12 days, made the first payment four weeks after the lockdown began, and reached more than 368,000 urban households in ten cities with 100,000 ariary payments.
The President asked the national social protection fund FID to extend its rural cash transfer machinery to the locked-down cities; FID and WFP built a rapid registration process, paid by mobile money and post, and partners aligned their transfers on the same amount and targeting.
Preconditions
A national social fund already running cash transfers and a payment chain
Donor finance that could be redirected within weeks
Mobile money and postal networks able to pay urban households
Where else it has been used: Madagascar later turned the urban response into a longer urban safety net under the World Bank project; similar lockdown transfers in Malawi (mwi-cuci) took ten months to pay, which shows how much the existing agency mattered.
Limits: The figures are in an eight-page UNICEF case study PDF (PDF page 4: first payment four weeks after the start of the urban lockdown, registration of almost 200,000 households in about 12 days, more than 368,000 households; PDF page 6: registration completed two weeks after the declaration of the state of health emergency, first payments within one month; the 368,000 by the end of 2020 also on PDF page 2). UNICEF co-led the cash working group and directly supported 8,500 of the households, so it is grading a response it helped build, and the case study gives no exact dates for the lockdown or the first payment. The 368,000 mixes government payments financed by the World Bank with humanitarian top-ups by UNICEF, WFP, the EU and NGOs using the same parameters; the World Bank's Social Safety Net Project ICR (2025, P149323, paragraph 43 and annex note 2) counts 215,093 households paid under its own financing, a smaller number on a different scope. The benefit was 100,000 ariary (about US$26), small against lost urban income, and coverage was limited to ten cities. The speed came from an agency that already ran rural safety nets, not from a standing urban programme.
๐ฒ๐บMauritiussince 2020delivered and closedbears on Institutional responsiveness
Mauritius made its revenue authority, the MRA, a payer of COVID-19 income support in March 2020, and by end June 2020 the MRA had disbursed Rs 9,985 million under the Wage Assistance Scheme (Rs 7,603 million) and the Self-Employed Assistance Scheme (Rs 2,382 million) to some 500,000 workers and self-employed people, after receiving 176,000 online SEAS applications within 10 days of the lockdown.
9,985 Rs million, wage assistance scheme and self-employed assistance scheme disbursements, march to june 2020, 2020-06. 176,000 applications, online seas applications received within 10 days during the lockdown, 2020. Mauritius Revenue Authority, Annual Report 2019/20, retrieved 2026-10-02.
The move, and what it needed
The government handed income support to the tax authority, which built online application forms, verified wage claims against employers' PAYE and pension returns, and paid employers and self-employed people directly into bank accounts.
Preconditions
A tax authority with near-universal electronic filing and employer payroll returns
Bank accounts for most workers, including informal ones
A fiscal buffer large enough to pay a month's wages for most of the private sector
Where else it has been used: The same tax-authority route was used by Australia's JobKeeper and Ireland's wage subsidy; in Mauritius it was extended to the informal self-employed, who had no prior payroll record.
Limits: The MRA reports on its own delivery. The 9,985 is the sum we computed from Table 3 of the annual report (Rs 7,603 million WAS plus Rs 2,382 million SEAS, March to June 2020; the text rounds it to Rs 10 billion); the 500,000 recipients is the Director General's 'some 500,000 workers and self-employed individuals', paid 'directly or indirectly', so wage subsidies reached employees through their employers and the count is approximate. The 176,000 applications in 10 days is a pull quote, not a table, and the report gives no time to first payment. The scheme rules date the speed: the MRA communique announcing the Wage Assistance Scheme is dated 23 March 2020 (mra.mu communique list, WageAssistanceScheme230320.pdf), and the first SEAS payment covered the period from 16 March 2020, but the annual report itself gives no date for the first payment. Mauritius had a complete PAYE and pension contribution register to verify claims against, which most countries in this group did not. The report does not show fraud, errors or employees laid off despite the scheme, and the WAS was extended for tourism-linked firms into 2021, so this record covers only the first four months.
๐ฑ๐ธLesothosince 2020delivered and closedbears on Institutional responsiveness
After Lesotho declared a drought emergency in October 2019, the government and the World Bank restructured the Social Assistance Project in January 2020 to reallocate US$7.4 million, and between February and September 2020 the Ministry of Social Development paid more than 40,000 Child Grants Programme households a LSL 500 monthly top-up for six months and added about 10,000 new households, which it kept on the programme at its own cost.
The social development ministry used its national household registry (NISSA) and the Child Grants Programme to add a temporary top-up for existing recipients and enrol new poor households in drought districts, financed by restructuring an existing World Bank credit instead of waiting for new money.
Preconditions
A household registry covering every rural council before the shock
A standing cash transfer with pay points in the affected districts
A lender willing to reallocate an existing loan within three months
Where else it has been used: Lesotho used the same route after the 2016 drought (27,000 recipients topped up in 2017, per the same ICR); NGOs also used NISSA to target their own responses.
Limits: The figures are in a PDF (the restructuring in paragraph 39 on PDF page 22; the top-up figures in paragraph 62 on PDF page 31). The World Bank financed the top-ups and rates its own project. The elapsed time is about four months from the October 2019 declaration to the first top-ups in February 2020, which we computed from dates in the report; the report gives no date for the first payment, only the February to September window. The results framework counts 67,000 beneficiaries of emergency top-ups against a target raised to 50,000, a different unit from the 40,000 households in the text, and the extra reach is partly attributed to exchange rate movements during COVID-19. Fewer than 10 percent of payments were digital, so most top-ups went out at cash pay points. The government paid the base benefit and absorbed the 10,000 new households; the top-ups themselves were donor money.
๐ฒ๐ผMalawisince 2020delivered and closedbears on Institutional responsiveness
Malawi declared a State of Disaster over COVID-19 on 20 March 2020 and enrolled 194,762 eligible urban households in a three-month cash transfer in four cities, but first payments came only after January 2021, and by 30 June 2021 just 32,000 households, 16 percent, had received all three payments while 29 percent had received none.
The social welfare ministry designed a new urban transfer from scratch, registering households door to door in hotspots and paying by mobile money, with no prior urban registry to draw on.
Preconditions
An urban beneficiary registry, which Malawi did not have, so registration took most of 2020
Mobile money accounts already verified in the recipients' names
Agent networks with enough cash for a sudden mass withdrawal
Where else it has been used: Madagascar's Tosika Fameno (mdg-tosika-fameno) paid in four weeks through an agency that already ran transfers; the contrast is the evaluation's own lesson about building delivery systems before the shock.
Limits: The figures are in a PDF (State of Disaster date on PDF page 19, printed page 18; Table 8 and the January 2021 payment plan on PDF page 41). The World Bank co-financed the intervention with KfW, the EU and others, and commissioned the evaluation, but its own report documents the delay rather than hiding it. The record is evidence of slow responsiveness: about ten months from the State of Disaster to the first planned payment, a span we computed from the report's dates, and 15 months later only one in six households was fully paid. The report attributes the lag to know-your-customer failures between registration data and mobile network operator records and to agent liquidity, not to political choice. The 199,000 households cited elsewhere in the report is the enrolment target; the 194,762 is the eligible count in Table 8. Payments continued after June 2021 and the final completion figure is not in this report.
๐บ๐ฌUgandasince 2020delivered and closedbears on Institutional responsiveness
The Bank of Uganda gave supervised financial institutions exceptional permission to restructure the loans of borrowers hit by COVID-19 from 1 April 2020, and banks restructured UGX 2.3 trillion of loans in April 2020 alone and UGX 5.9 trillion by July 2020, approving 893,018 of 895,241 applications.
The central bank suspended the usual classification and provisioning rules for restructured loans for 12 months, so banks could reschedule borrowers' repayments without eating their capital, and paired it with a liquidity facility for distressed lenders.
Preconditions
A banking system well capitalised enough to carry deferred repayments
A supervisor with power to waive its own prudential rules by circular
Credit reference and reporting systems to track restructured loans
Where else it has been used: The same forbearance was used across East Africa in 2020: the report itself notes that Rwanda had the highest restructured-loan ratio in the region at 39.0 percent at end June 2020.
Limits: The live Bank of Uganda site is now a JavaScript application and the original PDF path returns a web page, so the source is the Internet Archive copy of the central bank's own report, fetched today. The figures are in Box A on PDF pages 27 to 29 (printed pages 19 to 20) and in the policy summary. The central bank is the rule-maker and the reporter. Restructuring is forbearance, not new money: it postponed repayments and recognition of bad loans, and the same report warns that NPLs could materialise when the relief matured; the 5.9 trillion is a flow of restructurings, while the stock under relief at end June was UGX 4.9 trillion, 31.2 percent of loans. Granting relief was at each bank's discretion, so the approval rate reflects applications banks chose to process. The report dates the relief window from 1 April 2020 but not the decision itself, so the time from shock to rule is not shown here.
๐บ๐ฌUgandasince 2022delivered and closedbears on Disaster preparedness and recovery
Uganda's Ministry of Health contained the Sudan ebolavirus outbreak declared on 20 September 2022, for which no licensed vaccine existed, through national and district emergency committees, listing 4,793 contacts, and declared it over on 11 January 2023 after 142 confirmed and 22 probable cases and 77 deaths.
The health ministry activated standing national and district emergency committees, traced and followed contacts, and isolated and treated cases in dedicated treatment centres.
Preconditions
Standing outbreak response structures from earlier Ebola and Marburg outbreaks
A national reference laboratory able to confirm cases fast
District health teams able to trace thousands of contacts
Limits: The 113 days are computed here from the two dates in the WHO notice; the case, death and contact counts are WHO's summary of Ministry of Health data. Containment within four months with no vaccine is the claim, but the outbreak reached nine districts including Kampala, the case fatality was 47 percent, and the record cannot separate the government's response from luck in transmission chains or from the over US$27 million of international support WHO mentions. An epidemic is a different shock from the climate hazards the construct names first, filed on the same terms as the COVID-19 vaccination precedents, and a single 2022 outbreak says little about cyclone or flood preparedness. Uganda's earlier Ebola and Marburg experience is part of why the system existed.
๐ง๐ผBotswanasince 2021delivered and closedbears on Disaster preparedness and recovery
Botswana's health authorities reported to WHO that 1,951,054 people, 83 per 100 population, had received at least one COVID-19 vaccine dose by 31 December 2023, from a first dose given on 26 March 2021, with 1,663,490 (71 per 100) completing the primary series.
Botswana's health ministry ran a national COVID-19 campaign through its clinics and mass vaccination sites and reached over four in five residents with a first dose.
Preconditions
A routine immunisation programme with a cold chain beyond the capital
A supply of doses large enough for a national campaign, however obtained
Registers good enough to count who received a second dose
Where else it has been used: Bolivia (bol-covid-vaccination) and Venezuela (ven-covid-vaccination) are filed on the same terms.
Limits: The figures are Botswana's own reports to WHO (DATA_SOURCE 'REPORTING'), with no independent coverage survey behind them, read from one row of a CSV with no prose. First-dose reach was 83 per 100 and completion 71 per 100, among the highest in sub-Saharan Africa, which partly reflects a small population and a strong primary health system rather than a crisis-specific capability. Booster or additional doses reached 510,809 people, 22 per 100. The VACCINES_USED field is empty, so the record cannot show which products were used or where the doses came from, and it evidences distribution only. The per-100 rates divide by WHO's population estimate. Vaccination is a response to one shock, not proof of preparedness for cyclones, floods or droughts, and the brief for this column prefers a hazard-specific delivery; this is filed because none with an official figure was found in the time available. The cited URL is the blob the WHO dashboard serves, as in bol-covid-vaccination.
๐ฒ๐บMauritiussince 2021delivered and closedbears on Disaster preparedness and recovery
Mauritius's health authorities reported to WHO that 1,123,773 people, 88 per 100 population, had received at least one COVID-19 vaccine dose by 31 December 2023, from a first dose given on 26 January 2021, with 1,088,142 (86 per 100) completing the primary series.
Mauritius's health ministry vaccinated almost the whole adult population through its public hospital and health centre network, aided by a small island population and early dose supply.
Preconditions
A routine immunisation programme with a cold chain beyond the capital
A supply of doses large enough for a national campaign, however obtained
Registers good enough to count who received a second dose
Where else it has been used: Bolivia (bol-covid-vaccination) and Venezuela (ven-covid-vaccination) are filed on the same terms.
Limits: The figures are Mauritius's own reports to WHO (DATA_SOURCE 'REPORTING'), with no independent coverage survey behind them, read from one row of a CSV with no prose. Completion of 86 per 100 is near-universal, but Mauritius is a small island of about 1.3 million people with a dense public health network, so the scale says little about logistics capacity. Booster or additional doses reached 654,507 people, 51 per 100. The VACCINES_USED field is empty, so the record cannot show which products were used or where the doses came from, and it evidences distribution only. The per-100 rates divide by WHO's population estimate. Vaccination is a response to one shock, not proof of preparedness for cyclones, floods or droughts, and the brief for this column prefers a hazard-specific delivery; this is filed because none with an official figure was found in the time available. The cited URL is the blob the WHO dashboard serves, as in bol-covid-vaccination.
๐ง๐ซBurkina Fasosince 2020still operatingbears on Institutional responsiveness
Burkina Faso's national safety net programme (Burkin-Naong-Sa Ya) paid three rounds of FCFA 20,000 adaptive cash transfers to over 42,800 people between May and September 2020 in response to COVID-19, including 21,160 vendors from temporarily closed markets, and later raised the lean-season transfer from FCFA 20,000 to FCFA 45,000 in 2022 to align it with the food security response, paying three extra monthly payments to over 763,700 people in IPC3+ areas over the 2021 to 2023 lean seasons.
The safety net unit kept an 'adaptive' transfer window inside its regular programme and pointed it at whichever shock arrived: market closures and urban poverty in 2020, then IPC3+ lean-season food crises from 2021, raising the transfer amount when food prices rose.
Preconditions
A regular cash transfer programme with payment agencies and beneficiary lists already in place
A food security classification (Cadre Harmonise) to decide where adaptive payments go
Donor financing that allowed components to be reoriented without new approvals for each shock
Where else it has been used: The same adaptive safety net design runs through the Sahel Adaptive Social Protection Programme countries (Mali, Niger, Senegal, Chad, Mauritania); compare mli-pgtmu, where the government funded the response itself.
Limits: The World Bank financed the programme and reports on its own project; paragraphs 34 and 35 are on page 10 of the plain-text ICR. The COVID-19 response was small: over 42,800 people, against a regular programme the ICR targeted at 948,100 people and says reached close to 1.05 million, and the ICR mixes units (21,000 of the beneficiaries are described as households in Ouagadougou and Bobo Dioulasso). The ICR gives the payment window, May to September 2020, not the date of the first payment, so the most that can be read is that cash started within about two months of the first COVID-19 measures. The 2020 audit was qualified, finding FCFA 27 million of overpayments to COVID-19 beneficiaries by payment agencies and discrepancies in payment lists (page 24). The ICR says the 2022 increase was made to align with the food security response under the Sahel Adaptive Social Protection Program, not as an indexation rule, and it converts both the 2022 and 2023 amounts to about US$85, so the change kept pace with prices rather than raising real support. All the money is IDA and trust fund money, not the national budget, and the ICR records that the two military coups of 2022 (January and September) led to pauses in disbursements, so the responsiveness shown in 2020 did not hold steady through the period.
In response to COVID-19, the government of Cote d'Ivoire used its own budget to enrol a fourth cohort of 100,000 households (80,000 urban, 20,000 rural) in December 2020 into the World Bank supported productive cash transfer programme, through the programme's existing targeting, enrolment and digital payment systems, taking it to 227,000 households and over 1.3 million people.
Faced with the pandemic, the government put its own money into an existing donor-built cash transfer instead of creating a new scheme, so the new cohort used the programme's community targeting, enrolment and mobile payment machinery and was enrolled within the year.
Preconditions
A cash transfer programme already running with digital payments before the shock
Budget room to fund a cohort from domestic resources
A delivery unit able to extend targeting from rural to urban households, which the 2019 urban pilot had prepared
Where else it has been used: The same piggybacking on a standing programme as Honduras's 2020 top-ups (hnd-cct-shock-response) and Togo's Novissi, though slower than Novissi, which built a new digital transfer within weeks.
Limits: The World Bank financed the programme, rates its own project Highly Satisfactory, and reports results jointly for its project and the government-funded cohort, so the 100,000 is the financier's account of a government action it supported technically. Table 2 and paragraphs 32 to 34 are on page 13 of 48 in the plain-text version; paragraph 85 (page 21) says the pandemic triggered the expansion with government funding. The speed is modest: the first COVID-19 measures were in March 2020 and the cohort enrolled in December 2020, about nine months later, and the ICR gives no date of first payment to the new cohort. The separate emergency transfer the government announced in April 2020 through the Fonds special de solidarite (reported in the press at about 177,000 households) is not covered by this source and is not counted here. The ICR also says the project was never formally restructured to take in the government money and that the social registry (RSU) was underdeveloped, so the expansion rode on one programme's lists rather than a national registry. Households are counted at enrolment; the source reports payment receipt only through sample surveys.
The Republic of Congo's Ministry of Social Affairs and the Lisungi programme unit paid one-off emergency cash transfers of about US$82 to more than 206,800 mostly urban households hit by COVID-19 by April 2022, above a target of 200,000, while the social registry used to target them grew to more than 830,000 households and 95 percent of the transfers were paid digitally.
The social affairs ministry used the existing Lisungi conditional cash transfer unit and its social registry to register urban households hit by COVID-19 and pay each a one-off digital transfer, growing the registry in the process so that later programmes could target from it.
Preconditions
A standing cash transfer programme unit (Lisungi) with a registry to build on
Emergency IDA financing approved within three months of the shock
Digital payment providers able to reach urban households
Where else it has been used: The same piggybacking on a standing programme as civ-ptmp-covid and cod-step-kin, but with a single small payment per household rather than a recurring transfer.
Limits: The World Bank financed the transfers (an IDA credit of US$50 million, US$20 million for the ECTs) and reports on its own project; the figures are in paragraphs 23 to 25 of the plain-text project paper, pages 12 to 13 of 59. The project's later Implementation Status Report (June 2026, P174178) gives 231,562 households benefiting from emergency cash transfers, so the count moved after this paper; the record uses the earlier dated figure. The response was slow for an emergency: the project was approved on 24 June 2020 and became effective on 9 October 2020, and the paper gives no date for the first or last ECT payment, so the time from shock to cash cannot be stated, only that 206,800 households had been paid by April 2022. Each household got one payment of about US$82, small against lost income. The same paper says the focus on the crisis delayed the government's plan to scale Lisungi to national coverage, and that procurement delays and a change of ministry leadership in May 2021 held up the registry and digital payment work. The government asked for the emergency money; the money and the design were the World Bank's.
๐ฌ๐ญGhanasince 2022delivered and closedbears on Institutional responsiveness
After losing market access and requesting IMF support in July 2022, Ghana's government launched a domestic debt exchange on 5 December 2022 covering all medium- and long-term local-currency government, Daakye and ESLA bonds, and settled it on 21 February 2023 with 85 percent of the face value of bonds held by investors other than pension funds exchanged, giving almost GHS 50 billion of cash debt relief in 2023.
The Ministry of Finance offered domestic bondholders a menu of new longer, lower-coupon bonds at fixed ratios with a deadline, renegotiated the perimeter (exempting pension funds and adjusting terms for individuals) during the offer, and closed within eleven weeks so the IMF programme could be approved.
Preconditions
A domestic bond market concentrated in a few banks and institutions that could be negotiated with directly
A central depository able to swap instruments on a fixed settlement date
An external anchor (the IMF arrangement) that made the exchange a precondition for financing
Where else it has been used: Jamaica's 2010 and 2013 domestic exchanges are the template Ghana's design follows; Sri Lanka's 2023 domestic optimisation used a similar perimeter carve-out for pension funds.
Limits: The figures are in Box 2 of the staff report (PDF page 22, printed page 18) and the July 2022 request for IMF support is in the debt sustainability annex (PDF page 95); the URL is the Ministry of Finance's copy of the IMF report, because imf.org returned an HTML bot page to a script. The 78 days is computed by us from the two dates in Box 2. The IMF is a party to the programme the exchange was a prior condition for, so this is not an independent evaluation. The speed is of the rule change once decided: the fiscal and debt pressures that forced it built up over years and Ghana lost market access in 2022, so a responsive state would arguably have acted earlier and with less damage. The 85 percent covers only the eligible perimeter: pension funds, about 20 percent of eligible debt, were excluded after bondholder protests, and 85 percent equals only 28 percent of all domestic debt; dollar bonds and Cocobills needed later exchanges. The exchange imposed an average NPV loss of about 30 percent on bondholders, weakened bank balance sheets and was contested in court and on the street; the number shows the operation closed fast, not that its costs were well distributed.
๐ฌ๐ฒGambiasince 2020delivered and closedbears on Institutional responsiveness
The Gambia's government had its new social safety net project restructured within three months of the pandemic (restructuring dated 9 June 2020) to create Nafa Quick, which between July and October 2020 paid two transfers of GMD 3,000 to 78,359 rural households in 30 districts, against a target of 60,000, with all payments made before the end of the lean season.
The government asked the World Bank to repurpose its freshly started safety net project, and the project unit ran a separate, simple rural cash transfer, two payments timed to the lean season, through community targeting in 30 districts while the social registry and the regular programme were still being built.
Preconditions
An externally financed safety net project already effective, with money that could be reallocated
A project unit and community structures able to list households without a registry
Willingness to sacrifice the duration of the long-term programme to pay for the emergency one
Where else it has been used: Like hnd-cct-shock-response and cog-lisungi-covid, a donor project became the shock-response channel; The Gambia's case shows the cost explicitly, because the regular programme was cut to fund it.
Limits: The World Bank financed the project and rates its own efficacy High; the figures are in the plain-text ICR: the June 2020 restructuring on page 9 (paragraph 17), the 78,359 households and the 100 percent lean-season indicator in Table 2 on page 14 (paragraph 34), and 'restructured in three months' with the July to October 2020 payment window on page 24 (paragraph 70). The ICR says Nafa Quick ran before the delivery system existed, so no socio-economic data were collected on its beneficiaries and its targeting of the poor cannot be assessed. The speed was bought by cutting the longer-term programme: the extreme-poor Nafa transfer was halved from 36 to 18 months to pay for Nafa Quick, and the additional financing meant to restore it was not approved because a COVID-19 waiver expired first, so the remaining 18 months were deferred until a new project, The Gambia RISE (approved February 2024), began paying them from July 2024 (page 24, paragraph 69). Two payments of GMD 3,000 is a small, temporary transfer. The money came from an existing IDA project, not from the national budget.
๐ฒ๐ฑMalisince 2021delivered and closedbears on Institutional responsiveness
Mali's government launched its own emergency COVID-19 cash transfer programme in 2021, paying one-off transfers to 400,000 poor households across the country through the targeting and payment systems built by the Jigisemejiri safety net project, and kept it running in 2022, but only FCFA 42 billion of the FCFA 100 billion budgeted was disbursed because of the political crisis.
The government funded an emergency transfer from its own budget and ran it through the social registry, targeting procedures and payment contracts that a donor project had built over a decade, so it could reach 400,000 households without building new machinery.
Preconditions
A social registry (RSU, over 1.1 million household records) and a government safety net unit (UTGFS) already in place
Budget room for a FCFA 100 billion programme, which in the event only partly materialised
Political stability to sustain disbursement, which the 2020 to 2022 crisis undercut
Where else it has been used: The same reuse of donor-built delivery systems as civ-ptmp-covid; Mali's case shows a government paying with its own money, and the budget gap shows the limit of that.
Limits: The numbers are reported by the World Bank about a programme it did not finance, in paragraph 42 of the ICR of its own project (page 18 of 67 in the plain-text version), with no government source cited and no date of first payment; the 400,000 is a round figure and the ICR does not say whether it counts households paid or households planned. Elsewhere the ICR says 'approximately 400,000 households'. The response was slow: the programme was launched in 2021, a year or more after the first COVID-19 measures, and only 42 percent of its budget was disbursed (computed by us from FCFA 42 billion of 100 billion) because of the political crisis that followed the coups of August 2020 and May 2021. The same paragraph records that the government's regular safety net budget line (PGFS) disbursed FCFA 6.2 billion of 12.5 billion over 2017 to 2022. The ICR also records World Bank disbursement suspensions to Mali in August 2020, May 2021 and January 2022 for a cumulative 18 months (page 27), which slowed the systems the transfer relied on. A one-off transfer does not show a lasting change in the rules.
๐ธ๐ฉSudansince 2021dismantledbears on Institutional responsiveness
After Sudan cut fuel subsidies and unified its exchange rate in February 2021, the transitional government's Sudan Family Support Programme began paying cash in March 2021 and by 25 October 2021 had registered about 2.7 million families and paid about 30.6 billion Sudanese pounds to 1.4 million families, about 4.7 million people, before all activities, including cash delivery, were halted after the military takeover of that day.
The Ministry of Finance and the Ministry of Labour and Social Development built a near-universal cash transfer to offset fuel subsidy removal and exchange-rate unification, registering families through 352 centres linked to the national civil registry and paying through banks, mobile money and cash cards, with donor money held in a designated account; the coup stopped the money and the programme with it.
Preconditions
Donor financing large enough to compensate most households for a subsidy cut
A national ID registry to validate enrolment against duplicates
A government that the financiers continue to recognise, which is the precondition that failed in October 2021
Where else it has been used: The same compensation logic as Iran's 2010 subsidy reform cash transfer and Egypt's Takaful and Karama, but financed almost entirely from outside, which made it end the day the financiers paused.
Limits: The World Bank and the STARS donors financed the programme, decided its pause and wrote this report, so the publisher is grading its own operation and also chose to end it. The numbers are in a PDF: the subsidy and exchange-rate reforms of February 2021 on PDF page 10, the halt of all activities 'including delivery of cash to beneficiaries' on PDF page 16, the 2.7 million families registered and 1.4 million families (about 4.7 million people) paid on PDF page 23, and the per-state table (Table E4, 2,637,632 enrolled and 1,395,295 paid) on PDF page 85, which notes that the first 80,000 families were paid in March 2021 (PDF page 77). The 1.39 million families paid include families in two states where the World Food Programme, not the ministry, implemented payments (PDF page 75), so not every paid family was reached by the government's own delivery. Table E4 has visible inconsistencies (East Darfur shows 391,900 families paid against 4,291 enrolled), so the state rows should not be quoted. The speed claim is about seven months from first payment to 1.4 million families, against a target of 6.5 million; the report says the programme started late because of exchange-rate problems, and paid families reached only 22 percent of the target population. What responded was a transitional government acting inside an IMF and World Bank reform package that conditioned arrears clearance, not on its own timetable. The halt was a decision by the financier after a coup, so the dismantling reflects a political break and donor policy (OP 7.30) as much as institutional capacity; later safety-net payments in Sudan were routed through the World Food Programme rather than the ministry.
The Democratic Republic of the Congo's Ministry of Health, supported by WHO and partners, ran the response to the North Kivu and Ituri Ebola outbreak from August 2018 until it was declared over on 25 June 2020, vaccinating more than 303,000 people, registering 250,000 contacts and testing 220,000 samples, in an outbreak that counted 3,481 cases and 2,299 deaths.
The health ministry ran an incident management structure with WHO, traced contacts across three provinces and used ring vaccination with a new vaccine to cut chains of transmission, while working through community leaders in a conflict zone.
Preconditions
A health ministry with prior Ebola outbreak experience and field teams
A licensed or investigational vaccine and the cold chain to deliver it at minus 60 degrees or colder
International surge capacity and money for a response lasting years, not weeks
Where else it has been used: The same ring vaccination approach was used in Guinea in 2021 (gin-ebola-2021), where the outbreak was far smaller and ended in four months.
Limits: This is the second largest Ebola outbreak on record, and the vaccination count measures the size of the response, not its success: it ran for almost two years and killed 2,299 people, about two thirds of the cases, so the record evidences a delivery under extreme conditions rather than early containment. WHO and many partners carried much of the operation, the vaccine was used under a ring vaccination protocol, investigational for most of the outbreak until its licensure in late 2019, which the page does not discuss, and the page does not split national from international effort. The figures are WHO's summary of Ministry of Health reports, and the vaccination count is a rounded 'over 303 000'. The outbreak ran in an active conflict zone, which the page names but does not describe or quantify. A resurgence in North Kivu followed in February 2021, listed on the same page, so the declared end did not mean the virus was gone from the region.
๐ธ๐ฉSudansince 2017operating below its peakbears on Disaster preparedness and recovery
WHO and UNICEF estimated that Sudan's national DTP3 coverage among one-year-olds fell from 95 percent in 2017 and 84 percent in 2022 to 51 percent in 2023 and 39 percent in 2024, after war broke out in April 2023, before recovering to 71 percent in 2025.
Sudan's Expanded Programme on Immunization kept national DTP3 coverage above 90 percent for most of the 2010s through fixed clinics and outreach, then lost more than half of it when war closed facilities, displaced millions and broke the supply chain from Khartoum, and partly rebuilt it with catch-up campaigns.
Preconditions
A cold chain and clinic network that depends on the capital for supply and pay
Population registers and reporting that survive displacement
External partners able to fund and run catch-up campaigns in contested areas
Where else it has been used: Haiti (hti-immunization-erosion) shows a slower erosion under insecurity; Sudan's fall was faster and deeper, and its partial recovery came within two years.
Limits: Routine immunization is a standing health service, not a hazard-specific preparedness delivery, and this record is filed on the precedent of hti-immunization-erosion: it reads the war as the shock and coverage as the capacity to keep a basic service running through it and restore it. The figures are WHO and UNICEF estimates built from administrative reports and surveys, not a census, and estimates for a country at war with displaced populations and collapsed reporting are especially uncertain. The API rows mix revisions: 2000 to 2024 carry a July 2025 timestamp and 2025 a July 2026 one, so the latest revision may have restated earlier years that the API does not show. 95 percent in 2017 is the highest value in the published series, which starts in 2000. The fall is caused mainly by war and displacement, outside the health system's control, and the 2025 recovery may owe much to UNICEF, WHO and Gavi supported catch-up campaigns rather than the national programme alone; the data cannot separate them. The started year is the start of the comparison, following hti-immunization-erosion, not the founding of the national programme.
๐ฆ๐ดAngolasince 2025delivered and closedbears on Disaster preparedness and recovery
Angola's Ministry of Health, with WHO, UNICEF, the World Bank and the ICRC, vaccinated over 900,000 people with oral cholera vaccine in a five-day reactive campaign in January 2025, with 99.5 percent administrative coverage, in an outbreak that had reached 8,543 cases and 329 deaths, a case fatality rate of 3.9 percent, in 16 of 21 provinces by 23 March 2025.
The health ministry used the international oral cholera vaccine stockpile to mount a five-day mass campaign in the first affected areas, while case management and water and sanitation lagged behind the spread to 16 provinces.
Preconditions
Access to the global oral cholera vaccine stockpile and partners to fund delivery
Vaccination teams able to reach crowded urban districts within days
Treatment centres and oral rehydration close enough to patients to keep deaths low, which here was missing
Limits: The record shows a fast vaccination campaign inside a response that failed on its main outcome: WHO says a case fatality rate can be kept below 1 percent with timely care, and Angola's was 3.9 percent, with a third of deaths in the community, which WHO attributes to late presentation, gaps in surveillance and early response, and weak access to treatment. The 99.5 percent is administrative coverage of a target population the DON does not define, not a survey, and the 900,000 is rounded. The campaign was run with WHO, UNICEF, the World Bank and the ICRC, so the national share cannot be separated, and the DON does not say where the doses came from. The DON is a snapshot at 23 March 2025 while the outbreak was still spreading; this record does not show when or how it ended. The status 'concluded' refers to the January campaign, not the outbreak. The underlying cause, poor water and sanitation in crowded areas, is a standing gap the vaccination does not address; WHO notes cholera re-emerged in 2006 after a decade-long absence and recurs periodically.
๐ฌ๐ญGhanasince 2022delivered and closedbears on Disaster preparedness and recovery
Ghana's Ministry of Health declared the country's first Marburg virus disease outbreak on 7 July 2022, identified and monitored 198 contacts through 42 days, and declared the outbreak over on 16 September 2022 after three confirmed cases, all from one household, and two deaths.
The Ghana Health Service ran its integrated disease surveillance and response system: a suspected viral haemorrhagic fever was tested at the national reference laboratory, the ministry declared the outbreak, traced and followed every contact for 42 days and isolated the remaining case in a designated centre.
Preconditions
An integrated disease surveillance and response system reporting from district level
A national laboratory able to test for filoviruses
Designated isolation facilities and contact tracing teams ready before the event
Where else it has been used: Rwanda's 2024 Marburg response (rwa-marburg-2024) ran the same playbook at twenty times the case count.
Limits: Three cases in one household is a very small event, and containing it shows that surveillance and contact tracing worked once, not a standing national capacity; a small outbreak can end quickly by luck of the transmission chain. Detection was late for the first case: the index patient was admitted on 26 June and died the next day, and a second household member, a child, was infected and died before isolation, so two of three cases were fatal. The figures are WHO's restatement of Ministry of Health reports, so the government is reporting on its own response, with WHO and partners in support. The national laboratory (Noguchi Memorial Institute) confirmed the virus, but Institut Pasteur Dakar's confirmation was part of the timeline, as the DON describes. The record says nothing about Ghana's response to other shocks such as the 2023 Akosombo dam spillage floods.
๐จ๐ฎCรดte d'Ivoiresince 2021delivered and closedbears on Disaster preparedness and recovery
Cรดte d'Ivoire's health authorities reported to WHO that 13,568,372 people, 51 per 100 population, had received at least one COVID-19 vaccine dose by 31 December 2023, from a first dose given on 1 March 2021, with 12,164,304 (46 per 100) completing the primary series.
The health ministry ran a national COVID-19 campaign through its immunisation programme and vaccination sites, and reached about half the population with a first dose within three years.
Preconditions
A routine immunisation programme with a cold chain reaching districts
Donated or COVAX doses arriving in large enough volume for a national campaign
Registers able to count completion of the primary series
Where else it has been used: Bolivia (bol-covid-vaccination) and Venezuela (ven-covid-vaccination) reached similar partial coverage; among this batch's West African countries only Guinea (66 per 100 first dose) reported more.
Limits: COVID-19 vaccination is a response to one shock and a judgement call for this column, which prefers a hazard-specific delivery; no hazard-specific Ivorian case with an official figure was found. The figures are Cรดte d'Ivoire's own reports to WHO (DATA_SOURCE 'REPORTING') with no independent coverage survey. First-dose reach was barely a majority (51 per 100) and fewer than half completed the primary series (46 per 100), with boosters at 12 per 100, so this is a partial delivery on the same terms as ven-covid-vaccination; it is filed because it reached about half the population nationally, which most of its West African neighbours did not. The VACCINES_USED field is empty, so the record cannot show which products were used or where the doses came from, and evidences distribution only. The per-100 rates divide by WHO's population estimate. The file was last updated for this country on 31 December 2023. The WHO dashboard address covid19.who.int/who-data/vaccination-data.csv now returns an HTML page, so the cited URL is the blob the dashboard serves, the same one bol-covid-vaccination cites.
๐ฒ๐ฑMalisince 2014delivered and closedbears on Disaster preparedness and recovery
After Mali confirmed an imported Ebola case on 23 October 2014 and a second, unrecognised imported case started a chain of seven further cases in a Bamako clinic, the government with WHO and partners identified and followed 433 contacts for 21 days, set up a hotline receiving about 6,000 calls a day, built isolation and treatment facilities, and ended monitoring of the last contacts on 15 December 2014.
Mali's government, on alert since a false alarm in April 2014, traced every contact of two imported cases with medical students trained in epidemiology as extra tracers, isolated them, and used a hotline to find rumours and suspected cases.
Preconditions
A biosafety laboratory in the capital able to confirm the virus within days
Months of preparedness work and a rehearsal before the first case
A pool of trainable people, here medical students, to scale contact tracing quickly
Where else it has been used: Nigeria (nga-ebola-response) ran the same playbook three months earlier, and the WHO page groups the two as models.
Limits: The page is a WHO narrative written to present the response as a success, not a dataset, and it was written before Mali was declared free of transmission (WHO says that would happen on 18 January 2015 if no further cases arose). The second chain shows a detection failure: a patient from Guinea was admitted to a private clinic in Bamako with a diagnosis of kidney failure and died, and seven further cases followed, five of them fatal, including a doctor and a nurse who treated him; the page does not give a single total for cases and deaths. WHO and partner staff already in the country for preparedness shifted to the response, and the Bamako laboratory used was built with US National Institutes of Health support, so the national share cannot be separated. This is one small event seeded by two imported cases, and it says nothing about Mali's capacity for other hazards, including the conflict that has displaced people in the north and centre since 2012.
๐ฌ๐ฒGambiasince 2020delivered and closedbears on Disaster preparedness and recovery
The Gambia's government bought 264,500 bags of rice, 279,500 bags of sugar and 185,200 drums of oil for vulnerable households with about GMD909.9 million redirected in 2020, and the National Disaster Management Agency distributed them between May and 28 September 2020, with 228,367 bags of rice recorded as delivered to households, 36,133 fewer than procured.
The government redirected debt-relief money to buy staple food in bulk and used the disaster management agency and local authorities to hand it out through distribution points with coupons to households identified as vulnerable.
Preconditions
A disaster agency with a field network able to run distribution points nationwide
Household lists from local authorities, here incomplete and drawn up without criteria
Stock records and storage that let food be tracked from purchase to household, which here were missing
Limits: The audit office judged the distribution untimely, unequal and poorly recorded, so this is a record of a national relief delivery with large leaks, not a model. Procurement deliveries ran from 1 to 12 May 2020 but the last household distribution was on 28 September 2020, nearly five months later. The delivered counts come from the Kobo-Collect tool, which the audit says recorded what a household should receive rather than what it received, some households recorded as served could not be reached, and coordinators' reports disagree with it, so 228,367 is an upper bound on what reached people. The same delivered figure appears for rice and sugar in the audit table. Households were selected without a needs assessment, some vulnerable households in two Kanifing wards were never registered, and food stored in the open was damaged without records. The audit examined Banjul, Kanifing and West Coast Region in detail and left the other regions to a later audit. The number of households reached is not reported. The figures are in a 156-page PDF (executive summary, pages 18 to 24).
๐ฆ๐นAustriasince 2020delivered and closedbears on Institutional responsiveness
Austria's social partners drafted a COVID-19 short-time work model on 13 and 14 March 2020, the AMS adopted its directive on 18 March and the ministers approved it on 19 March, three days after the first lockdown began, and in 2020 113,771 firms with 1.25 million employees used the scheme, against 508 firms in 2009.
The government let the social partners (the chambers of commerce and labour and the trade unions) write the crisis model over a weekend, the employment service turned their outline into a directive within three days, and parliament adapted the existing short-time work law by initiative motion on 15 and 20 March.
Preconditions
A standing short-time work instrument and an employment service that already processed it
Social partners with the authority to agree terms for the whole economy in days
Fiscal room to pay out EUR 6.4 billion in a year
Where else it has been used: The same corporatist route as Germany's Kurzarbeit (deu-kurzarbeit), made faster and more generous; the audit office's verdict is the price of that speed: overpayment and repeated retroactive change.
Limits: The figures are in a PDF: the 113,771 firms and 1.25 million employees, the EUR 7.849 billion paid from March 2020 to March 2021 and the EUR 500 million overpayment are in the summary (PDF page 10, repeated in TZ 18); the timeline (social partner talks on 13 and 14 March, the punctation handed to the AMS on 15 March, the AMS board decision on 18 March, ministerial approval on 19 March, retroactive to 1 March) is in TZ 2 (PDF page 21). The three days are our arithmetic from those dates. The audit office is critical, not celebratory: it found that speed came at a cost, with an unsuitable flat-rate method and over-compensation of municipal tax causing overpayments of about EUR 500 million in phase 1, a directive amended twelve times by March 2021, often retroactively, the AMS and the ministry's labour market experts left out of the design, and no concept for targeted detection of abuse. The 1.25 million counts employees in participating firms over the year, not workers on short time at one moment. Austria already had a short-time work instrument, so the move was a redesign of an existing rule, not a new one, and the number of jobs it saved is an estimate the office treats cautiously.
๐ฉ๐ฐDenmarksince 2020delivered and closedbears on Institutional responsiveness
Denmark's government, employers and unions agreed a wage compensation scheme on 14 March 2020, the Danish Business Authority opened applications 11 days later on 25 March and decided about 90 percent of applications within 30 days, paying DKK 12.7 billion in the first phase (March to August 2020) on 66,631 applications from 34,042 firms.
The government agreed the scheme with the employers' and workers' organisations in one tripartite deal, and the Danish Business Authority built the application system on existing case systems and public register data, with automatic checks deciding low-risk cases at once and a later control for the rest.
Preconditions
Tripartite bargaining that can reach a national deal in a weekend
Register data on firms and employees that can check applications automatically
Parliamentary approval of spending by account (aktstykke) instead of a new law
Where else it has been used: The Netherlands' NOW scheme and Austria's Kurzarbeit (aut-corona-kurzarbeit) took similar weeks; what Denmark adds is an audit of the trade-off, which shows the after-control lagging the payments by more than a year.
Limits: The figures are in a PDF: the 14 March agreement and the 25 March opening on printed page 7 (PDF page 14), Table 1 with applications, firms and payments on PDF page 11, and the 90 percent decided within 30 days on page 4 (PDF page 10). The audit office states that the ministry normally needs several months to set up such a scheme; the comparison is the ministry's own claim. Speed was bought with deferred control: the office found that control relied mainly on an after-control that had not yet been carried out, had been postponed by about 11 months to as much as a year and a half after payment, and had no earlier check of the riskiest cases, so some firms may have received compensation they were not entitled to and that cannot be recovered (Statsrevisorernes bemaerkning, PDF pages 3 to 4). The scheme was later extended into four phases, with DKK 19.4 billion paid by October 2021; the claim covers the first phase only. The record counts firms and applications, not employees: the office's report gives no count of workers covered.
๐ง๐ชBelgiumsince 2020delivered and closedbears on Institutional responsiveness
Belgium's National Employment Office treated every COVID-19 work stoppage as temporary unemployment for force majeure and paid allowances to 939,532 workers for March 2020 and 1,117,133 for April, when 1,221,412 workers, 35.3 percent of private and contractual employees, were declared temporarily unemployed; almost 1.4 million received at least one payment during 2020.
The employment office widened an existing rule, temporary unemployment for force majeure, to cover every corona stoppage and simplified the employer procedure, so firms could suspend contracts without proving economic grounds and existing payment bodies paid workers within the month.
Preconditions
A standing temporary unemployment regime with electronic social risk declarations from employers
Payment bodies, including trade union funds, already paying unemployment benefits
A legal concept (force majeure) broad enough to be reinterpreted by the administration
Where else it has been used: France's activitรฉ partielle (fra-activite-partielle) followed the same route of stretching an industrial instrument to the whole economy; Belgium's version paid workers directly rather than reimbursing employers.
Limits: The monthly counts and the 35.3 percent share are on the ONEM press page; the 1.4 million for 2020 and the monthly average of 504,657 are on the annual report page (https://www.onem.be/espace-presse/rapport-annuel-onem-2020). The press page gives no date for the rule change itself, so the record shows speed only as payment for the first month of the crisis, not as days from shock to rule. ONEM does not pay most allowances itself: the payment bodies (the trade union funds CSC, FGTB and CGSLB, and the public fund CAPAC) do, and the press page splits the amounts between them. The 35.3 percent is computed on 3,458,136 salaried employees excluding civil servants. A worker counted in April was unemployed for 14 days on average, and only 20 percent for the whole month, so the head count overstates lost hours. The cost was large: ONEM's annual report records EUR 4.042 billion of extra temporary unemployment spending in 2020. ONEM later ran controls on corona temporary unemployment; their results are not reflected here.
๐ฎ๐นItalysince 2020delivered and closedbears on Institutional responsiveness
Italy's Cura Italia decree of 17 March 2020 created wage supplementation with a Covid-19 cause across ordinary CIG, the solidarity funds and CIG in deroga, and INPS recorded 5.27 million beneficiaries in April 2020 and 2.8 billion authorised hours from March to August, but about 83 percent of firms asked INPS to pay workers directly and the institute's own report records the resulting delay in payments for the first period.
5,274,000 beneficiaries, beneficiaries of wage supplementation with covid-19 cause in the month, 2020-04. 2.8 billion hours, hours authorised with covid-19 cause, march to august 2020, 2020-08. INPS, XIX Rapporto annuale (October 2020), retrieved 2026-10-02.
The move, and what it needed
The government extended the existing wage supplementation funds to a Covid-19 cause by decree, covered firms outside the ordinary funds through the solidarity funds and the regional CIG in deroga, and let firms shift payment to INPS, which then had to authorise and pay an unprecedented volume at once.
Preconditions
Standing wage supplementation funds with an authorising and paying institute
Decree-law powers that take effect on publication
Payment capacity at the paying institute, which is the precondition that strained here
Where else it has been used: Belgium (bel-chomage-temporaire-corona) also paid workers directly through existing bodies and paid within the month; Italy's direct-payment option moved the cash-flow burden from firms to INPS and slowed the first payments.
Limits: The figures are in a 482-page PDF: the monthly beneficiary counts are Tavola 1.7 (PDF pages 43 to 44, in thousands: 4,240 in March, 5,274 in April across ordinary CIG, solidarity funds and CIG in deroga), the 2.8 billion hours on PDF page 16 and again on page 33, and the direct-payment share and the delay in footnote 12 (PDF page 40). The institute reports on its own delivery. Speed is mixed: the rule changed within days of the lockdown, but the decree let any firm request direct payment by INPS without proof of financial difficulty, about 83 percent did, and the report says this produced a delay in payments, especially for the first period requested; it gives no figure for the delay. Authorised hours are not hours used: the report estimates take-up at 63 percent for March and 60 percent for April. The report itself judges the solidarity funds that covered smaller firms evidently inadequate, and the CIG in deroga ran through the regions, which added a step before payment.
๐จ๐ฟCzechiasince 2020delivered and closedbears on Institutional responsiveness
The Czech government approved the Antivirus employment support programme on 31 March 2020, with costs eligible from 12 March, and the Labour Office paid CZK 51.2 billion through it from 2020 to 2022 to almost 71,000 employers for 1.2 million employees.
The government used an existing power in the Employment Act to declare a targeted programme by resolution, so the Labour Office could reimburse employers for wage compensation paid to workers in quarantine or kept idle, backdated to the start of the emergency.
Preconditions
A statutory power to create targeted employment programmes by government resolution
A Labour Office able to sign agreements with tens of thousands of employers
Labour law that already obliged employers to pay wage compensation during obstacles to work
Where else it has been used: Slovakia's First Aid (svk-prva-pomoc) was approved the same day; the Czech audit shows the cost of a rule written for speed with no sharing of the burden and no impact monitoring.
Limits: The figures are in the audit office's PDF (CZK 51.2 billion and 70,971 employers in the summary table on PDF page 2; the government resolution No. 353 of 31 March 2020 and the eligibility period from 12 March 2020 in Table 1 on PDF page 7) and in its press release of 4 December 2023 (https://www.nku.cz/cz/pro-media/tiskove-zpravy/miliardy-na-udrzeni-pracovnich-mist-v-dobe-covidu-19-dostaly-od-statu-i-firmy--kterym-rostl-zisk-a-pocet-zamestnancu-id13610/), which gives almost 71,000 firms and 1.2 million employees. The audit office's verdict is negative: part of the money was spent ineffectively, CZK 9.8 billion went to industries where output did not fall, the A Plus regime paid 100 percent of wage compensation against the principle of cost sharing used elsewhere in the EU (CZK 4.4 billion could have been saved), ten of 31 audited firms drew CZK 1.9 billion while paying dividends, and the ministry and Labour Office did not monitor or evaluate the programme's effects. The 19 days from the eligibility date to the government decision are our arithmetic and are the second metric; the 1.2 million employees covered are in the press release only; the record does not show when the first payments arrived. The rule change was speedy; whether it was well targeted is what the audit disputes.
๐ธ๐ฐSlovakiasince 2020delivered and closedbears on Institutional responsiveness
Slovakia's government approved the labour ministry's job retention scheme on 31 March 2020, labour offices began accepting applications on 6 April with support backdated to 13 March, and over two years the scheme paid almost EUR 2.5 billion to 176,000 employers and self-employed people, 5.7 million monthly payments covering more than 770,000 employees.
The labour ministry designed a state aid scheme paid through the network of labour offices, with separate measures for closed businesses, firms with falling revenue and the self-employed, and the government approved it so offices could take applications within a week.
Preconditions
Labour offices that already contracted with employers for active labour market measures
European Social Fund money that could be reallocated to the scheme
EU state aid temporary framework rules allowing wage subsidies
Where else it has been used: The Czech Antivirus (cze-antivirus) was approved the same day with a similar design; both were later criticised for poor targeting.
Limits: The page is the labour ministry's own scheme site and its closing summary is promotional: it quotes the minister calling the ministry the most successful state body at delivering aid, and cites a supportive analysis by the ministry's own Institute of Social Policy. The figures are rounded ('almost 2.5 billion', 'more than 770 thousand') and the page has no publication date; the eligible period it states runs from 13 March 2020 to 28 February 2022. The six days are our arithmetic from the two dates on the page; the record does not show how long the first payments took after applications opened, and this record has not read any audit of the scheme. Employees counted are those in supported firms, not jobs shown to have been saved, and nearly half of the money went to micro-enterprises, per the institute's director quoted on the page.
๐ญ๐ทCroatiasince 2020delivered and closedbears on Institutional responsiveness
The governing board of the Croatian Employment Service decided on 20 March 2020 to introduce job preservation grants for employers hit by COVID-19, the service began implementing them on 23 March, taking applications online, and in 2020 it paid HRK 7.70 billion to 109,431 employers covering 697,126 workers on 225,390 applications.
697,126 workers, workers covered by job preservation grants in the year, 2020. 3 days, days from the board decision (20 march 2020) to the start of implementation (23 march 2020), 2020-03-23. Hrvatski zavod za zapoลกljavanje (HZZ), Godiลกnjak 2020, retrieved 2026-10-02.
The move, and what it needed
The employment service's governing board created a flat per-worker grant by decision, took applications online through a dedicated site, and paid employers monthly for March to May, then extended and adjusted the grant through the year as restrictions continued.
Preconditions
An employment service whose board can create a measure without new legislation
An online application channel built in days
European Social Fund and state budget money that could be redirected
Where else it has been used: Croatia reused the same instrument for later shocks: in 2026 the service ran a job preservation grant for employers hit by wildfires in Omiลก and Braฤ (mjere.hr).
Limits: The employment service reports on its own measure. The figures are in a PDF: the 20 March decision on PDF page 9 and again on page 41, the 109,431 employers, 697,126 workers and HRK 7,703,685,673 on PDF page 10 and page 41, the 23 March start and the 225,390 applications on page 41. The three days are our arithmetic. The yearbook counts workers covered at any point in 2020, and the same employer could apply repeatedly, so workers are not a monthly peak. To run the grants the service suspended part of its regular active labour market programmes from March to May 2020, which the yearbook states; the record does not show how many jobs the grants preserved or whether any money went to firms that would have kept their staff anyway.
๐ฎ๐นItalysince 2016still operatingbears on Disaster preparedness and recovery
Italy's Extraordinary Commissioner for reconstruction after the 2016 Central Italy earthquakes reported that by 31 May 2026 private reconstruction had received 36,149 grant applications for 17.82 billion euros, approved 24,650, granted 12.59 billion and paid 7.99 billion, and that 14,968 of 23,361 authorised building sites had been completed across the 138 municipalities of the crater.
Parliament created a dedicated extraordinary commissioner with ordinance powers, regional special reconstruction offices and a single digital platform for grant files, and the commissioner pays reconstruction grants for private homes and runs a programme of public works across four regions.
Preconditions
A national grant scheme that pays the full cost of rebuilding damaged primary homes
Regional reconstruction offices able to process tens of thousands of technical files
A legislative decree (D.L. 189/2016) that keeps the commissioner in place across changes of government
Where else it has been used: The same model was applied in 2025 to the Ancona and Umbertide earthquakes, according to the commissioner's own site; the Emilia 2012 and L'Aquila 2009 reconstructions used regional and municipal structures instead.
Limits: The commissioner reports on its own programme, and the executive summary is written as advocacy ('La cura funziona'). The figures are in a PDF, on PDF pages 3 to 5. The host sisma2016data.it rate-limited scripted requests (HTTP 429), so the PDF was read through a fetch tool, not curl. Ten years after the August 2016 shock, about 64 percent of authorised private sites were complete and less than half of the 17.82 billion requested had been paid; the report itself says 69 percent of the payments were made from 2023 onwards, so most of the recovery happened six to ten years after the earthquake, which is evidence of slow recovery as much as of capacity. Applications and sites are counts of files and works, not households rehoused; the report gives one household figure, over 1,300 families (about 2,800 people) returning home in the last year, without a cumulative total. Public reconstruction, 3,667 programmed works for over 4.85 billion euros, is reported as reaching 40 percent of sites under way or finished. The record shows recovery after a shock, not preparation before one: Italy's seismic prevention and the emergency response in 2016 are outside it.
๐ญ๐ทCroatiasince 2020still operatingbears on Disaster preparedness and recovery
Croatia's Ministry of Construction, Physical Planning and State Assets and Ministry of Health had disbursed 16 percent of the 200 million US dollar World Bank earthquake recovery loan approved in June 2020 by November 2024, when the closing date was extended for the second time to 30 June 2027, and by May 2026 had restored 40 percent of the targeted health and education buildings.
After the March 2020 Zagreb earthquake the government borrowed from the World Bank to rebuild damaged hospitals and schools and to fund public health preparedness, while spending EU Solidarity Fund grants first; the loan's complex hospital works then took years to tender and its closing date was extended twice.
Preconditions
A rapid damage and needs assessment that sized the loan within three months
Implementing units inside the construction and health ministries that survive changes of minister
A construction market with capacity to absorb recovery works, which was lacking after the second earthquake
Limits: The World Bank finances the project and rates it (moderately satisfactory on both counts in June 2026), so this is a lender reporting on its own loan. The 40 percent is an intermediate indicator in the ISR's results table; the 16.32 percent disbursed and the extension from 31 December 2024 to 30 June 2027 (after an earlier extension from 30 June 2024) are in the Second Restructuring Paper of November 2024 (https://documents.worldbank.org/curated/en/099110624203564259/text/P173998-7d9a1555-e106-4a16-b043-2253208f955b.txt, paragraphs 1, 4 and 5). The loan is a small part of Croatia's recovery: the restructuring paper itself says about 1 billion euros of EU Solidarity Fund grants had to be spent within 18 months and became the government's priority, which is one reason the loan lagged, so this record shows the slow track, not the whole national response, and Croatia's own reconstruction fund figures for private homes were not read at an official source. The paper also cites a government change, four changes of ministry leadership in four years, the December 2020 Petrinja earthquake and construction labour shortages. The project builds hospital and school buildings that are not finished, so the record documents recovery in progress, not completed. Status is operating, not eroded, because no part of a delivered peak has been lost; the delay is the finding.
๐ณ๐ดNorwaysince 1950still operatingbears on Disaster preparedness and recovery
The Norwegian Civil Defence tested 1,130 public warning sirens in its national warning test on 14 January 2026 and reported an uptime of 89 percent, which it called a clear improvement on the January 2025 test and a slight fall from the June 2025 test.
The national civil defence keeps a network of public warning sirens in populated areas and sounds every one of them in a national test twice a year, publishing the share that worked and following up each failure.
Preconditions
A state civil defence organisation that owns and maintains the sirens rather than leaving them to municipalities
A central control system able to trigger and log every installation at once
Regular public tests, which double as a reminder of the signals
Where else it has been used: Germany dismantled its federal siren network after the Cold War (deu-siren-network) and is rebuilding it; Norway kept its network under the civil defence and tests it on a fixed calendar.
Limits: The civil defence reports on its own system, in a short news item that does not print the publication date; the test date is given as Wednesday 14 January and the comparison with 'January 2025' places it in 2026. A test is not a shock: the record shows that the warning hardware works on a set day, not that people were warned, sheltered or evacuated in a real event. Uptime of 89 percent means roughly one installation in nine failed, mostly from frost and cold according to the item, and installations disconnected for maintenance or relocation were not tested and are not in the denominator, which flatters the figure. The item does not give the earlier test percentages it compares with, nor the share of the population within siren range. Norway has had a cell broadcast system (nรธdvarsel) since 2024, which this figure does not cover. The started year is an approximation for the post-war civil defence siren network and was not read at the cited page.
๐ธ๐ฎSloveniasince 2023still operatingbears on Disaster preparedness and recovery
Slovenia's Ministry of Natural Resources and Spatial Planning reported in July 2025 that after the floods of 4 August 2023 it had completed all intervention and emergency works on 850 km of watercourses by June 2024, paid 35.7 million euros in advance payments for home repairs to more than 7,350 individuals, transferred 218 million euros in advances to municipalities and completed more than 430 watercourse repair sites, within a five-year recovery programme funded at 2.3 billion euros.
The government passed an intervention law, paid flood-hit households and municipalities in advance instead of after damage assessment, mobilised more than a thousand machines on rivers at once, and then adopted a five-year recovery programme with dedicated funding through a new reconstruction office.
Preconditions
A water directorate and concession contractors able to scale works more than tenfold in a year
Fiscal room and a law allowing advance payments before final damage assessment
EU Solidarity Fund support to cover part of the eligible public costs
Limits: The ministry reports on its own response at a press conference marking two years since the floods, and calls its results 'an excellent result', so the framing is the government's. The figures are stated as 'more than' and are not broken down by date, apart from the June 2024 completion of emergency works. Advance payments to 7,350 people are not completed repairs: the same page says more than 5,000 applications from individuals and firms for building repair were still being processed and that 'more than half' of beneficiaries had received all the funds they needed, so the household recovery was incomplete two years on. The 2.3 billion euro programme is money allocated, not spent; the ministry says watercourse works rose more than tenfold after the floods, with 240 million realised in 2024 and 332 million forecast for 2025. A separate release of the cohesion ministry (12 June 2025) says Slovenia had used 88 percent (376.6 million euros) of the 428 million euro EU Solidarity Fund grant, so part of the recovery was EU money. The record shows recovery after the shock, not the warning or evacuation before it; the August 2023 floods killed several people and damaged a large share of municipalities, which this record does not measure.
๐ฌ๐ทGreecesince 2023still operatingbears on Disaster preparedness and recovery
Greece's Ministry of Climate Crisis and Civil Protection paid almost 65 million euros of first state aid (proti arogi) to about 14,000 people affected by the September 2023 Thessaly floods within the first month after the storm, and by 15 February 2024 had paid 154.94 million euros to 45,292 households, firms, farms and livestock units in the first round.
A general secretariat for disaster recovery runs a single online platform where victims apply, cross-checks applications against tax and property data held by the revenue authority, and pays first aid in weekly batches within weeks of the event, before full damage assessment.
Preconditions
A tax administration with property and bank account data that can be matched to applicants
A standing legal framework for state aid after disasters, so no new law is needed per event
A ministry unit dedicated to recovery that survives between disasters
Where else it has been used: The same platform paid the Evros fire, Samos earthquake and Chios fire victims; the release pattern is identical across events.
Limits: The ministry reports on its own payments in press releases, and the one-month figure comes from the deputy minister's statement in a separate release of 13 October 2023 (https://arogi.gov.gr/deltia-typou/hr-triantopoylos-8i-katavoli-protis-arogis-65-ekat-eyro-se-14000-polites-mesa-se-peripoy-ena-mina/), worded as 'almost 65 million' to 'about 14,000'. Recipients mix households, businesses, farms and livestock units, so the count is not people or homes; first aid is an advance against later compensation and covers basic needs and household contents, not rebuilding. The release gives no figure for the number of eligible victims, so coverage cannot be computed. Speed of payment is a recovery measure, not preparedness: the record says nothing about warning, evacuation or the flood defences of Thessaly, where Storm Daniel caused deaths and flooded large areas. The 15 February 2024 cumulative covers households and firms hit by the flood 'in Thessaly, but also other areas', so it is not strictly limited to Thessaly.
๐ฉ๐ฐDenmarksince 1950operating below its peakbears on Disaster preparedness and recovery
Denmark's Emergency Management Agency counted about 4,693,000 places in public shelters, protective concrete trenches and building shelters at the end of 2001, and the municipal count it collected in 2022, the first since then, recorded about 3.54 million, with places in concrete trenches down from 152,335 to 107,658 and in building shelters (sikringsrum) from 3,743,822 to 3,307,400, after municipalities had stopped requiring shelters in new buildings.
The state built public shelters and protective trenches and required shelters in new buildings during the Cold War, then devolved shelters to municipalities in 1993, which stopped requiring new ones and let some be closed, until the 2022 security shock prompted the first national count in twenty years.
Preconditions
A building rule requiring shelters in new construction, which is what kept the stock growing and what lapsed
A national register of shelter places kept up to date, which Denmark did not keep after 2002
Municipal budgets for maintaining rooms that have no peacetime function
Where else it has been used: Germany (deu-public-shelters) and Norway (nor-tilfluktsrom) ran their shelter stock down after the Cold War in the same way; Finland (fin-civil-defence-shelters) kept the building requirement and its stock.
Limits: The 2022 total is computed here by adding the four national totals printed at the foot of the two-page table (107,658 trench places, 60,432 other public shelter places, 3,307,400 building shelter places and 63,105 supplementary places); the agency itself prints no grand total. Its release (https://www.brs.dk/da/nyheder-og-publikationer/publikationer2/alle-publikationer/2022/beskyttelsesrum-2022/) says the data were not yet validated, that building shelters are under-registered in the building register, and that it expects the consolidated count of building shelter places to be 'somewhat higher' than the table but lower than in 2002, so the size of the fall is uncertain. The 2001 figure is from the agency's August 2002 report on shelter preparedness (https://www.brs.dk/globalassets/brs---beredskabsstyrelsen/dokumenter/ovrige-emner/2020/-rapport_om_beskyttelsesrumsberedskabet-.pdf, PDF page 23), which also gives the categories; the supplementary category fell most (720,000 to 63,105) and may not be defined the same way in both counts. The same 2002 report says the need under the old target of one place per resident was estimated at 6.5 million places in 1991, so the stock never met the target. The erosion follows a policy choice: shelters have been a municipal responsibility since 1993 and municipalities stopped imposing shelters in new buildings. The record measures places on paper, not whether the rooms could be readied in time. The started year is approximate for the post-war shelter programme and was not read at either source.
๐ฑ๐ปLatviasince 2022still operatingbears on Disaster preparedness and recovery
Latvia's State Fire and Rescue Service tested the national early warning siren network on 23 May 2025 and reported that 160 of the 162 sirens tested worked without faults, out of 164 installed, against 119 of 155 in the test of 20 May 2024.
The fire and rescue service sounds every national warning siren twice a year, publishes how many worked by region, and repairs or relocates the failures before the next test.
Preconditions
A remotely triggered siren network owned by a national service
A published test result after every test, which makes failures visible
Maintenance funding, which rose after 2022 with the security situation
Where else it has been used: Norway's civil defence publishes the same kind of test result for a network seven times larger (nor-tyfonvarsling).
Limits: The fire service reports on its own system. A test is not a shock: the record shows that sirens sounded on a set day, not that people were warned or acted in a real event, and the item does not say what share of the population lives within hearing range of 164 sirens. Sirens not tested because they were being moved, repaired or renewed (two in May 2025, nine in May 2024) are outside the denominator. The 2024 figure is from the service's item of 20 May 2024 (https://www.vugd.gov.lv/lv/jaunums/parbaudot-155-trauksmes-sirenas-bez-traucejumiem-nostradaja-119-sirenas); other items in the series give 152 of 158 (May 2023), 138 of 159 (November 2023) and 151 of 158 (November 2024), and in June 2022 the remote activation software failed and the sirens had to be started by hand, so performance has swung between tests rather than improved steadily. No test result after May 2025 was found in the service's sitemap. The started year is the first test in the series read (June 2022), not the founding of the network; Latvia also sends cell broadcast and app alerts, which this figure does not cover.
๐บ๐ฆUkrainesince 2022delivered and closedbears on Institutional responsiveness
After Russia's full-scale invasion on 24 February 2022, Ukraine's Cabinet allocated UAH 21,450 million for a one-off eSupport payment by Resolution 199 of 4 March 2022, opened applications in the Diia app on 8 March to insured employees and sole proprietors in regions with active fighting, paid the first 6,500 hryvnia transfers on 21 March, and by 27 March had paid 2.7 million people UAH 17.9 billion against 4 million applications.
The Cabinet reused the existing eSupport payment rail, let insured workers in combat regions apply in the Diia app, checked them against social insurance records and paid a flat sum to their bank cards within four weeks of the invasion.
Preconditions
A national digital identity app with millions of users and verified bank cards
A social insurance register that could confirm employment and location quickly
A reserve fund the Cabinet could reallocate by resolution
Where else it has been used: The same rail paid later wartime transfers, and it is the Ukrainian version of what Togo (Novissi) and Brazil (Auxilio Emergencial) did in 2020: pay by phone first and check later.
Limits: The 2.7 million and the UAH 17.9 billion are the Ministry of Digital Transformation's own running count on the Government portal, not an audited outturn; the final number of recipients is not on this page. The live page at kmu.gov.ua answered a captcha to a script on the day of retrieval, so the record cites the Internet Archive copy of 31 March 2022, which carries the same text. The release says the service opened on 8 March and was open to employees for whom the single social contribution was paid and to sole proprietors of all groups in 14 listed regions and Kyiv, so eligibility was by region, not by a test that the person's work had stopped. The 25 days is our count: the start date of payments comes from a second Government portal release of 22 March 2022 ('Pochalysia pershi vyplaty 6500 hrn yePidtrymky', kmu.gov.ua), which says payments began the day before and had reached almost 500,000 people, and the allocation and its date come from Cabinet Resolution 199 of 4 March 2022 on kmu.gov.ua. Speed was bought with eligibility checks that were later found loose: press reports from August 2022 cite tens of thousands of payments to ineligible people, a figure this record does not carry because no official page with it was opened. A one-off flat transfer to workers in listed regions is a fast payment, not a durable programme change, and it relied on a payment rail (the Diia app and the existing eSupport scheme) that already existed before the shock.
๐ท๐ดRomaniasince 2020delivered and closedbears on Institutional responsiveness
Two days after the state of emergency was declared on 16 March 2020, Romania's Government made the unemployment insurance budget pay 75 percent of base salary to employees whose contracts were suspended (Emergency Ordinance 30 of 18 March 2020), and by mid-June 2020 the employment agency ANOFM had approved the indemnity for 2,067,970 employees and paid RON 2,937 million, with 241,453 employer requests approved within the legal deadline and 61 late.
The Government took the Labour Code's existing technical unemployment, which employers paid themselves, and by emergency ordinance moved its cost to the unemployment insurance budget for the emergency, with electronic filing and a statutory payment deadline for the county employment agencies.
Preconditions
An existing statutory form of temporary contract suspension that employers already used
County employment agencies able to process employer lists against the employment register
An unemployment insurance budget with a surplus the Government could redirect by ordinance
Where else it has been used: Most EU members ran a short-time work or suspension scheme in 2020, financed in part by EU SURE loans; Romania's was a suspension scheme rather than Kurzarbeit and reached about two million employees in three months.
Limits: The live PDF at curteadeconturi.ro returned a Cloudflare challenge to a script on the day of retrieval, so the record cites the Internet Archive copy of the same file. The figures are in a 162-page PDF: Table 9 (payments at 15 June 2020) and Table 10 (requests, beneficiaries and on-time approvals by region) on printed page 59, the 2,067,970 and the 2,483,890 employees requested on printed page 60, and the ordinance date (OUG nr. 30/18.03.2020) in the annex on PDF page 132. The audit is independent but the counts are 'date furnizate de ANOFM'. The legal deadline was the agency's own procedural one, and Table 10 counts requests, not days, so the record shows that the agency kept its deadline, not the median time to payment. 409,876 employees' requests were still pending at the cut-off. The indemnity was paid to employers, who passed it to staff; the report finds overpayments and payments to ineligible people among its RON 38.3 million of damages across all emergency spending. The scheme suspended contracts rather than shortening hours, so it is a wage replacement for idle workers, and its speed owes much to an existing Labour Code mechanism that the ordinance financed from the state budget.
๐ญ๐บHungarysince 2020delivered and closedbears on Institutional responsiveness
On 18 March 2020 Hungary's Government introduced a repayment moratorium until the end of 2020 that applied automatically to household and corporate loans disbursed before that date, and by June 2020 about 1.6 million of the roughly 2.7 million household borrowers used it on at least one loan, while 44 percent of corporate loans outstanding were in moratorium at the end of the second quarter.
1.6 million household customers, household customers using the moratorium on at least one loan (around 60 percent of about 2.7 million), 2020-06. 44 %, share of corporate loans outstanding in moratorium, 2020-Q2. Magyar Nemzeti Bank, Financial Stability Report, November 2020, retrieved 2026-10-02.
The move, and what it needed
The Government, on the central bank's proposal, suspended principal and interest repayments by decree for every existing loan unless the borrower opted out, which needed no application and no new administrative channel.
Preconditions
Emergency decree powers that let the Government change private loan contracts overnight
A banking supervisor that could tell banks how to book the suspended instalments
Banks capitalised enough to absorb nine months of deferred cash flow
Where else it has been used: Most EU members ran payment moratoria in 2020, usually opt-in and bank-led under EBA guidelines; Hungary's blanket opt-out design reached more borrowers faster and was harder to unwind.
Limits: The numbers are in a PDF: the 1.6 million household customers on PDF page 38 (Box 3), the 44 percent of corporate loans on page 39, and the 18 March 2020 start and its automatic application on page 48 (Box 5). The 1.6 million is the MNB's own calculation from bank-reported data. Because the moratorium was opt-out, participation measures inertia as well as need, and the report says the moratorium does not prevent credit risk rising. The speed figure is the date of a rule, not of a payment: a moratorium costs the budget nothing up front and shifts the cost to banks and to borrowers who keep accruing interest. The report records that the Government would extend it from January 2021 for vulnerable debtors; the record does not cover the later extensions or the losses when it ended. The MNB proposed the measure, so it is reporting on its own advice.
๐ท๐ธSerbiasince 2020delivered and closedbears on Institutional responsiveness
After the state of emergency of 15 March 2020, Serbia's Government presented a EUR 5.1 billion economic package on 31 March that included paying micro, small and medium firms one minimum wage per employee for three months if they cut staff by no more than 10 percent, and on 7 May 2020 the Ministry of Finance paid about RSD 32 billion as the first minimum wage for 1.055 million employees of 232,000 businesses.
The Ministry of Finance paid a flat minimum wage per employee straight into the accounts of every small private employer that opted in through the tax administration's portal, using payroll tax filings as the register, with a staff-retention condition.
Preconditions
Monthly individual payroll tax returns that already listed every employee by employer
A tax administration portal through which firms could opt in without a separate application
Fiscal space from a prior surplus, which the minister cited as what made the payment possible
Where else it has been used: A cruder, faster cousin of short-time work schemes; North Macedonia's 14,500 denar per employee measure of 2020 used the same flat design.
Limits: The page reports the Finance Minister announcing on 6 May that the payment would be made 'tomorrow', so the record rests on a ministerial statement of a payment due the next day, not on a Treasury outturn, and the figures are approximate ('priblizno 32 milijarde dinara'). The package date and the eligibility condition come from a second Government portal release of 31 March 2020 ('Za ublazavanje posledica epidemije na privredu 5,1 milijarda evra', srbija.gov.rs/vest/456168), which also names 15 March as the cut-off for firms that stopped trading; the 37 days are our count between the two releases. The subsidy was nearly universal for private small firms that applied, so it paid many employers who would have kept staff anyway, and the same release says firms also deferred RSD 29.5 billion of taxes and contributions. The page does not give the second and third payments or a final total. The 'condition' that firms not cut staff by more than 10 percent is a rule, and this record does not show whether it was enforced.
๐ง๐ฌBulgariasince 2020delivered and closedbears on Institutional responsiveness
Bulgaria's National Social Security Institute paid employers 60 percent of the wages and contributions of staff kept on during the COVID-19 emergency under Council of Ministers Decree No. 55 of 2020, starting with 68 employers and 1,065 employees on 15 April 2020, and by 4 March 2021 the Ministry of Labour and Social Policy reported BGN 884 million paid under the measure and 300,000 jobs kept.
The Council of Ministers decreed that the social security fund would pay 60 percent of wages and contributions for workers whose employers kept them on, with the employer paying the other 40 percent, and the Employment Agency approved applications that the social security institute then paid and published employer by employer.
Preconditions
Monthly employer declarations to the social security institute that identified each insured worker
An Employment Agency able to process employer applications in county offices
A social security fund with the liquidity to pay employers before the budget was revised
Where else it has been used: A short-time work design of the German type with a fixed state share; Bulgaria's distinctive move was publishing every paid employer by name, which made the scheme auditable by the public.
Limits: The BGN 884 million and the 300,000 jobs are a minister's statement on the ministry's own site, with no breakdown; 'jobs kept' is the number of supported positions, not a counterfactual estimate of jobs saved. The start figures are our sums from the National Social Security Institute's first payment list (nssi.bg, 'Izplateni kompensatsii po PMS No. 55 ot 2020 g.', file pms55_15042020.xls: 68 employers, 1,065 persons, BGN 274,274), one of 45 lists the Institute published to December 2021; one of them (14 May 2020) is truncated at the source and cannot be read, so we do not give a total from the lists. Speed was modest: the ministry reported only 643 employer applications covering 7,231 employees in late April 2020 (mlsp.government.bg, '643 zayavleniya ot rabotodateli sa postapili po skhemata 60/40'), and uptake grew through May and June. The decree date and the date of the emergency declaration are not on the pages opened, so the record times the response from the first payment, not from the shock.
๐ท๐ธSerbiasince 2014delivered and closedbears on Disaster preparedness and recovery
After the May 2014 floods evacuated about 32,000 people and flooded the Kolubara open-pit mines that fuel 40 percent of Serbia's generating capacity, the state utility EPS dewatered the Tamnava West mine by the first quarter of 2015 and kept supply at pre-flood levels with no recorded shortages, and the Directorate of Water Management rebuilt or upgraded 71 km of flood protection that by October 2019 covered 24,944 households against a target of 9,000.
The government used an emergency loan to buy power while the state utility pumped out its flooded lignite mines, then redirected the savings into rebuilding river defences to a higher standard.
Preconditions
A vertically integrated state utility able to import power at short notice
An emergency lending window that disburses within months
A water directorate with designs ready for embankment works
Limits: The figures are the World Bank's, in a completion report on an emergency loan it financed and rated; the event figures are in para. 1 (printed page 6), the power and mine results in para. 45 (pages 16 to 17), and the flood protection figures in para. 51 (page 19). Most of the loan bought electricity imports for the 2014 to 2015 winter, so 'no shortages' partly measures money, not preparedness, and the original power target was cut when demand proved lower. The 24,944 households exceeds its target mainly because funds were moved into flood works at restructuring, with a 22-month extension; the target itself was set low. Five substations were procured but never installed for lack of permits, and two meteorological radars were lost to mis-procurement, both before closing. The record is recovery from one event and does not show that warnings or evacuation worked in 2014, when 51 people died.
After the May 2014 floods displaced about 90,000 people and directly affected about a million in Bosnia and Herzegovina, the two entities' agriculture and water ministries and Brcko District distributed US$27 million of emergency goods and restored 446 public service facilities to pre-flood condition against a target of 200, including more than 260 km of rural roads and 53,460 m of flood defences, by the project's June 2019 closing.
The state treasury on-lent an emergency loan to the two entities and Brcko District, whose ministries bought relief goods through strategic reserves and then rebuilt municipal and regional infrastructure from a list drawn from the joint needs assessment.
Preconditions
A joint recovery needs assessment agreed by all levels of government within weeks
Entity-level strategic reserves able to procure and distribute goods
Municipalities with project lists ready for rebuilding
Limits: The figures are the World Bank's, from a completion report on its own US$100 million emergency loan; the event figures are in the context section (printed page 5), the outcome table on page 13, and the road and flood defence figures on page 18. The 446 facilities exceeds its target partly because unused emergency-goods funds were moved into infrastructure in January 2016, so the target was not a fixed plan. 'Facilities' counts very different things, from a culvert to a bridge, and the report does not say how many were fully rebuilt. Emergency goods reached beneficiaries according to third-party audits in 2017, but some agricultural goods arrived late. The project closed six months after its December 2018 original date, five years after the floods, so this is slow recovery, not rapid response, and it covers one loan, not the whole post-flood programme financed by the EU and others.
๐ท๐ดRomaniasince 2004delivered and closedbears on Disaster preparedness and recovery
Romania's Ministry of Administration and Interior and Ministry of Regional Development completed seismic retrofitting of 44 high-risk public buildings, where about 23,350 people worked, and rolled out a modernised emergency communication system to 48 sites under a project approved in May 2004, after cutting the retrofit target from 84 buildings to 40 in 2010 and closing in June 2012 instead of December 2009.
The government used a development loan to strengthen the most dangerous public buildings in the Vrancea seismic zone and to give the emergency inspectorate a shared voice and data network, then trimmed the list as costs rose.
Preconditions
A seismic risk inventory that ranks public buildings
Counterpart budget that survives a fiscal crisis
Building codes aligned with the EU norms the designs must meet
Limits: The figures are the World Bank's, from a completion report on its own loan, in the PDO indicator table near the front of the document and the restructuring section. The report itself says there was no opportunity to test the retrofitted buildings in the absence of a major earthquake, so the record shows preparation, not demonstrated performance in a shock. The 44 exceeds a target already cut from 84 to 40 because of construction cost increases and currency depreciation, and three buildings were left unfinished when the 2010 to 2011 budgets did not cover ongoing works. The 2007 EU accession forced redesigns, and the government changed seven times in eight years. More than 350 public facilities were assessed at risk at baseline, so 44 is a small share, and the separate Bucharest programme for residential buildings at seismic risk is not covered.
๐ฒ๐ฉMoldovasince 2010delivered and closedbears on Disaster preparedness and recovery
Moldova's State Hydrometeorological Service, with the emergency situations department, installed a Doppler radar by June 2014 and extended the lead time of severe weather warnings from 10 minutes to 1 hour at baseline to 3 to 6 hours by 2016, against an original target of 12 hours, under a project approved in August 2010 that closed in September 2016, two years late.
The environment ministry's weather service bought a Doppler radar and forecasting systems and wired its warnings into the emergency department's command centre, so warnings reach responders hours rather than minutes ahead.
Preconditions
A national weather service with staff to run a radar and models
A link from the forecaster to the civil protection command centre
Budget to maintain the radar once the loan ends
Limits: The figures are the World Bank's, from a completion report on its own credit, in the PDO indicator table near the front of the document; the achieved lead time is a range (3 to 6 hours) and the record's metric takes its upper bound. The 12-hour target was formally revised down, so 'achieved' is against the lowered target. No named severe weather event is shown in which the warnings were issued and acted on, so the record documents capacity built, not capacity used; it is filed on the same terms as other warning-infrastructure records in this corpus. The report notes repeated management changes at the hydromet service, which threaten sustainability, and gives no later figure for radar uptime.
๐ฆ๐ฑAlbaniasince 2020still operatingbears on Disaster preparedness and recovery
Albania's government reported in November 2021, two years after the 26 November 2019 earthquake, that more than 33,400 families had received reconstruction and repair grants worth ALL12.7 billion, that 13,325 people were receiving rent bonuses while new homes were built, and that 27 schools had been completed, while promising 4,500 individual houses and 12,000 apartments by spring 2022.
The government created a reconstruction ministry and programme within months of the earthquake, paid rent bonuses to displaced families and repair grants to owners through municipalities, and built new housing blocks and schools on cleared sites with state and donor funds.
Preconditions
A damage assessment that classifies each building, so grants match damage
A state budget line and donor pledges large enough to fund rebuilding at once
Municipalities able to administer grants and clear sites
Limits: The source is the government's own promotional page on its flagship programme, in Albanian; the live kryeministria.al page sits behind a bot wall, so the cited copy is the Internet Archive capture of July 2026. The figures are counts of grants, with no report of how many repaired homes were verified, and 'more than 33,400' is a floor. The completion promises for spring 2022 (4,500 houses, 12,000 apartments, no trace of the earthquake) are on the same page, and no official figure was found showing whether they were met, so status is left as operating. The page says EUR300 million of grants pledged at the donors' conference were active in more than 30 projects; the delivery is the government's, but the money is not all domestic. Nothing here measures preparedness before the earthquake.
๐ฆ๐ฟAzerbaijansince 2020delivered and closedbears on Institutional responsiveness
After Azerbaijan's first COVID-19 case on 28 February 2020 and the first lockdown in March, the government paid a monthly allowance of AZN 190 to 600,000 unemployed citizens, formal and informal, from April to June 2020, and in 2020 also granted targeted social assistance to 9,000 newly poor families without the usual household assessment and extended expired assistance for 9,408 households.
The government created a temporary flat allowance for anyone who registered as having lost work, including informal workers, through an online registration with the State Employment Service, and relaxed the targeted social assistance rules so newly poor families could be added without a home assessment.
Preconditions
An online registration route with a telephone fallback through district employment centres
Budget resources to pay first and refinance part of the cost later from a donor loan
A targeted social assistance programme whose eligibility checks could be relaxed by decision
Limits: The figures are in paragraph 14 and footnote 4 on page 6 of 9 of a World Bank project information document, written a year later to prepare a loan that would retroactively finance the payment for 170,000 of the 600,000 recipients; it relays the government's figures and is not an evaluation. The document says 'from April to June 2020' but not whether all 600,000 were paid in April or how many months each received, so the speed shown is the month, not days; registration was online, with a phone route through district employment centres. 600,000 is a round number. The first case date and 'three lockdowns between March 2020 and January 2021' are in paragraph 4 (page 4). The document calls AZN 190 commensurate to the cost of living, and the payment stopped after three months; the 9,000 families added to targeted social assistance were few next to 600,000 allowance recipients, which suggests the standing safety net expanded little.
๐ฌ๐ชGeorgiasince 2020delivered and closedbears on Institutional responsiveness
Georgia, which had no unemployment benefit, announced a temporary GEL 200 monthly benefit for laid-off formal workers and a one-off payment for self-employed and informal workers on 24 April 2020, a month after declaring a state of emergency, and by September 2021 had paid 231,487 unemployed people and 307,236 self-employed people.
The government used the Revenue Service's payroll tax records to identify laid-off formal workers and pay them a flat monthly benefit through the State Employment Support Agency, and opened an online portal for a one-off payment to self-employed and informal workers with no declared income.
Preconditions
Monthly payroll tax declarations linking workers to employers and bank accounts
An employment agency able to verify eligibility against social assistance lists
External emergency finance (World Bank and AIIB) committed within weeks
Limits: Both counts are in one sentence on page 1 of the status report, with no table behind them and no date of first payment; the timing comes from the project appraisal document (http://documents.worldbank.org/curated/en/949581588712278110/pdf/Georgia-Emergency-COVID-19-Response-Project.pdf: state of emergency on 21 March 2020 in paragraph 9, printed page 12; the Prime Minister's 24 April announcement in paragraph 26, printed page 18; the GEL 200 benefit for up to six months and the GEL 300 one-off benefit in paragraphs 45 and 46). The speed shown is therefore 34 days from emergency to announced rule, not to money paid. The World Bank and AIIB co-financed the programme, the report notes that its indicator values cover only the IBRD share of some outputs, and the counts are cumulative people paid at least once, not people covered for six months. The benefit was temporary by design and Georgia still has no permanent unemployment insurance, so the rule change did not outlast the shock. The self-employed figure is described as people 'whose activity was affected', a looser category than the one-off benefit's eligibility rule.
๐ฏ๐ดJordansince 2020delivered and closedbears on Institutional responsiveness
Jordan's National Aid Fund reused the Takaful cash-transfer platform after the March 2020 curfew, opened online registration for ten days in mid-April and had paid about 150,000 newly identified households by 10 May 2020, and 237,000 households had received three months of temporary cash support by March 2021.
The National Aid Fund opened the Takaful online registration platform to informal workers who had lost income, selected households with the Takaful targeting formula against linked administrative data, and paid them into e-wallets and bank accounts authorised by the central bank.
Preconditions
A cash-transfer registry and online enrolment platform built the year before
Central bank rules allowing the Fund to open e-wallets for beneficiaries remotely
Domestic charity accounts able to fund the first months before external finance arrived
Where else it has been used: The same reuse of a Takaful-style platform appears in Morocco (mar-tadamon-covid), where a health-assistance register played the role of the cash-transfer registry.
Limits: The 150,000 figure is in paragraph 60 of the appraisal document (printed page 28 of 60), and the curfew dates are in paragraphs 2 and 3 (printed page 9). The 237,000 figure is in a second World Bank document, the Implementation Status and Results Report of 26 April 2021, sequence 2 (http://documents.worldbank.org/curated/en/429691619464662394/pdf/Disclosable-Version-of-the-ISR-Jordan-Emergency-Cash-Transfer-COVID-19-Response-Project-P173974-Sequence-No-02.pdf, page 1), which also says the transfers cost about US$114 million, of which the government paid about US$103 million through national charity accounts and the project financed US$11 million retroactively. Both documents come from the lender to the programme, and the PAD figure is an 'about'. The speed is the delivery system's, not a new law's: the Takaful platform, its registry of almost one million households and digital payments were built from 2019 with World Bank, UNICEF, WFP and other donor support, so the response mostly reused prior investment. The emergency transfer lasted three months per household and the project's own PDO indicator for cash support through the project still read zero in that report because the loan became effective only on 11 November 2020: the fast part was financed domestically. Refugees, about 655,000 registered, were outside this programme and were served by UN agencies.
๐ฐ๐ฟKazakhstansince 2020delivered and closedbears on Institutional responsiveness
During the March to May 2020 state of emergency, Kazakhstan's government paid people who lost income because of business closures and lockdown KZT 42,500 per person per month, and about 4 million people received the payment.
The government declared a state of emergency and attached to it a flat monthly payment for people who had lost income because of closures and lockdown, paid through the state benefit system for the duration of the emergency.
Preconditions
Fiscal room for an anti-crisis package the World Bank put at 5.7 percent of GDP
Population and tax registers to check applicants against
A state benefit payment system already reaching most households
Limits: The figure is a rounded 'about 4 million people' in Box 1 on printed page 7 (PDF page 8) of a World Bank economic update; the Bank did not finance or evaluate the payment and relays the government's own figure. The government release with the exact count (gov.kz, Ministry of Labour and Social Protection) renders only through JavaScript and could not be opened for verification. The update gives the state of emergency's dates (quarantines from 19 March, lifted on 11 May 2020) but no date of first payment and no count of people paid in each month, so the speed claim rests on the payment having been made during an emergency that lasted under two months. The update publishes no targeting, error or rejection figures, so it cannot show how well the payment reached the people it was meant for. The update puts the whole anti-crisis package at 5.7 percent of GDP, of which this payment was one part. The payment was temporary and ended with the emergency.
๐ฑ๐งLebanonsince 2021still operatingbears on Institutional responsiveness
Lebanon's emergency cash transfer for the economic collapse that began in 2019 was approved by the World Bank in January 2021, rewritten by Parliament in March 2021 and declared effective only in July 2021, so that in October 2021 the number of households receiving basic cash transfers still stood at its 15,153 baseline against a target of 147,000; the programme later reached 158,770 households by March 2025.
The government borrowed for a new cash transfer rather than extending an existing one, and the loan had to pass Parliament, which amended it, so the programme waited for a revised law, a new government and fiduciary conditions before it could pay.
Preconditions
A cash-transfer instrument that requires parliamentary ratification of each external loan
No functioning domestic financing for safety nets during a banking and currency collapse
A donor-funded legacy programme whose end, in June 2024, showed how little was domestically owned
Where else it has been used: The contrast with Jordan (jor-naf-covid-cash), which paid 150,000 new households within two months by reusing a standing platform and domestic funds, is the point of the record.
Limits: The 15,153 is the PDO indicator table on pages 2 and 3 of the status report, identical at baseline (30 June 2020), March 2021 and October 2021; the approval, ratification and effectiveness dates are on page 1, which says Parliament's changes cut budget allocations for fiduciary activities and that the Bank judged some of them would impede implementation, and that the government still had to meet disbursement conditions so that payments could begin before the end of 2021. The 158,770 is in paragraph 4 of the World Bank restructuring paper of 13 May 2025 (https://documents.worldbank.org/curated/en/099051325103513216/pdf/P173367-bed938d8-7d2a-4120-80a5-3da01c595ba8.pdf), which says about 94,000 households were enrolled in 2022 and 2023 and 65,000 were migrated in October 2024 from the National Poverty Targeting Program after it closed in June 2024 for lack of donor funding. The baseline households were beneficiaries of that older programme, not of this one. The record documents slowness, not failure: once paying, the programme reached roughly its target. The publisher is the lender, and part of the delay was the World Bank's own conditions and the process of agreeing them with Parliament across a change of government in September 2021; the record does not apportion blame. The start year is the year of approval, not the start of the crisis in late 2019.
๐ฒ๐ฆMoroccosince 2020delivered and closedbears on Institutional responsiveness
Morocco created a Special Pandemic Fund in March 2020 and, through its Tadamon budget line, paid emergency cash transfers of MAD 800 to 1,200 to households of informal workers, reaching 4.2 million households in the first wave in April 2020 and 5.45 million across the first and second waves, for a total of MAD 16.15 billion by October 2020.
The Ministry of Economy and Finance opened a dedicated budget line in a pandemic fund, used the RAMED health-assistance register to pay listed informal-worker households by mobile phone without a new application, and opened a one-week web platform for informal households outside the register.
Preconditions
A large existing register (RAMED, about 30 percent of the population) with household composition and phone contacts
A payment channel that could reach unbanked households through mobile cash withdrawal
A fund able to pool budget and private contributions within days
Where else it has been used: The emergency transfers fed directly into the 2020 royal social protection reform and the unified social register; RAMED was replaced by AMO-Tadamon in December 2022.
Limits: The figures are in paragraphs 13 and 14 of a PDF (printed pages 12 and 13 of 103, PDF page 18): 4.2 million households in the April 2020 first wave, 1.25 million added later through the grievance mechanism, 5.45 million across the first two waves (2.8 million RAMED and 2.65 million non-RAMED), a third wave in July 2020, and MAD 16.150 billion paid by October 2020. The World Bank wrote this appraisal to finance the programme retroactively (US$160 million for the Tadamon transfers), so it is a lender describing a programme it was about to fund, not an independent evaluation, and the document gives no exact payment dates beyond the month. Speed is shown only in months: the fund was set up in March 2020, enrolment for non-RAMED households was open on 10 to 16 April, and the first wave was paid in April. The 5.45 million counts households paid at least once, not households kept whole: transfers were flat amounts by household size and stopped after three waves. The 1.25 million added through grievances means the first list missed about a fifth of the households it finally paid, so targeting was fast but rough. The fund was financed in large part by contributions from public enterprises, banks, firms and individuals, not only the budget.
๐ท๐บRussiasince 2020delivered and closedbears on Institutional responsiveness
Russia's government adopted rules on 11 May 2020, the day the President announced it, for a one-off payment of 10,000 rubles per child aged 3 to 16, applied for through the state services portal, and on the first day of payment, 1 June 2020, the Pension Fund paid it for 15.3 million children, more than 150 billion rubles.
The government made the payment universal by age, took applications only through the state services portal and the Pension Fund, and checked them automatically against the digitised civil registry and tax data, with a legal limit of five days to decide and three to pay.
Preconditions
A national e-government portal already used by tens of millions of people
A digitised civil registry of births linked to parents
Accumulated reserves, which the President cited when announcing the payment
Limits: The 15.3 million children are the Minister of Labour Anton Kotyakov's spoken figure at a briefing on 2 June 2020, the day after payment began, given as the count 'as of this moment' and described as paid in one day; the 'more than 150 billion' rubles he added is that count times 10,000 rubles, not a separate payment total. It is not a statistical release. He said about 22 million citizens were expected to use the payment. The decree date (No. 652 of 11 May 2020), the President's announcement on the same day, the 5-day review and 3-day payment limits and the 1 June start are in the government's notice of 11 May 2020 (http://government.ru/news/39668/), and the 21 days is our count from 11 May to 1 June. The government and its ministers are reporting on their own measure. The payment was universal for the age group, not targeted at income loss, so speed was bought by not checking need: eligibility needed only the child's age and the parent's application, which the portal could match to civil registry records that, the head of the Federal Tax Service said at the same briefing, had been back-filled with more than 25 million records. A universal per-child cash payment is a broad response; it says nothing about how fast the state adjusted targeted programmes such as unemployment benefits.
๐น๐ณTunisiasince 2020delivered and closedbears on Institutional responsiveness
After the March 2020 lockdown, Tunisia's government announced in April 2020 a vertical and horizontal expansion of the AMEN Social programme: a TND 50 top-up for two months to all 260,000 poor households, two TND 200 transfers to 450,000 vulnerable households on the subsidised health card, and a one-off TND 200 transfer to 300,000 further vulnerable households enrolled through a new simplified platform.
The Ministry of Social Affairs topped up existing cash-transfer beneficiaries, paid households already listed for subsidised health care, and opened a simplified enrolment platform to reach vulnerable households outside the registry.
Preconditions
A social registry with accurate identification for most poor and vulnerable households
A legal framework for the AMEN Social programme adopted before the shock
Data cross-checks against other administrative registries to vet new applicants quickly
Limits: All three figures are in paragraph 23 of a PDF (printed page 15 of 86, PDF page 21), written a year after the event by the World Bank as it prepared a loan to finance the programme's next phase; it is a lender's description, not an audit, and the counts are round numbers. The document dates the announcement to April 2020 and says the households received the transfers, but it gives no payment dates, so the speed claim rests on the month of the announcement, not on the month money arrived. The 300,000 are households the Ministry of Social Affairs enrolled through a platform it built for the crisis, a horizontal expansion; the other two groups were already known to the state. The amounts were small (TND 200 is about US$73 at the document's rate), and the transfers were one-off or two-month measures. The responsiveness rested on the AMEN Social programme and registry built since 2019 with earlier World Bank support, which the same paragraph credits.
๐บ๐ฟUzbekistansince 2020delivered and closedbears on Institutional responsiveness
Uzbekistan's government expanded its COVID-19 unemployment support from a one-off to multiple benefits, which the World Bank followed by restructuring its emergency project on 12 August 2020, and with that financing had paid US$50.77 million in low-income family allowances to more than 1,000,000 families and US$0.65 million in unemployment benefits to 19,700 people by 1 January 2021.
The Ministry of Finance channelled an emergency donor credit into the existing low-income family allowance and a temporary unemployment benefit, changing the benefit from one payment to several when the shock outlasted the first plan.
Preconditions
An existing means-tested family allowance with a beneficiary list
Emergency external finance, here an IDA credit approved within weeks
A presidential resolution to authorise spending, which was the slow step
Limits: The figures are in paragraphs 2 and 6 of a World Bank restructuring paper, the lender's account of its own project. The record shows slowness as much as speed: the project was approved on 24 April 2020 but became effective only on 14 July 2020, the paper says it took the government three months to issue the Presidential Resolution enabling implementation, and payments under the component began only in November 2020, more than six months after the project was approved. The 1,000,000 families is 'more than', the families were recipients of an existing allowance financed partly from the project rather than new beneficiaries, and the paper does not say how many were added because of the pandemic. Unemployment benefit reached only 19,700 people, which says how narrow formal unemployment support was. Support the government paid from its own budget outside this project is not covered by this source.
๐ฐ๐ฟKazakhstansince 2024delivered and closedbears on Disaster preparedness and recovery
Kazakhstan's government reported that during the spring 2024 floods, which it called the largest in 80 years, more than 120,000 people were rescued and placed in temporary accommodation centres, and that by December 2024 it had bought 5,767 homes, fully repaired 9,156 and built 2,680 new houses for owners of the 19,359 flooded homes inspected, while 36,455 families received one-off assistance.
The Ministry of Emergency Situations coordinated a national and regional headquarters that evacuated whole settlements by land and air, and the government then used its reserve to buy housing on the market, repair, and build new houses within the same year.
Preconditions
A civil protection system able to mobilise army, police and National Guard under one command
A government reserve fund large enough to pay for housing within months
A functioning housing market from which homes could be bought quickly in the affected regions
Limits: The government reports on its own response in a press review, with no independent verification, and the page exists only in Kazakh (the English and Russian addresses redirect to the home page); figures were translated by me. 'Rescued and placed in temporary accommodation' counts evacuations, not people saved from death, and the page gives no count of deaths or of people who stayed. The housing figures answer only for homes that were the owner's sole dwelling (19,359 inspected), so second homes and renters are outside them, and the three remedies (purchase, repair, new build) add to 17,603 against 19,359 inspected by my sum, so about 1,750 cases are not accounted for on the page. The page itself says the scale could not have been forecast, which bears on preparation: the response was large but the warning system did not anticipate the event, and the President afterwards created a Committee for the Prevention of Emergencies inside the ministry. Spending figures are nominal tenge. Status 'concluded' rests on a December 2024 account; later complaints about the quality of new houses, if any, were not checked.
๐ฆ๐ฒArmeniasince 1994delivered and closedbears on Disaster preparedness and recovery
Armenia's Ministry of Economy, through a project implementation unit, restarted reconstruction in the zone of the December 1988 Spitak earthquake that had stopped when Soviet work crews withdrew in 1991, and by April 1997 had provided 2,857 housing units against 2,069 planned, completed 10 factories against 7 and laid 41 km of water pipes against about 27 km, under an IDA credit closed in June 1997.
A small, politically insulated implementation unit under the economy ministry took over half-finished Soviet reconstruction sites and completed them through competitive local contracts, choosing nearly finished buildings to maximise habitable units per dollar.
Preconditions
Sunk investment in partly built blocks that could be finished cheaply
Competitive local contractors willing to bid in dollar-equivalent contracts
An external credit when the state budget could not fund reconstruction
Limits: The World Bank wrote the report on a credit of about US$28 million. The unit counts are in Table 5 (Key Indicators for Project Implementation, printed page 17), which is OCR text with errors (the bathhouse row repeats the housing figure). The earthquake left about 530,000 people homeless (Part I, para. 1), so 2,857 units is a small fraction of the need, delivered five to eight years after the shock; the record shows recovery capacity in a newly independent state restarting a stalled Soviet programme, not a timely national recovery. The ICR says beneficiaries were slow to move out of temporary 'domiks', and that many families were still in temporary shelter in 1997. Units count apartments of different kinds (completion of unfinished blocks, repair and strengthening) and the project's single-family starter houses were scaled back. The ICR rates implementation satisfactory and credits an independent implementation unit, not a ministry line department.
๐ฐ๐ฌKyrgyzstansince 1999delivered and closedbears on Disaster preparedness and recovery
After record floods in southern Kyrgyzstan in 1998 damaged about 1,000 houses, the Department of Water Resources rebuilt or built 27.4 km of river protection embankments against 8.8 km planned, directly protecting about 40,000 people, and in 2003 a flood greater than that of 1998 passed the new embankments without significant damage or overtopping.
After a record flood the water department rebuilt the failed river defences to a 1-in-200-year standard with more durable rock and gabion-mattress linings instead of patching them, and created a dedicated agency to maintain embankments.
Preconditions
Emergency credit available within months of the flood
Hydrological records long enough (from 1927) to set a design flood
A maintenance budget for the new agency, which the report flags as limited
Limits: The World Bank wrote the report on an IDA credit of about US$10 million and rated the outcome Satisfactory. The figures are in the outputs section and the summary of achievements; the 2003 test is described with a reference to Table 11.3 of the report, which is OCR text of a 2004 PDF. The 40,000 people protected is the report's estimate, not a count, and 'no significant damage' is the report's judgement; it also says sections downstream of the rebuilt embankments were weakened by the same 2003 floods and needed emergency repairs. The works took until March 2004 against a planned closing of September 2001, in part because floods in 2002 and 2003 swept away half-finished works and Uzbekistan closed its border to rock supplies. The record evidences preparation for the next flood at a few sites in three raions, not national flood preparedness, though the project set up a Flood and Mudflow Protection Agency responsible for some 460 km of embankments nationwide; whether that agency is still funded was not checked.
๐น๐ณTunisiasince 1984delivered and closedbears on Disaster preparedness and recovery
After the October 1982 flood in Sfax killed at least 70 people and destroyed about 700 houses, Tunisia's Urban Hydraulics Directorate built an eight-km belt canal and rehabilitated the city's wadi channels and dikes, completing the works in June 1988 as scheduled, and the January 1990 floods that damaged much of central and southern Tunisia did not damage the protected area.
After a flood with an estimated 130-year return period, the urban hydraulics directorate built an intercepting canal around the city to divert runoff from the catchment above it, plus storm drains inside it, and set up a regional maintenance section.
Preconditions
A 1978 master plan and long hydrological series that let a feasibility study finish within three months
Land expropriation for a canal corridor on the urban edge
A maintenance unit with a recurring budget, which the report doubted
Limits: The World Bank, which lent US$25 million (US$22.3 million disbursed), wrote the report. The belt canal length is in the project description and the completion date in the timetable table (Table 2, planned and actual June 1988); 'completed according to the original schedule' is para. 5.01. The evidence that the works protected the city is one sentence noting that the January 1990 floods did not damage the project area, which the report says was acknowledged by the press and local residents; no rainfall or discharge comparison is given, so the 1990 event may have been smaller in Sfax than 1982. The report warns that protection depends on maintenance funding without cost recovery, and recommends an impact evaluation within five years; none was read. This is a 1984 to 1988 build and says nothing about current Tunisian flood preparedness. The text is OCR of a scanned 1991 report.
๐ฆ๐ฟAzerbaijansince 2019delivered and closedbears on Disaster preparedness and recovery
Azerbaijan's Minister of Emergency Situations reported on 5 December 2019, ten months after the February 2019 Shamakhi earthquake, that 580 houses had been built and 1,175 houses restored in Shamakhi district with state funds, while three apartment buildings with 174 flats were, by the President's account on the same page, still under construction.
The Ministry of Emergency Situations acts as the builder after an earthquake, demolishing damaged houses and building replacements with state money within the year, under direct presidential instruction instead of through insurance.
Preconditions
A ministry with its own construction capacity and direct budget access
Oil-revenue fiscal room to fund reconstruction without insurance
Political priority set from the top, which also makes the reporting promotional
Limits: The figures are spoken by the emergency minister during a presidential visit and published on the President's website, in a text that is largely praise of the President; there is no independent verification, no count of houses damaged to set them against, and no cost. The President says on the same page that the three high-rise buildings with 174 flats were still under construction and that work was also done in Aghsu and Ismayilli districts, which the figures do not cover. The page says there were no fatalities, which reflects the earthquake's size as much as preparedness. The earlier May 2012 Zagatala and Gakh earthquake, where the ministry's commission found 3,336 private houses unusable in Gakh alone and construction had started on 266 by September 2012 (president.az articles 4823 and 5940), shows the same model at larger scale but with no completion figure read. The record shows rapid state-financed reconstruction, not risk reduction: nothing on the page concerns seismic codes or warning.
๐บ๐ฟUzbekistansince 2020delivered and closedbears on Disaster preparedness and recovery
After the Sardoba reservoir dam broke on 1 May 2020 in Syrdarya region, Uzbekistan's government evacuated about 90,000 people from 24 mahallas of three districts, and by October 2020 had built 37 five-storey blocks with 1,480 flats in Sardoba district's Dustlik mahalla, given free to flood victims, and 86 five-storey buildings in Sardoba and Mirzaabad districts.
The president set up a government commission, assigned each region of the country a mahalla to rebuild with its own construction teams, and the state built free apartment blocks for flood victims within five months.
Preconditions
Central control over regional governments and state construction firms
Budget and donor money (including a private US$10 million donation) available within weeks
Land under destroyed houses left with owners, which eased acceptance of flats
Limits: The shock was itself a failure of preparedness: the president's own site calls it a 'man-made accident on the reservoir', a dam that failed within a few years of filling, so the record shows recovery after a failure the state caused, not readiness. The figures come from presidential news items (the 5 May address for the evacuation, the 13 October item for the flats, and 'Construction projects reported', 26 October 2020, https://president.uz/en/lists/view/3916, for the 86 buildings) with no independent check, and they mix counts of flats and of buildings. The plans of 15 May 2020 (lists/view/3572 to 3574) name about 4,500 affected homeowners and further blocks of 360 flats in Mirzaabad and 800 in Akaltyn plus repairs to hundreds of cottages; whether those were finished was not read, so the 1,480 flats show part of the programme. The evacuation figure is a round number in a presidential address, with no breakdown or timing beyond 'immediately', and the same address reports a separate storm in Bukhara region days earlier. Builders were mobilised from every region, so the speed reflects command mobilisation rather than a standing recovery system.
๐น๐ฏTajikistansince 1998delivered and closedbears on Disaster preparedness and recovery
After floods in 1998 caused damage estimated at US$40 million, about 2 percent of GDP, Tajikistan's Republican Project Coordination Unit for Liquidation of Consequences of Natural Disasters had repaired about 210 km of roads and fully completed seven bridges and structures by the credit's closing on 31 December 2001, with about 80 km of road repairs and three bridges unfinished, and the World Bank rated the outcome unsatisfactory.
A government emergency coordination unit repaired flood-damaged bridges, roads and embankments with an emergency IDA credit through local contractors, without the supervision capacity to hold them to quality.
Preconditions
An emergency lending window that disburses within months
Local contractors with experience of the work, which were scarce after the civil war
A coordination unit with financial controls the lender trusts, which it lacked
Limits: The World Bank wrote the report and rated the outcome Unsatisfactory and sustainability Unlikely, mainly over poor works quality and financial management; it refused a further extension and closed the credit as planned. The record is filed as evidence of weak recovery capacity, not of success. The report says the roads actually damaged totalled about 105 km of the 210 km repaired, that works quality was poor because local firms lacked experience and supervision, and that only three of eight planned power substations were completed. The project came a year after the end of a five-year civil war, with a supplemental credit after further floods in 1999, so capacity was at a low point. The text is OCR of a scanned 2002 report, and the data sheet and narrative disagree on the original closing date (January 2000 against June 2001).
๐ท๐บRussiasince 2014still operatingbears on Disaster preparedness and recovery
Russia's federal hydrometeorological service Roshydromet raised the accuracy of its warnings on hazardous hydrometeorological events delivered to regional emergency agencies from 81 to 83 percent to 91 percent, and of its warnings on natural hazards from 88 to 95.8 percent, between 2014 and the end of 2022 under its second national modernization project.
The national weather service modernised its observation network, computing and forecasting models over two successive World Bank projects, and measured itself on the verified accuracy of hazard warnings passed to emergency agencies.
Preconditions
A national hydromet service with a single federal budget and network
Successive projects over a decade, the first (RHM-1) building the baseline
A verification method agreed with WMO so accuracy is measured the same way each year
Limits: The World Bank wrote the report without site visits or formal interviews, because it suspended disbursements on 19 March 2022 (para. 1), so the indicator values are Roshydromet's own figures as reported to the Bank. The figures are in para. 49 and the results annex (ICR pages 18 and 37). Warning accuracy is a forecast-verification statistic, not a measure of whether warnings reached people or prompted action, and the record ties it to no single disaster; it is a preparedness capacity figure of the same kind as the South African radar availability record. The report says the improved reliability let Roshydromet sign, for the first time, binding agreements with the Ministry of Emergencies on hazard warnings. The final cost was US$174.89 million, 25 percent above appraisal, with the government financing what the Bank suspension left. Status 'operating' assumes the system still runs; nothing after 2022 was read.
๐ช๐ฌEgyptsince 2021delivered and closedbears on Disaster preparedness and recovery
Egypt's health authorities reported to WHO that 56,907,319 people, 56 per 100 population, had received at least one COVID-19 vaccine dose by 31 December 2023, from a first dose given on 24 January 2021, with 42,337,175 (41 per 100) completing the primary series.
Egypt's health ministry ran a national COVID-19 campaign through its primary care network and mass vaccination sites and reported progress to WHO.
Preconditions
A routine immunisation programme and cold chain
A supply of doses large enough for a national campaign, however obtained
Registers good enough to count who received a second dose
Limits: The figures are Egypt's own reports to WHO (DATA_SOURCE 'REPORTING'), with no independent coverage survey behind them. Completion of the primary series reached 41 per 100 and boosters 15,217,352 people (15 per 100), so the record shows a partial delivery. The row reports 102,191,677 doses in total, 100 per 100 population. The VACCINES_USED field is empty, so the record cannot show which products were used or where the doses came from, and it evidences distribution only. The per-100 rates divide by WHO's population estimate. Vaccination is a response to one shock, not proof of general preparedness, and it is filed under the precedent of bol-covid-vaccination and ven-covid-vaccination as a judgement call because no hazard-specific delivery with an official figure was found for this country in this batch. The source is a CSV with no prose; the numbers are read from one row. The cited URL is the blob the WHO dashboard serves, the same one bol-covid-vaccination cites.
๐ฏ๐ดJordansince 2021delivered and closedbears on Disaster preparedness and recovery
Jordan's health authorities reported to WHO that 4,821,579 people, 47 per 100 population, had received at least one COVID-19 vaccine dose by 21 August 2022, from a first dose given on 13 January 2021, with 4,558,313 (45 per 100) completing the primary series.
Jordan's health ministry ran a national COVID-19 campaign through its primary care network and mass vaccination sites and reported progress to WHO.
Preconditions
A routine immunisation programme and cold chain
A supply of doses large enough for a national campaign, however obtained
Registers good enough to count who received a second dose
Limits: The figures are Jordan's own reports to WHO (DATA_SOURCE 'REPORTING'), with no independent coverage survey behind them. Completion of the primary series reached 45 per 100 and boosters 682,211 people (7 per 100), so the record shows a partial delivery. The row was last updated on 21 August 2022, earlier than most, so later doses are missing; first-dose reach stayed below half of WHO's population estimate. The VACCINES_USED field is empty, so the record cannot show which products were used or where the doses came from, and it evidences distribution only. The per-100 rates divide by WHO's population estimate. Vaccination is a response to one shock, not proof of general preparedness, and it is filed under the precedent of bol-covid-vaccination and ven-covid-vaccination as a judgement call because no hazard-specific delivery with an official figure was found for this country in this batch. The source is a CSV with no prose; the numbers are read from one row. The cited URL is the blob the WHO dashboard serves, the same one bol-covid-vaccination cites.
๐ฑ๐งLebanonsince 2021delivered and closedbears on Disaster preparedness and recovery
Lebanon's health authorities reported to WHO that 2,740,227 people, 40 per 100 population, had received at least one COVID-19 vaccine dose by 22 December 2022, from a first dose given on 14 February 2021, with 2,414,223 (35 per 100) completing the primary series.
Lebanon's health ministry ran a national COVID-19 campaign through its primary care network and mass vaccination sites and reported progress to WHO.
Preconditions
A routine immunisation programme and cold chain
A supply of doses large enough for a national campaign, however obtained
Registers good enough to count who received a second dose
Limits: The figures are Lebanon's own reports to WHO (DATA_SOURCE 'REPORTING'), with no independent coverage survey behind them. Completion of the primary series reached 35 per 100 and boosters 660,249 people (10 per 100), so the record shows a partial delivery. The campaign ran through Lebanon's 2019 to 2022 financial collapse and the August 2020 port explosion, and first-dose reach stopped at 40 per 100, which is the more informative reading: a partial delivery by a state in crisis. Donor financing of the doses, if any, is not shown in the CSV. The VACCINES_USED field is empty, so the record cannot show which products were used or where the doses came from, and it evidences distribution only. The per-100 rates divide by WHO's population estimate. Vaccination is a response to one shock, not proof of general preparedness, and it is filed under the precedent of bol-covid-vaccination and ven-covid-vaccination as a judgement call because no hazard-specific delivery with an official figure was found for this country in this batch. The source is a CSV with no prose; the numbers are read from one row. The cited URL is the blob the WHO dashboard serves, the same one bol-covid-vaccination cites.
๐ณ๐ฟNew Zealandsince 2020delivered and closedbears on Institutional responsiveness
Cabinet agreed a 12-week wage subsidy on 16 March 2020, the Government announced it on 17 March, and the Ministry of Social Development paid employers on average within three and a half days of an application, paying more than NZ$13 billion across three stages by the end of December 2020.
Cabinet approved a flat per-employee subsidy paid as a lump sum to employers, and the Ministry of Social Development ran it through its existing benefit payment systems on applicant self-declaration, with Inland Revenue data used for checks after payment.
Preconditions
A payment agency with bank details and identity data for most employers and workers already on file
A fiscal position that allowed NZ$5.1 billion to be committed within a day and raised to NZ$9.3 billion within a week
Political acceptance of overpayment risk, recovered afterwards by audit and voluntary repayment
Where else it has been used: The same flat-rate, employer-paid design was used in Australia's JobKeeper (March 2020); New Zealand reused its own scheme for the August 2020 and 2021 resurgences.
Limits: The live OAG pages answer scripted requests with HTTP 403; the text was read on 2026-10-02 from the Internet Archive copy (https://web.archive.org/web/2026/https://oag.parliament.nz/2021/wage-subsidy/part3.htm), with the dates in Part 1 (paragraphs 1.12 to 1.13) and the 3.5 days in paragraph 3.3. The average is the Ministry of Social Development's figure as reported by the auditor, with no distribution: it does not say how long the slowest payments took. Speed was bought with a high-trust design in which applicants declared their own eligibility; by 5 March 2021 the Ministry had received 20,973 repayments worth NZ$726.2 million, about 5 percent of payments, and nearly NZ$300 million went to public organisations before decisions were made to return money. The NZ$13 billion is a spending figure, not a measure of jobs saved: the Government's estimate that the scheme indirectly supported about 1.8 million employees is not an outcome measure, and the auditor notes it could not determine with certainty how many recipients could have used other support. The audit covers the scheme to early 2021 and not the later 2021 stages.
๐ฏ๐ฒJamaicasince 2020delivered and closedbears on Institutional responsiveness
From March 2020 the Government of Jamaica built the CARE programme on its PATH cash transfer and tax registration systems, paying cash transfers to 458,295 beneficiaries (183,326 men and 274,969 women) for job and income losses and a COVID-19 PATH grant to about 266,100 PATH beneficiaries within the first year of the pandemic.
The Ministry of Finance created a family of temporary grants (SET Cash for those who lost jobs, general grants for registered self-employed workers, compassionate grants for others) and paid them by matching applications against tax registration numbers, while the Ministry of Labour and Social Security paid an extra PATH grant through the existing conditional cash transfer.
Preconditions
A tax registration number held by most adults, which let the government verify applicants it had no benefit record for
An existing conditional cash transfer (PATH) with a beneficiary roll and payment channel
Fiscal buffers built during the 2013 to 2019 consolidation, which paid for the response before external finance arrived
Limits: Both totals are published in the ICR's results framework (Annex 1, page 23: 458,295 and 266,100) and equal the sums of the male and female counts in Table 3 (page 18); the text on page 12 says 'more than 450,000 beneficiaries', 6.6 percent above target. The CARE measures were prior actions of a World Bank development policy loan approved in 2021, so the lender wrote the targets and rates its own operation (Moderately Satisfactory overall). The construct is speed, and the ICR gives no elapsed time from shock to first payment: it dates the measures to Cabinet decisions from March 2020 and says the PATH grant reached households 'within the first year of the pandemic', with beneficiaries determined in mid-2020. Eligibility for SET Cash covered job losses from 10 March 2020, so the rule was written to the shock, but the record cannot say whether money arrived in weeks or months. The ICR notes the government did not collect data on all indicators (for example, tourism jobs retained under BEST Cash). Beneficiary counts are not household counts and may overlap across grants.
๐ง๐ฉBangladeshsince 2020delivered and closedbears on Institutional responsiveness
Bangladesh's Department of Social Services paid old age, widow and disability allowance entitlements early from 25 March 2020 to 30 June 2020, reaching about 4.36 million beneficiaries through its digital government-to-person payment system, and the World Bank reimbursed US$165.84 million within 16 days of the government's July 2020 request.
The social services department advanced payment of existing allowances through the government-to-person system that pays straight into beneficiaries' bank and mobile money accounts, and a contingent emergency component in an existing World Bank loan refinanced the outlay after the fact.
Preconditions
A government-to-person payment platform already paying allowances into individual accounts
Mobile money accounts held by most beneficiaries
A contingent emergency clause in an active loan, so that refinancing did not need a new operation
Limits: The figures are in a PDF: the 4.36 million in paragraph 44 (page 10) and indicator I.11, and the dates in Annex 6 (page 31). The World Bank financed the project and the ICR praises its own instrument: the 16 days measure how fast the Bank reimbursed spending already made, not how fast the government acted, since the financing was retroactive for transfers paid from 25 March to 30 June 2020. The government's own change was modest: it paid existing beneficiaries of existing allowances early and through mobile money, and the source reports no new beneficiaries; other COVID-19 transfers paid outside these allowances are not covered by this record. The ICR also records the limits of shock response: Component 5, a cash transfer for people injured in the 2024 uprising, had policies and systems in place but was never paid, because the government's prefinancing did not arrive, and the allocation was cancelled. The poverty database meant to target new shocks (the National Household Database) never became usable.
๐ฒ๐ฒMyanmarsince 2020dismantledbears on Institutional responsiveness
Under its COVID-19 Economic Relief Plan, Myanmar's government distributed food to about 5.4 million households in April 2020 and cash transfers that had reached about 6 million vulnerable households by November 2020, and the World Bank reported in July 2021 that these transfers had come to a sudden halt after the military takeover of 1 February 2021.
The civilian government widened its maternal and child cash transfer and social pensions and set up emergency food and cash distributions through ward and village administrators, using community identification of the neediest households because no national poverty registry existed.
Preconditions
A functioning civilian administration down to ward and village tract level
Donor and budget finance that could be reallocated within weeks
Political continuity, which is the precondition the 2021 takeover removed
Limits: The 6 million and the halt are in one sentence on printed page 38 of the July 2021 Monitor ('reached about 6 million vulnerable households by November 2020 have come to a sudden halt'). The source does not say who stopped the transfers, by what decision, or whether every government transfer stopped: the line on printed page 43 that 'all government cash transfer programs and remittances stopped' is an assumption of the Bank's poverty simulation (method in Annex 1, page 51), not an observation, so the record does not claim that the maternal and child cash transfer or the social pensions ended. The food distribution to about 5.4 million households in April 2020 and repeated cash payments of 40,000 kyat to 5.4 million households in July and August 2020, rising to 5.6 million in September and 6.1 million in November, are the government's figures as relayed by the December 2020 Monitor (Coping with COVID-19, https://documents.worldbank.org/curated/en/906171608086222905/pdf/Myanmar-Economic-Monitor-Coping-with-COVID-19.pdf, printed pages 52 and 66), which notes its own survey found lower numbers. The figures are approximate and dated by month only, so the speed of the 2020 response is shown only by month. The COVID-19 transfers were emergency rounds of modest size (the December 2020 Monitor calls the support far from meeting needs), so part of what stopped was temporary by design; the reversal is that they stopped suddenly during a deeper crisis rather than being replaced. The halt followed a coup, not a policy judgement on the programme, and the source does not separate loss of finance, loss of donor support and loss of administrative capacity. A World Bank country monitor is an international source reporting on the government's programme, not the government's own accounts.
๐ฒ๐ณMongoliasince 2020delivered and closedbears on Institutional responsiveness
Mongolia's parliament passed the Law on Exemptions for Social Insurance Contributions and Assistance from the Unemployment Insurance Fund on 9 April 2020, waiving contributions retroactively from 1 April to 30 September 2020 for voluntary contributors and affected employers and workers, and 152,475 workers had received relief under the World Bank-financed part of the measure by 31 December 2020.
Parliament legislated a time-limited waiver of social insurance contributions for employers and workers hit by the pandemic, administered by the social insurance administration through its contribution records, with a development loan refinancing part of the foregone revenue.
Preconditions
A contribution-based social insurance system with employer and worker records
A parliament able to pass emergency legislation within weeks
External finance to backfill the insurance fund
Limits: The 152,475 (71,894 women) is an indicator in an implementation status report, not a completion report, and it covers only the part of the relief the World Bank refinanced (US$10 million); the Project Appraisal Document of June 2020 (https://documents.worldbank.org/curated/en/815051592878000582/pdf/Mongolia-Emergency-Relief-and-Employment-Support-Project.pdf, paragraph 22) gives the law's date and expected the whole measure to reach about 120,000 voluntary contributors and over 400,000 workers and 38,000 employers at about US$228 million, so the national total is larger and is not in this source. The 8 days is our arithmetic from those two dates and flatters the speed: the law was retroactive, so it says nothing about when employers actually stopped paying. The relief was part of a March 2020 stimulus package; waived contributions weaken the social insurance fund, a cost the indicator does not show. The ISR is the lender reporting on its own project.
๐ฑ๐ฐSri Lankasince 2022delivered and closedbears on Institutional responsiveness
After Sri Lanka suspended external debt service on 12 April 2022, the government topped up Samurdhi and categorical cash transfers to Rs 5,000 to 7,500 per family per month from May to December 2022 for 2.5 million existing recipients and 0.8 million waitlisted recipients, raising spending on these programmes to Rs 142 billion in 2022.
The government used the existing Samurdhi and categorical allowance rolls, plus the waiting list of households already screened but not enrolled, to add a flat monthly top-up within weeks of the debt standstill, with multilateral emergency loans paying for it.
Preconditions
Existing transfer rolls and a waiting list of pre-screened households
Emergency multilateral financing available within weeks of a default
Payment channels that kept working during the 2022 fuel shortages
Limits: The recipient counts and the Rs 142 billion are the government's own statement in its Memorandum of Economic and Financial Policies (attached to the Letter of Intent of 6 March 2023), published by the IMF, not an IMF measurement; the 3.3 million is our sum of the 2.5 and 0.8 million, which the staff report also gives as about 3.3 million beneficiaries at end-2022. The eLibrary HTML has no page numbers. The top-ups were financed by emergency assistance from the World Bank and the Asian Development Bank, so the speed reflects donor money as well as the government's decision. The source gives the months, not the date of first payment, so the elapsed time from the April 2022 debt standstill to payment is known only to within a month or two. The staff report itself says the real value of cash transfers eroded in 2022, with poorer households facing higher inflation, and that Samurdhi targeting was poor (38 percent of the poorest quintile covered in 2019, 12 percent of the richest receiving transfers), so the speed of the top-up does not show that the money reached the right households. The 2023 replacement of Samurdhi by the Aswesuma scheme is a separate, slower reform and is not covered here.
๐ณ๐ตNepalsince 2020delivered and closedbears on Institutional responsiveness
Nepal Rastra Bank's monetary policy of 17 July 2020 expanded its subsidised refinance facility to five times the available fund for COVID-19 affected businesses, and by mid-July 2021 Rs 148.75 billion had been approved to 48,890 borrowers, against Rs 7.49 billion of refinance outstanding a year earlier.
The central bank turned an existing subsidised refinance window into a pandemic credit programme by multiplying its size, setting sector rates (1 percent special, 2 percent small business, 3 percent general) and letting commercial banks, development banks, finance companies and microfinance institutions on-lend it.
Preconditions
A standing refinance facility with procedures banks already used
A banking system with liquidity and branch reach to on-lend quickly
A central bank willing to take credit-allocation decisions by sector
Limits: The URL serves a PDF, which NRB labels an unofficial translation; the 7.49 billion baseline is in the Monetary Policy for 2020/21 (https://www.nrb.org.np/ofg/monetary-policy-2020-21-full-text-english-version/, paragraph 48) and the expansion in paragraphs 100 to 103 of the same document. The central bank reports on its own scheme. 'Approved' is not disbursed, and the two figures are different measures (an annual stock against a cumulative approval), so the comparison shows scale, not a like-for-like increase. The rule changed about four months after the March 2020 lockdown, at the regular annual monetary policy rather than in an emergency statement, so the speed is moderate. Uptake was uneven: under the same document the business continuity loan for tourism and small firms to pay wages reached only Rs 956.7 million, and 30 percent of refinance could go through customer evaluation or microfinance lump sums that the source does not break down. Refinance at 1 to 3 percent with borrowers paying 3 to 5 percent went to borrowers banks chose, and the document gives no count of first-time or small borrowers. Other relief in the same section (repayment extensions for Rs 93.63 billion of loans, restructuring of Rs 129.21 billion) is not added here.
After restricting non-essential activity in March 2020, the Government of Trinidad and Tobago paid a Salary Relief Grant of TT$1,500 a month for three months through the National Insurance Board, spending TT$226.3 million on it in fiscal 2020 against a TT$490 million allocation, and the Ministry of Social Development and Family Services paid income support to 46,533 people outside the national insurance scheme.
46,533 persons, persons receiving covid-19 income support (ministry of social development and family services), 2020-09. 226.3 million TT$, salary relief grants paid in fiscal 2020 (october 2019 to september 2020), of a tt$490 million allocation, 2020-09. Ministry of Finance, Trinidad and Tobago, Review of the Economy 2020, retrieved 2026-10-02.
The move, and what it needed
The government split its pandemic income support by registry: the National Insurance Board paid a flat grant to registered workers who were terminated, suspended or lost income, and the social development ministry paid income support and food cards to people outside the insurance scheme.
Preconditions
A national insurance register with employer and contribution records
A social assistance ministry with an existing public assistance and food card system
Fiscal room from energy revenues and the Heritage and Stabilisation Fund
Limits: The figures are in a PDF: the 46,533 is footnote 54 on printed page 85 (Labour Force and Employment), the TT$226.3 million is on printed page 98 (Transfers to Households) and the TT$490 million allocation is in Box 4 that follows. The ministry is reporting on its own government's programme in a pre-budget document, with no audit. The source gives no date of first payment and no count of Salary Relief Grant recipients, so speed is shown only by the fiscal year: money was paid within the six months to September 2020. Less than half of the Salary Relief Grant allocation was spent in fiscal 2020, which may mean fewer eligible claims, slow processing or payments falling into fiscal 2021; the source does not say which. Eligibility for the grant depended on national insurance registration, so informal workers relied on the smaller income support channel. Persons and grants may overlap across programmes.
๐ฒ๐ฒMyanmarsince 2017delivered and closedbears on Disaster preparedness and recovery
Myanmar's Ministry of Construction and rural roads department had used US$37.2 million of a US$200 million IDA credit approved in July 2016 for recovery from the 2015 floods and landslides, rehabilitating 60 km of rural roads, 10 percent of the target, and starting no national road works in Chin State, when disbursements stopped after the military takeover of February 2021, and the project closed on its original December 2021 date with an Unsatisfactory outcome rating.
After the 2015 floods the government took an emergency recovery credit for roads, bridges and community works, but slow procurement, safeguards breaches and then the 2021 coup left most of the money unspent.
Preconditions
Agencies able to design and procure civil works quickly after a disaster
Safeguards compliance on politically sensitive roads
A government the lender could keep financing
Where else it has been used: Same stop date as mmr-nep; the contrast with Nepal's earthquake housing (npl-ehrp), delivered through cash grants rather than procured works, is the useful one.
Limits: The disbursement figures are on the ICR datasheet (pages 1 to 2) and the 60 km rural road figure, 10 percent of target, is in the component review (page 12); all were read from the plain-text version. The shortfall was mostly in place before the coup: by September 2020, four years after approval, US$27.4 million had been disbursed (ISR table, page 3), the national roads in Chin State never started because donor and government-financed linked works broke safeguards, design work for the large roads was unfinished, and in July 2020 the government requested a 23-month extension that was never processed. The World Bank's pause of all disbursements in February 2021 under its policy on de facto governments then fixed the outturn, so part of the Unsatisfactory rating reflects the lender's stop rather than the ministries' work. The ICR had no contact with the authorities after the takeover and no borrower report, so later works financed by other sources are not counted. US$12.4 million of goods under component 3 were procured and distributed with UNOPS support, so that part was not delivered by national procurement. Recovery of one flood event says little about warning or evacuation capacity, which this project did not measure. The 2015 floods themselves, and any relief delivered in 2015, fall outside it. Nothing delivered was lost, so this is a slow delivery and not a reversal.
๐ต๐ฐPakistansince 2006delivered and closedbears on Disaster preparedness and recovery
Pakistan's Earthquake Reconstruction and Rehabilitation Authority inspected all 463,243 houses due for reconstruction in Azad Jammu and Kashmir and Khyber Pakhtunkhwa after the October 2005 earthquake and certified 438,789 of them, 95 percent, as compliant with seismic-resistant standards by the time the World Bank credit closed in May 2011.
A federal authority with army support paid owner-driven housing grants in four tranches, each conditional on an inspection against a seismic-resistant design, and trained masons to build to it.
Preconditions
A damage census done within weeks
Military logistics in mountainous terrain
Bank branches or mobile payment points reaching rural households
Where else it has been used: Nepal (npl-ehrp) copied the tranche and inspection design after 2015; it is Sri Lanka's tsunami ICR (lka-tsunami-housing), not this one, that names Pakistan's programme as having learned from Sri Lanka's homeowner-driven model.
Limits: The table is in Annex 2 (printed page 23) and was read from the plain-text version. The same annex's text gives a different base, 353,931 houses to be reconstructed of which 337,177 were completed, and says 3.8 percent of eligible houses saw no work at all; the record uses the table, and the two figures cannot be reconciled from the document. The data are ERRA's own, compiled from army monitoring teams and partner organisations. Compliance at certification is not occupancy or durability. The programme cost about US$1.5 billion, of which the World Bank credit financed part, and the credit was later enlarged to cover the 2010 floods, which are outside this figure. It is a 2006 to 2011 delivery and says nothing about Pakistan's response to the 2022 floods.
๐ฑ๐ฐSri Lankasince 2005delivered and closedbears on Disaster preparedness and recovery
Sri Lanka's government paid owner-driven cash grants through two state banks, with IDA refinancing, to 47,864 Phase I beneficiaries after the December 2004 tsunami, of whom a 100 percent audit found 44,944 eligible, and the physical completion rate for Phase I houses was 94 percent when the programme closed in September 2008.
The government paid fixed cash grants in tranches through two state banks to households rebuilding on their own land, with divisional secretaries certifying eligibility and IDA refinancing the banks.
Preconditions
State banks with branches in every affected division
A buffer-zone policy settled enough to know who rebuilds where
Local administrations able to certify eligibility and progress
Where else it has been used: The ICR names Pakistan's 2005 earthquake programme (pak-erra-housing) as built on this model.
Limits: Both figures are in paragraph 2.2.6 of the ICR (printed pages 6 to 7), read from the plain-text version. They cover Phase I houses outside the coastal buffer zone only, about 68 percent of the homeowner programme's Phase I, not the whole national reconstruction of more than 110,000 damaged houses, and the donor-driven programme is excluded. About 6 percent of those paid were found ineligible, section 2.4.6 reports that 93 percent of beneficiaries with unfinished houses in the north and east named insufficient funds as the obstacle, and Annex 2 reports further damage to reconstructed houses after hostilities resumed in 2006, especially in Vaharai in Batticaloa District. The completion rate is from the government's own monitoring, reported by the World Bank. The figures describe recovery from one event and not warning or evacuation.
๐ฑ๐ฆLaossince 2010delivered and closedbears on Disaster preparedness and recovery
Laos's Ministry of Public Works and Transport repaired about 1,100 km of roads, including small and wooden bridges, damaged by Typhoon Ketsana in September 2009 across five southern provinces, out of about 1,842 km the post-disaster assessment found damaged, through a recovery component budgeted at US$8.8 million, and in 2011 used the contingency fund the project created to spend US$3.63 million repairing 28.6 km of National Road 13 North after Typhoons Haima and Nock Ten.
The public works ministry folded post-typhoon road repairs into a running road project and created a standing contingency fund that it could draw on within weeks of the next storm.
Preconditions
A damage assessment done soon after the event
A road project already under way with procurement in place
A contingency line the ministry can trigger without a new loan
Limits: The damaged length is in paragraph 3 (page 5), the component in paragraph 11, the contingency fund use in paragraph 33, and the 1,100 km result in the results annex, all read from the plain-text version. The results table dates the achievement to February 2017 against a revised target date of September 2012; paragraph 32 says 55 percent of the damaged roads, against a target of 60 percent, had been rehabilitated by the end of 2011, about two years after the storm, and that the remaining 45 percent was completed by the government with other funds, at a date the ICR does not give. The 1,100 km is a rounded target reported as met, not a measured count. The repair money came from the IDA Crisis Response Window, and the component spent US$10.6 million against US$8.8 million at approval (Annex 3), so the delivery is the ministry's but the financing was external. The World Bank rated the project Highly Satisfactory.
๐ณ๐ฟNew Zealandsince 2011delivered and closedbears on Disaster preparedness and recovery
New Zealand's Earthquake Commission had practically completed 66,252 home repairs under its Canterbury Home Repair Programme by 30 June 2015, after the 2010 and 2011 earthquakes damaged about 180,000 homes, missing the December 2014 completion target it had brought forward from December 2015, with about 8 to 10 percent of repairs needing additional work after completion.
The state natural-disaster insurer took on managed repairs itself, contracted one construction firm to project-manage tens of thousands of jobs, and paid for it from a pre-funded disaster fund and reinsurance.
Preconditions
A pre-funded public catastrophe fund with reinsurance
One contractor able to mobilise a regional workforce
Building inspection capacity to check repairs as they are signed off
Limits: The 66,252 figure, the missed December 2014 target and the 8 to 10 percent rework estimate are in Part 2 of the 2015 follow-up audit; the June 2013 figure is from the Auditor-General's 2013 overview (https://ao.parliament.nz/2013/eqc). The Auditor-General's site at oag.parliament.nz refused scripted requests and the same pages were read at ao.parliament.nz. The Auditor-General says it could not reach an overall conclusion on quality: an MBIE review reported in Part 3 of the same audit found 26 of 74 sampled EQC repairs non-compliant with the structural or durability provisions of the Building Code, and the rework rate is EQC's own estimate. Repair counts are Fletcher Construction's and EQC's figures. The programme covered homes with damage under the EQC cap; about 20,000 more expensive claims went to private insurers and are outside it. Later remedial work on repaired homes, which continued for years and generated litigation, is not quantified here.
๐ฏ๐ฒJamaicasince 2021still operatingbears on Disaster preparedness and recovery
Jamaica's government received confirmation in November 2025 of a full US$150 million payout under the catastrophe insurance it had bought from the World Bank, backed by a catastrophe bond issued in 2024 and renewing cover first placed in 2021, after Hurricane Melissa met the bond's parametric triggers in October 2025.
The finance ministry bought parametric hurricane cover through the World Bank's bond programme, renewed it when it expired, and received a full payout within days of a storm that met the pressure and track triggers.
Preconditions
A finance ministry with a written disaster risk financing strategy
Budget to pay premiums in years without a storm
An intermediary able to place the risk in capital markets
Limits: The payout figure is the World Bank's announcement that the triggers were met; it does not say when the money reached the Treasury. The damage estimate is the Planning Institute of Jamaica's, reported by the Jamaica Information Service on 3 March 2026 (https://jis.gov.jm/total-loss-and-damage-from-hurricane-melissa-estimated-at-1-952-trillion/), and the payout covered about 1.2 percent of it (our division), so the record shows a pre-arranged instrument paying in full, not adequate protection. The bond is issued by the World Bank and priced by capital markets; Jamaica's institutional move is buying and renewing the cover and building the layered risk-financing strategy around it. Other layers (CCRIF, contingency funds, contingent credit) are not quantified here. The press release is the lender's own statement, which praises the government's strategy.
๐ง๐ชBelgiumsince 2006dismantledbears on Disaster preparedness and recovery
The Federal Public Service Public Health held a strategic stock of masks for pandemic influenza in military warehouses in Namur from 2006, sized at 32 million surgical masks when the Bioethics Committee reviewed the pandemic plan in 2009. It destroyed the expired masks in operations in late 2015 and late 2018 without replacing them, so Belgium entered the COVID-19 crisis in early 2020 with no strategic mask stock.
From 2006 the federal health administration stored a national reserve of masks under its pandemic plan, then destroyed expired stock in 2015 and 2018 without a replacement contract, so the reserve ran down to nothing between two pandemics.
Preconditions
A pandemic plan that names a stock size but no owner for rotating and replacing it
Storage borrowed from another department (Defence), with conditions that degraded the stock
Limits: The zero is not a counted figure. It is our reading of the committee's prose statement that no strategic mask stock existed when the crisis began. The 32 million is the stock named in the 2009 bioethics advice and the plan's target; the report never gives an audited count of masks actually held in any year. Some of the stock was probably unusable before it was destroyed: the report records that the Namur warehouses lost their heating in 2012 and the masks stayed in unheated storage. The dates of destruction (late 2015 and late 2018) come from a chronology entry built from hearing testimony, which the committee marks as contradicted by other witnesses, not from a finding of its own. Responsibility for the decision not to replace the stock is disputed in the report itself between the former health minister and the experts heard. A mask stock is one narrow input to pandemic preparedness. It says nothing about the pandemic plan, testing, hospital surge capacity or the later rebuilding of a stock.
๐ง๐ชBelgiumsince 2017still operatingbears on Disaster preparedness and recovery
The National Crisis Centre of the Federal Public Service Interior has run BE-Alert, the national system that lets municipalities and provinces alert the population by SMS, including location-based SMS, and information screens, since 2017. By 1 September 2025, 537 of 565 municipalities, covering 95 percent of Belgian territory, were affiliated, 1,369,659 citizen addresses were registered, and the system had been used 273 times in 2025.
The federal crisis centre built one national alerting platform and offered it to municipalities and provinces, which keep the legal power to warn. It added channels over time (registered SMS, location-based SMS, screens) and runs an annual national test.
Preconditions
Emergency-planning powers held by mayors and governors, so a federal tool needs local uptake to work
Mobile operators obliged or willing to send location-based messages
A federal budget for a platform that municipalities can join voluntarily
Limits: This is the crisis centre's own news release, not a statistic or an audit. Affiliation means a municipality can send alerts, not that it does so well or in time: the release gives no measure of how fast or how accurately any alert reached people in a real emergency. The 273 uses in 2025 mix real alerts, vigilance messages and tests, and the release does not split them. One person can register up to five addresses, so the count is not people. Location-based SMS reaches unregistered phones, but this page gives no reach figure for it. Coverage took eight years to reach 95 percent of territory, and 28 Walloon municipalities were still outside it in 2025. The page says nothing about the July 2021 floods in Wallonia, the largest recent test of national warning, so it cannot show that the system works when a major shock arrives. A warning channel is one part of preparedness and says nothing about recovery.
๐ฏ๐ดJordansince 2016operating below its peakbears on Institutional responsiveness
After the February 2016 London conference, Jordan's Ministry of Labour opened formal work permits to Syrian refugees and waived their fees. Annual permits issued rose from 36,788 in 2016 to a peak of 90,152 in 2023. In July 2024 the waiver was not extended for flexible or employer-sponsored permits, and issuance fell to 54,586 in 2024.
Jordan's Ministry of Labour turned a donor compact into labour-law changes: it waived permit fees, created flexible permits not tied to one employer, and let cooperatives and trade unions act as sponsors. When the donor-funded waiver lapsed in July 2024, it charged Syrians full foreign-worker fees, and issuance fell by about 40 percent.
Preconditions
Concessional financing tied to the refugee response that paid for the waived fees
A labour ministry able to issue and track permits at volume, with intermediaries such as agricultural cooperatives and unions
Domestic tolerance for refugee labour in sectors open to foreigners, which is what came under fiscal and political pressure later
Where else it has been used: Ethiopia's 2019 refugee proclamation and Turkey's 2016 work-permit regulation followed the same compact logic. Both kept permit numbers low relative to their refugee populations.
Limits: The construct is how fast rules and programmes change when conditions change, but a permit count measures uptake. The speed of the 2016 rule change is shown only by the timeline and the 2016 bar. The annual figures are chart labels in a one-page UNHCR factsheet, and the factsheet credits them to the Ministry of Labour. They count permits, not people: renewals, short-term and retroactive permits are all counted, so 340,000 cumulative permits overstates the number of refugees who ever held one. The World Bank and UNHCR copies disagree slightly (45,000 against 47,762 for 2019). The waiver ended in July 2024, so 2024 is only half a post-change year. The 2025 figure (35,702 to July) is partial and is not annualised here. Fall in permits is not fall in work: refugees may have moved back into informal jobs, and the series cannot show that.
๐ญ๐บHungarysince 2013delivered and closedbears on Disaster preparedness and recovery
During the June 2013 Danube flood, which crested at a record 891 cm in Budapest, the government declared an emergency and Hungary's national disaster management directorate ran six days of intensive defence along 807.4 km of levees; of 206,000 people the flood directly endangered, 1,570 had to be evacuated.
The government declared an emergency by river section, and the national disaster management directorate, unified under the interior ministry, took over command of the defence from mayors in 81 settlements and drew soldiers, police, students, public-works employees and volunteers onto the levees.
Preconditions
A national levee system built and maintained over generations by the water directorates
A single disaster management body with legal power to take command from local authorities in an emergency
Upstream gauge readings that gave days of warning before the crest reached Budapest
Limits: The source is the agency that ran the defence, writing about itself six days after the crest, so it is self-reporting with no independent verification. The page gives no count of deaths, injuries or damage cost, and no recovery figure, so it evidences the preparing and defending half of the construct and not the recovering half. A low evacuation count could reflect where the river happened to break out (47,285 hectares, mostly forest and farmland, were flooded) as much as the defence. Much of the protection is a stock of levees rather than a 2013 move, and 36,780 civilian volunteers, 19,515 civil guards and 8,306 soldiers took part, so the outcome is not the directorate's alone. This is one event; it says nothing about the capacity against other hazards such as the 2010 Ajka red sludge spill. The 206,000 is a rounded figure.
Brazil's central bank specified, built and ran a mandatory instant payment rail that settled 7.98 billion transactions in July 2026, for 152 million people and 14 million firms transacting in that month.
The central bank wrote the standard, mandated participation for every institution above a size threshold, ran the settlement infrastructure itself and set the price to zero for individuals.
Preconditions
A central bank with regulatory authority over the institutions it needs to compel
Near-universal account ownership to build on
A national identity number to key accounts to
Where else it has been used: The instant rail is not novel and India built one first. What Brazil added was compulsory participation and a zero price, and those are the parts most copies drop.
Limits: Transaction volume shows adoption of one payment rail. It says nothing about cost or schedule performance on a capital project, which is what this indicator asks for. Pix was specified and enforced by a single institution with regulatory power over the participants, so it is weak evidence about delivery that needs several agencies to hold position.
๐ง๐ทBrazilsince 2019still operatingbears on Large project delivery
Brazil consolidated federal public services behind one identity platform, reporting 175 million active accounts and 5,179 digital services in May 2026.
One identity provider for every federal service, with assurance levels tied to records the state already held, and a rule that services migrate to it rather than build their own login.
Preconditions
Existing federal databases good enough to verify identity against
Authority to force agencies off their own systems
Limits: An account count measures registration and a service count measures listing. Neither shows whether a service completes, how long it takes, or what it costs to run. Both numbers are published by the programme about itself, with no external audit attached.
๐ช๐ชEstoniasince 2001still operatingbears on Large project delivery
Estonia has run one data exchange layer for the whole state since 2001, and in July 2026 it carried 374.9 million queries between 1,000 connected organizations, 251 of them government bodies, across 3,785 services.
A distributed exchange layer between agency systems, with a legal once-only rule: an agency may not ask a citizen for data another agency already holds.
Preconditions
A legal mandate strong enough to bind every agency
A national identity system
Small enough administration to move together
Where else it has been used: Finland, Iceland and others run the same software. The once-only law is the part that does the work and the part that is hardest to pass.
Limits: Query volume measures use of a layer that grew over 25 years. That is evidence of sustained operation and not of bringing one large project in on cost and on schedule, which is what this indicator asks for. The factsheet is published by the authority that operates the system.
๐ฎ๐ณIndiasince 2014still operatingbears on Large project delivery
Existing banks were given a national target and a stripped-down account product, and the identity system removed the documentation barrier that had excluded people.
Preconditions
A bank branch and agent network with national reach
A digital identity covering the population
Political priority high enough to hold banks to a target
Limits: An account count measures opening. Part of the stock is dormant or holds very little, and the release carries no external audit. Reaching 590 million people through banks that already existed is a different problem from delivering a physical project on a schedule.
๐บ๐พUruguaysince 2010still operatingbears on Large project delivery
Uruguay rebuilt its electricity supply between 2010 and 2017 through public tenders and long-term contracts, and in 2025 fossil fuels produced 293 GWh of the 15,855 GWh generated, so 98.2 percent came from hydro, wind, biomass and solar.
Long-term power purchase agreements at fixed prices, awarded by competitive auction, with the state utility as a single creditworthy buyer carrying the offtake risk.
Preconditions
A solvent state utility that private investors will contract with
A wind or solar resource worth bidding for
Cross-party agreement to hold the policy across governments
Where else it has been used: The auction plus single-buyer design is the most transferable piece and has been used across Latin America. The decade of political continuity behind it is not a design choice.
Limits: The share is computed from the published generation series by source and is not itself a published headline figure. A generation mix is an outcome, and it hides the cost and schedule record of the projects behind it. A wet or dry year moves the hydro share by several points, so one year is not the programme.
๐บ๐พUruguaysince 2007still operatingbears on Large project delivery
Uruguay gave a connected laptop to every child in public primary school from 2007, the first country to do it nationally, and the household survey found a Ceibal computer in the homes of 35.8 percent of the population in 2018.
A dedicated agency outside the education ministry procured devices and connectivity at national scale and used schools as the distribution channel.
Preconditions
A small enough country to reach every school
Budget held outside the ministry it serves
Electricity and connectivity to the schools
Limits: Household reach peaked at 40.1 percent in 2014 and the series stops in 2018, so this number says nothing about the programme now. A device in the home measures distribution and says nothing about learning. The record is here because the delivery is documented and the measure of it is weak.
๐ง๐ทBrazilsince 1988still operatingbears on Large project delivery
Brazil wrote a right to health into the 1988 constitution and built a single public system that covers everybody in the country, and the WHO service coverage index for Brazil rose from 71 in 2000 to 84 in 2023.
A right to health written into the constitution, funded by earmarked federal transfers and delivered by municipalities under national rules.
Preconditions
A constitutional moment able to carry the commitment
A municipal layer capable of delivering services
Earmarked funding that survives annual budget fights
Limits: A service coverage index counts whether services reach people. It says nothing about waiting times, quality, or the difference between a capital city and the interior, and it credits private provision alongside the public system. Brazil spends less per head on public health than most countries with a comparable index, which this number cannot show.
๐ง๐ทBrazilsince 1973operating below its peakbears on Large project delivery
Brazil's national immunisation programme reached 99 percent DPT coverage in 2003 and held above 95 percent for a decade, then fell to 68 percent by 2021 and had recovered to 91 percent by 2024.
91 % of children, dpt immunisation coverage, children aged 12-23 months, 2024. 99 % of children, dpt immunisation coverage at its 2003 peak, 2003. World Bank, from WHO and UNICEF estimates, retrieved 2026-08-26.
The move, and what it needed
Free universal vaccination delivered through the primary care network, with national campaign days that made coverage a visible public event.
Preconditions
A primary care network with national reach
A cold chain
Public trust in the health system, which is what later eroded
Limits: This record is here because it runs both ways. A delivery capability that reached the whole country lost a third of its coverage in five years and has not fully recovered, which is evidence about durability rather than about peak performance. Coverage is administratively reported, DPT is one antigen among many, and the fall has several causes that the series cannot separate.
๐ง๐ทBrazilsince 2003delivered and closedbears on Large project delivery
Brazil connected the last rural households to electricity through a national programme run with the state and private distributors, taking access from 94.4 percent of the population in 2000 to 99.8 percent in 2024.
The federal government paid the connection cost and delivered through the existing distribution concessionaires, with connection targets written into their contracts.
Preconditions
Concessionaires already operating in the territory
A grid close enough to extend
Subsidy large enough that the last connections are not loss-making
Limits: Access counts a connection. It says nothing about how often the power fails, what it costs, or the isolated Amazon communities that make up most of the remaining gap. Access was already rising before 2003, so the programme accelerated a trend rather than starting one.
๐ง๐ทBrazilsince 2003still operatingbears on Large project delivery
Brazil built one national registry of low-income households and paid conditional transfers through it, going from 6.6 million benefits in December 2004 to a peak of 21.0 million in 2022 and 18.6 million in December 2025.
One registry of low-income households shared across every social programme, with payments made directly to a card held by the mother, conditional on school attendance and health checks.
Preconditions
A payments network reaching poor municipalities
Civil registration good enough to identify households
Municipal staff to maintain the registry
Where else it has been used: The conditional transfer plus single registry has been adopted across Latin America, Africa and Asia. The registry, not the transfer, is the reusable asset.
Limits: A benefit count is administrative and says nothing about whether the transfers reduced poverty. The programme was renamed and restructured twice between 2021 and 2023, so the series covers more than one design. The capability worth copying is the registry and the targeting, and neither has a number in this record.
๐ง๐ทBrazilsince 1996still operatingbears on Large project delivery
Brazil has run every election on electronic voting machines since 2000, across an electorate that reached 155.9 million voters in 2024, with results published the same evening.
One election authority owns the machines, the software and the count, with public audit ceremonies and results published the same evening.
Preconditions
A single national election authority with its own budget
Enough institutional standing to be trusted with the count
Logistics able to reach every polling place
Limits: The electorate measures the scale of the operation and not the speed or the integrity of the count. The Superior Electoral Court publishes result timing and audit results separately and neither is in this number. Electronic voting has been contested politically in Brazil despite repeated public audits, which is a fact about trust rather than about delivery.
๐ง๐ทBrazilsince 1997still operatingbears on Large project delivery
Brazil developed deepwater extraction, found the pre-salt fields in 2006 and produced them at depths of more than 5,000 metres, with national oil production rising from 71,844 thousand cubic metres in 2000 to 219,032 in 2025.
A state oil company sustained deepwater research for decades, and licensing rules concentrated the learning from each field in one operator.
Preconditions
A state company with a real research budget
Reserves worth the exploration risk
Capital markets willing to fund very long payback
Limits: Production is an outcome shared between Petrobras and foreign operators, and several individual platforms ran years late and over budget, which is exactly the cost and schedule record this indicator asks for and which this number hides. Expanding oil output also sits against Brazil's climate commitments, and a production figure cannot weigh that.
๐ฒ๐ฝMexicosince 2003dismantledbears on Large project delivery
Mexico built public health insurance for people outside social security from 2003 and abolished it in 2020, and CONEVAL measured the share of the population without access to health services rising from 16.2 percent in 2018 to 39.1 percent in 2022.
The abolition is the mechanism recorded here: a new government replaced a funded, enrolment-based entitlement with a budget-line service that required no affiliation, and affiliation-based coverage disappeared faster than the replacement could deliver care.
Preconditions
A programme identified with the previous political coalition, which made it removable
Centralised budget authority able to redirect earmarked health funds
No statutory or constitutional floor under the entitlement
Where else it has been used: The replacement, INSABI, was itself dissolved in 2023 and replaced by IMSS-Bienestar, the second replacement in four years. The case travels as a warning: an entitlement without a legal floor is one election from removal.
Limits: The metric measures the dismantling, not the programme: CONEVAL's deprivation number is self-reported affiliation, not care received, and Seguro Popular at its peak was itself criticised for shallow coverage and uneven state-level spending. The 2018-2022 change spans the pandemic, which stressed every health system, so not all of the 23-point rise is the reform. CONEVAL itself lost autonomy in a later reorganisation, so the continuity of this series is not assured.
England built a national network of early-childhood centres reaching 3,632 sites by August 2009, then closed or hollowed out a large part of it after funding was cut, with the official count at 3,123 by October 2017 and independent estimates near 1,000 closures.
3,123 centres, children's centres open, official count, 2017-10. 3,632 centres, children's centres open at the peak, 2009-08. The Sutton Trust, Stop Start report summary, retrieved 2026-08-27.
The move, and what it needed
Central government funded local one-stop centres for under-fives, ring-fenced the money, and targeted the most deprived areas first; when the ring-fence was removed in 2010, local authorities under cuts converted, merged or closed centres faster than any national decision was ever taken.
Preconditions
Local authorities able to deliver a nationally specified service
A ring-fenced grant, whose removal was also what let the network erode
Cross-departmental agreement that early childhood was infrastructure
Where else it has been used: The 2025 Best Start family hubs programme is the same state rebuilding a version of what it closed, fifteen years later, which is the strongest available evidence that the erosion was a loss.
Limits: A site count is the crudest measure of the service: many surviving centres kept the building and lost the staff, which is why the Sutton Trust's estimate of real closures is roughly double the official fall shown here. The count stops in 2017 and later parliamentary answers show further decline. The record shows the erosion of delivered infrastructure under budget cuts, not the effect on children; the IFS separately measured lasting health benefits from the programme at its peak, which is what makes the erosion a loss rather than housekeeping.
๐ฟ๐ฆSouth Africasince 1994operating below its peakbears on Large project delivery
South Africa ran one of the world's cheapest reliable grids and extended it to most of the population after 1994, then under-invested and lost the surplus, reaching 335 days of load shedding in 2023 before recovering to 83 days in 2024.
The erosion is the mechanism recorded here: a state monopoly was told to electrify without being allowed to charge for or build ahead of demand, new-build decisions were delayed for a decade, and the two giant plants finally ordered were too large for the delivery capability that remained.
Preconditions
A single vertically integrated utility, so one balance sheet carried every deferral
Tariffs held below cost for political reasons, which financed under-maintenance
Procurement institutions too weakened to deliver a mega-project on schedule
Where else it has been used: The failure pattern, defer, then order two of the largest coal plants ever attempted, is a caution for any state that skips a build cycle and tries to catch up in one leap.
Limits: The day counts are read from the annual chart in the CSIR report rather than a stated table, and a day with any load shedding counts the same as a day of stage six, so the unit understates severity differences. The 2024 recovery also leans on private rooftop solar that households bought in self-defence, which the report itself notes, so the improvement is not all Eskom's. What the record shows is a delivered capability, generation adequacy, lost over roughly fifteen years of deferred maintenance and failed new build, with Medupi and Kusile late and over budget.
๐บ๐ธUnited Statessince 2000operating below its peakbears on Large project delivery
The United States eliminated endemic measles transmission by 2000 through school-entry vaccination requirements, and falling coverage brought 2,777 confirmed cases by August 2026, the most since 1991, against 285 in all of 2024.
The erosion is the mechanism recorded here: vaccination requirements are enforced at school entry by states, so a delivery that took federal purchasing and decades of practice is being undone state by state through widening exemption rules and falling trust, with no single decision anywhere.
Preconditions
Enforcement located at fifty state legislatures rather than one national rule
A generation of parents with no memory of the disease
An information environment that carries vaccine doubt faster than health agencies answer it
Where else it has been used: The same coverage decline is visible across several high-income countries; the American case is the cleanest because elimination was formally certified in 2000, which fixes the baseline the erosion is measured against.
Limits: A case count is the lagging end of the erosion; the leading end is kindergarten vaccination coverage, which CDC reports separately as falling from about 95 percent before the pandemic to under 93 percent, below the outbreak-prevention threshold. The 2026 number is a year in progress and will rise. Elimination status is a WHO-defined property of transmission chains, not of case counts, and had not been formally revoked at retrieval; what the record documents is a maintained public-health achievement being lost while the infrastructure that delivered it still exists.
๐ง๐ทBrazilsince 1952still operatingbears on Large project delivery
Founded in 1952, Brazil's federal development bank provides long-term financing and investment across the national economy, and it disbursed R$169.7 billion in 2025.
The federal development bank converts public capital and long-term finance into patient support for firms and infrastructure projects that commercial lenders may not carry through their formation period.
Preconditions
A public balance sheet able to lend beyond ordinary commercial horizons
A pipeline of firms and projects able to turn finance into productive capacity
A mandate that can survive changes of government
Where else it has been used: Development banks exist across emerging economies, but the Brazilian case shows the value and the risk of keeping one institution large enough to shape an industrial system: patient capital can build capability, while political allocation can weaken it.
Limits: Disbursement volume measures the bank's activity, not whether the projects it finances arrive on time, within budget or with durable public value. The source is the institution reporting on itself, and the number mixes market-rate lending with policy-directed finance.
๐ง๐ทBrazilsince 1694still operatingbears on Large project delivery
Founded in 1694, Brazil's national mint has remained a public institution through colonial, imperial and republican regimes and now produces currency, identity documents and passports after more than 330 years of operation.
A permanent public production institution carries the state's credential infrastructure across political and monetary transitions, preserving practical capability while its outputs change with the regime around it.
Preconditions
A public mandate for products whose authenticity the state must guarantee
Security and production capabilities that can be renewed across technology shifts
Administrative continuity strong enough to outlast constitutional change
Limits: Institutional survival and the ability to produce state credentials show continuity, not the cost, quality or strategic value of every product made by the mint. The institution also changed its technologies and functions over time, so continuity does not mean an unchanged organisation.
๐ธ๐ฌSingaporesince 1960still operatingbears on Large project delivery
Singapore set up the Housing and Development Board in 1960 to address a severe housing shortage, and it built more than 21,000 flats within its first three years.
A public housing authority combined land, planning, finance and construction into one delivery system, allowing a severe national shortage to be treated as a programme rather than as separate private projects.
Preconditions
Public authority over enough land for coordinated development
Finance and construction capacity that can be directed to a common programme
A long-term administrative mandate that survives individual projects
Limits: The early construction count shows delivery at scale, not housing quality, affordability or equal access over the full history of the programme. The account is published by the agency responsible for the programme.
๐บ๐พUruguaysince 1952operating below its peakbears on Large project delivery
Uruguay's state railway suspended passenger service in January 1988 to focus on freight; by 1990 about half of the national network had been declared non-operational, leaving 1,484 kilometres active from a 2,998-kilometre network, before passenger services returned only partially in 1993.
The state railway concentrated scarce operating capacity on freight, suspended passenger service and then allowed half the network to fall outside the active system; the later partial restart could not simply recreate the connections and maintenance routines that had been dropped.
Preconditions
A state railway carrying both public-service and commercial obligations
Budget and management pressure strong enough to prioritise freight over access
Alternative road transport available to absorb passengers when rail service stopped
Where else it has been used: Rail contractions are common where a network is asked to behave like a freight business while serving as public infrastructure. Uruguay's current Central Railway rebuild is a second attempt to restore a network after the operating base was allowed to thin.
Limits: Track kilometres are an infrastructure count, not a measure of trains, reliability, passengers or freight capacity, and the source does not establish that every kilometre declared non-operational was physically removed. Passenger service returned on selected routes in 1993, so this is a contraction and partial recovery rather than a permanently closed railway. The network statement is a government description and does not evaluate the decision's social or economic effects.
๐ซ๐ทFrancesince 2007operating below its peakbears on Large project delivery
France's Flamanville 3 EPR took about 16.5 years from construction beginning in September 2007 to completion in early 2024, and its commissioning was the country's first new nuclear reactor since Civaux 2 in 1999, a 25-year gap.
A state-backed utility attempted a first-of-kind reactor after a long pause in domestic nuclear construction, while the regulator accumulated inspections and prescriptions as construction, manufacturing and testing problems appeared; the project delivered, but the decade-plus gap exposed how much build-and-supply-chain capability had gone unused.
Preconditions
A utility and regulator able to retain nuclear engineering knowledge through a long construction pause
A supply chain capable of qualifying safety-critical components to current standards
Political and financial capacity to carry a first-of-kind project through repeated redesigns
Where else it has been used: The EPR experience is now being used in France, Finland, China and the United Kingdom as a common warning about first-of-kind nuclear delivery. The lesson that travels is to rebuild the industrial base before promising a fleet, not to copy the reactor design alone.
Limits: The durations are calculated from the start and completion dates and from the previous commissioning date stated by the regulator; they do not isolate the causes of delay, the project's cost or the reactor's long-run availability. Flamanville eventually entered service and the regulator reports nearly 600 construction inspections, so the case documents a thin and difficult delivery capability rather than a project that never completed. The regulator describes its own oversight and is not an independent cost auditor.
๐จ๐ฆCanadasince 2016operating below its peakbears on Large project delivery
Canada rolled out the Phoenix federal pay system in 2016, then spent more than C$2.6 billion by April 2022 to support and stabilise it and correct pay-file errors, with a further roughly C$1 billion allocated through the 2024-25 fiscal year.
A central agency replaced a complex government-wide payroll system through a schedule-driven rollout with insufficient independent oversight; once errors spread across departments, the state converted the project into a permanent stabilisation operation while designing a successor.
Preconditions
A payroll system serving many departments with different rules and legacy data
Central oversight strong enough to challenge a schedule and budget that look achievable on paper
A fallback payroll process that protects workers while the digital system is repaired
Where else it has been used: Phoenix is now used inside Canada's digital-government controls as a failure case. The transferable lesson is governance: independent review has to see the whole operating complexity before a government-wide launch, not after the backlog appears.
Limits: Spending on stabilisation and correction measures shows the scale of a delivery failure, not the total social cost of late or incorrect pay, nor whether a replacement will perform better. The Auditor General notes that the total implementation cost was still not known, and the figures combine technical support, error correction and later stabilisation rather than one clean project budget. The system continued to operate while the government explored its replacement.
The Netherlands ended roughly sixty years of Groningen gas extraction with the permanent closure of all production locations on 19 April 2024; by the end of 2024 the response system had handled more than 500,000 damage reports and paid over โฌ2.5 billion in compensation.
The state and a private operator ran a national energy system for decades, then progressively restricted and finally closed the field as evidence of induced earthquakes and resident harm made the old production objective untenable; the institutional challenge shifted from extraction to repair and trust.
Preconditions
A public-private operating structure able to produce and distribute gas nationally
Seismic monitoring and a legal route for government to reduce production and close the field
Administrative capacity to process large-scale damage claims after the revenue stream ended
Where else it has been used: The case travels as a warning about lock-in: a system that reliably delivers an economic objective can still require an equally large exit and repair capability when its external costs accumulate.
Limits: Damage reports and compensation measure the burden of the extraction system and the state's response, not the engineering reliability or economic value of the gas infrastructure. The figures include cases handled by successor agencies and do not show how many homes are safe or how long recovery will take. Closure was a deliberate response to induced earthquakes, so the case is evidence of a capability being withdrawn after its harms became politically unavoidable, not a verdict on every use of natural gas.
The Netherlands has run DigiD as a shared login for public services since 2005; in 2024 it recorded 550,153,271 successful authentications across 16,829,696 accounts and reported 99.94 percent availability against a 99.5 percent norm.
The state provided one identity rail that agencies could reuse instead of each building a separate login, then raised assurance levels through SMS, an app and an identity-document check while publishing service-level figures for the shared platform.
Preconditions
A population register and national identifier that public services can use consistently
A central operator able to set security, availability and integration standards for many agencies
A support and authorisation route for people who need another person to act for them
Where else it has been used: Shared government identity rails have spread across Europe, but the difficult part is not the login screen: it is agreeing one trust model, keeping the service available and providing assisted access when digital identity is not enough.
Limits: Authentication volume, account stock and uptime show a durable shared service, not whether online procedures finish, whether users can access it equally or what it cost to build and operate. Logius reports its own service figures, and account counts include people who may use DigiD rarely or not at all. The case is evidence of sustained digital delivery rather than proof that every public service is easy to use.
๐ณ๐ฑNetherlandssince 1954delivered and closedbears on Large project delivery
After the 1953 flood, the Netherlands built 13 Delta Works between the first project in 1954 and the completion of the Maeslant Barrier in 1997, creating a long-lived national defence system of dams, sluices and storm-surge barriers.
13 major flood-defence works, delta works delivered, 1997. 43 years, programme duration to final delivery, 1997. Rijkswaterstaat, institutional history, retrieved 2026-08-31.
The move, and what it needed
A national authority turned a catastrophic shock into a sequenced portfolio: it set common safety standards, built dams and movable barriers one after another, retained specialist engineering knowledge and kept the programme alive across governments until the final barrier closed the original build cycle.
Preconditions
A widely shared national threat that can sustain investment after the disaster fades from daily politics
A public engineering authority able to carry designs, standards and procurement knowledge across decades
Stable funding and coordination with ports, fisheries, municipalities, businesses and environmental interests
Where else it has been used: The Delta Works became a reference point for coastal-defence programmes elsewhere. The transferable move is a durable portfolio and standard-setting institution; copying a barrier without the maintenance, modelling and governance around it copies only the visible object.
Limits: Counting structures and years shows continuity of a national programme, not the cost, construction productivity, maintenance burden or actual reduction in flood probability. Rijkswaterstaat reports the programme's own history and says the works protect the area behind them, but this record does not independently test performance or distributional effects.
๐ณ๐ฑNetherlandssince 1995operating below its peakbears on Large project delivery
The Netherlands opened the Betuweroute freight railway in 2007, but the Auditor General found that state operating support in 2006โ2013 reached โฌ170 million against a โฌ76 million maximum subsidy and required another โฌ152 million through 2015; the line was not cost-covering and its promised environmental gain was never measured.
The state built a specialised freight corridor on the expectation that private operators and a road-to-rail shift would make it commercially self-sustaining; when those assumptions did not hold, public money filled the operating gap while the promised environmental test remained undone.
Preconditions
A public sponsor able to separate construction approval from a credible long-term operating plan
Freight operators and port users with enough demand to shift traffic onto one corridor
An independent appraisal that tests both revenue assumptions and environmental counterfactuals before opening
Where else it has been used: The case is a warning for infrastructure justified by a future market: a completed asset can still erode national capability if operating incentives, demand and evaluation are left to catch up after the ribbon-cutting.
Limits: The support figures show that the operating model missed its financial promise, not the railway's current freight volume, safety or strategic value. The Court of Audit says the environmental modal-shift benefit was never measured, so this is not evidence that the line failed every objective. The 2006โ2013 support period also excludes construction costs and later operating decisions.
The Netherlands runs MijnOverheid as one portal for government mail and personal data; in 2024 it carried 103,429,342 messages through 11,015,895 accounts with reported availability of 99.99 percent against a 99 percent norm.
103,429,342 messages, messages delivered through mijnoverheid, 2024. 11,015,895 accounts, mijnoverheid accounts, 2024. Logius, 2024 service figures, retrieved 2026-08-31.
The move, and what it needed
The state made one citizen-facing mailbox and data view that agencies could feed through common services, letting residents find notices and records in one place instead of learning a separate portal for every authority.
Preconditions
A national identity rail such as DigiD that can authenticate residents across agencies
Authoritative registers and interoperability standards so agencies can publish consistent messages and data
Paper, phone and assisted-digital routes for residents who cannot or do not want to use the portal
Where else it has been used: Many countries have built citizen portals, but a single front door only works when agencies agree to standards, notifications and support obligations; a website alone does not create a joined-up government.
Limits: Message volume, account stock and uptime show a durable shared service, not whether a notice is understood, whether a public procedure finishes or whether people can use the portal equally. Logius reports its own service figures, and an account can be inactive. The case is evidence of sustained digital delivery rather than proof that every government interaction is simple or inclusive.
The V250 trains for the NetherlandsโBelgium Fyra service entered operation six years late and the service was cancelled after only 40 days, leaving the intended passenger service on the HSL-Zuid unrealised and triggering a parliamentary inquiry.
The state, rail operators and an international supplier committed to a technically ambitious service without a credible fallback, and the timetable moved ahead before the trains had demonstrated reliable operation; once the fleet was withdrawn, the promised network could not be delivered as contracted.
Preconditions
An independent acceptance regime that tests a complex system in real operating conditions before commercial launch
A contract and concession design with a usable fallback when a supplier misses reliability and certification milestones
Clear accountability across the state as sponsor, the operator as concession holder and the cross-border partner
Where else it has been used: The case travels as a warning about megaproject coupling: a finished track, a concession and new rolling stock can each look plausible while the combined system lacks a tested path to service.
Limits: The 40-day service and six-year delay show a failed launch and delivery process, not that the HSL-Zuid railway has no later passenger value or that every actor shared the same responsibility. The metric does not capture the eventual alternative services, the line's wider benefits or the full financial loss. The parliamentary inquiry was convened to examine the interacting procurement, oversight and governance failures rather than a single technical defect.
Translink and the Dutch public-transport operators rolled out OVpay nationwide in 2023, embedding bank-card and mobile check-in across the country; in 2024 the system processed 390 million travel transactions, 15 percent of all transit transactions.
390,000,000 transactions, ovpay travel transactions processed, 2024. 15 % of transit transactions, ovpay share of all travel transactions, 2024. Translink, 2024 annual report, retrieved 2026-08-31.
The move, and what it needed
Translink gave all operators a common back office and reader software while banks and card networks supplied the payment credentials; a pilot was expanded in stages until passengers could use a familiar bank card or phone on every Dutch operator without buying a separate transit card.
Preconditions
One national transaction processor and technical standards that competing operators will share
Banks, card schemes and transport authorities willing to align liability, settlement and fare rules
A phased migration that keeps the legacy card and assisted support available while the new rail proves itself
Where else it has been used: Open-loop transit payments are spreading internationally. The Dutch case shows that the visible tap is the easy part: nationwide acceptance depends on a common back office, operator coordination and a migration path for concessions, refunds and passengers who are not well served by bank cards.
Limits: Transaction volume and share show that a national payment rail is being used, not its construction cost, reliability, fare equity, privacy effects or the quality of the migration for people who still need an OV-chipkaart. Translink reports the figures and says the rollout and functions such as capping and product sales were still being extended in 2024. Adoption therefore measures delivery and network reach, not a completed replacement of every legacy process.
On 1 January 2024 the Netherlands put the Omgevingswet and its Digital Environment System into operation, joining municipalities, provinces, water boards and the national government behind one Omgevingsloket for rules, maps and permit applications.
4 government tiers, government tiers represented in the shared environment portal, 2024. 1 portal, national environment portal, 2024. Rijksoverheid, Omgevingsloket description, retrieved 2026-08-31.
The move, and what it needed
The government consolidated the legal framework and gave four administrative tiers a shared data and application front door, backed by common digital standards and staged chain testing so a resident can check rules and apply without choosing the responsible authority first.
Preconditions
A legal framework that can align national, provincial, municipal and water-board powers without removing local rule-making
Common location, rule and application data standards that every authority can publish and maintain
A transition programme with testing, support and fallback arrangements for authorities moving from separate legacy portals
Where else it has been used: One-stop planning portals are being attempted in many jurisdictions. The Dutch case makes the hard dependency visible: one interface works only when each authority keeps its underlying rules current and agrees to shared standards, governance and testing.
Limits: The four-tier and one-portal counts show a national institutional and digital launch, not whether local rules are complete, applications are faster, the interface is accessible or the legal simplification has delivered better environmental outcomes. The government describes the portal as under continuous improvement and the transition includes significant local implementation work. This is evidence of coordinated delivery at the point of launch, not a completed evaluation of the new system.
๐จ๐ฑChilesince 2007operating below its peakbears on Large project delivery
Chile launched the integrated Transantiago busโmetro system in 2007; an independent 2016โ19 evaluation found the state subsidy per passenger fare rose from CLP 640 per CLP 1,000 paid in 2016 to CLP 790 in 2019 even as passenger production rose 7%.
790 CLP subsidy per CLP 1,000 passenger fare, state subsidy per clp 1,000 passenger fare, 2019. 640 CLP subsidy per CLP 1,000 passenger fare, state subsidy per clp 1,000 passenger fare, 2016. DIPRES, independent evaluation of the National Public Transport Subsidy, retrieved 2026-08-31.
The move, and what it needed
The state integrated bus routes and metro under concessions and used subsidies to stabilise fares, but the evaluation exposed weak outcome baselines and growing fiscal dependence. The case shows why a complex service reform needs a measurable fallback plan as well as a unified operating design.
Preconditions
A unified transport authority able to coordinate routes, metro, fares and concession contracts
Reliable passenger, coverage, cost and service-quality baselines before the reform begins
Contract and contingency mechanisms that can be revised when demand, operations or fiscal exposure diverge from plan
Where else it has been used: Integrated transport reforms are common worldwide. Transantiago travels as a caution that coordination and physical integration do not substitute for outcome measurement, realistic demand assumptions and contracts that preserve room to correct course.
Limits: The subsidy ratio is not the system's total cost, a schedule overrun or a direct measure of service quality. The evaluation reported missing coverage baselines and global outcome indicators, and the transport network continued operating under later reforms and branding. Rising subsidy dependence is a warning about delivery and contract design, not proof that one institutional choice caused every passenger or fiscal problem.
๐จ๐ดColombiasince 2013still operatingbears on Large project delivery
By September 2024, Colombia's 4G road programme comprised 30 concessions at 86.99% aggregate progress, with 14 in operation, 13 under construction, two in pre-construction and one in early termination.
Colombia bundled road building and rehabilitation into a portfolio of public-private concessions, using a national infrastructure agency to standardise contracts, sequence projects and track construction while private concessionaires supplied capital and delivery capacity.
Preconditions
A central agency with authority to structure, procure and renegotiate long-term road concessions
A pipeline of bankable projects and a legal framework for public-private partnerships
Transparent progress reporting and a route to terminate or restructure concessions that cannot meet their obligations
Where else it has been used: Large concession portfolios travel well as a financing and delivery pattern, but the Colombian case shows why a headline completion percentage must sit beside stage, termination, cost and service measures rather than stand in for them.
Limits: ANI's aggregate progress and stage counts are agency-reported and combine projects with different scopes, concession structures and baselines. They do not establish on-time or on-budget delivery, road quality, travel-time gains or equitable regional benefits; one project had already entered early termination. The portfolio is evidence of sustained delivery capacity and correction inside a large programme, not proof that every concession met its original promise.
๐จ๐ดColombiasince 2007operating below its peakbears on Large project delivery
A 2021 Contralorรญa fiscal-responsibility ruling on Colombia's Reficar modernization held COP2.945 trillion in damage after finding USD997 million in additional investments that added no value to the project.
The state-owned refinery modernization expanded through weak cost, contracting and supervision controls; after the project entered service, Colombia's supreme audit institution traced losses, assigned fiscal responsibility and used the finding to make a failed delivery visible rather than treating the asset's operation as proof of success.
Preconditions
A public audit institution with access to contracts, transactions and project records after delivery
Baseline scope, cost and schedule data that allow later investigators to distinguish justified change from value-destroying additions
Legal recovery and accountability mechanisms that can impose responsibility on officials and contractors
Where else it has been used: Reficar travels as a warning for state-led megaprojects: an operating asset can still represent poor delivery if governance cannot control scope, productivity and contract changes, and if the state lacks an independent route to learn from the loss.
Limits: A fiscal-responsibility ruling is an oversight finding about quantified damage, not a complete account of every cost increase, engineering decision or criminal question; the refinery itself continues operating. The two figures come from a later accountability summary and mix a peso liability with a dollar project finding, so they should not be added. The case demonstrates failed governance and correction through audit, not that large state-led industrial projects are impossible.
๐ฏ๐ตJapansince 2019still operatingbears on Large project delivery
Japan's MEXT GIGA School programme reached a one-device-per-student environment in 1,810 of 1,812 local school authorities (99.9%) by the end of fiscal 2022, with the remaining two scheduled to finish during fiscal 2023.
The national ministry set a clear one-device standard and channelled grants, procurement guidance and network support through local school authorities, while vendors supplied hardware and software and schools adapted use to their own classrooms.
Preconditions
Central funding and a simple, measurable service standard that local authorities can procure against
Procurement, connectivity and technical-support capacity in every school authority
Teacher training, maintenance and replacement plans that continue after the initial hardware push
Where else it has been used: Rapid school-device rollouts travel during crises, but Japan's experience shows that physical coverage is only the first gate; sustained capability depends on support, content and renewal budgets after the delivery headline fades.
Limits: The completion count means a device was in a student's hands and a school network was usable; it does not measure learning, teacher workload, device quality, replacement costs or equal use at home. MEXT reports the authority survey and the last two authorities were still finishing after the reference year. Delivery of hardware is therefore a national implementation milestone, not proof of better education.
๐ฏ๐ตJapansince 1994dismantledbears on Large project delivery
Japan decided in December 2016 not to restart the Monju prototype fast-breeder reactor and moved to decommissioning after a 1995 sodium leak, a 2010 restart followed by equipment trouble and 2012 inspection failures; the approved plan targets roughly 30 years of decommissioning.
After repeated technical and inspection failures, Japan stopped trying to restart a strategically important prototype, created a government-led decommissioning team and converted the remaining work into a staged thirty-year plan with external oversight and local coordination.
Preconditions
A regulator and review process able to expose inspection and management failures
Political authority to end a sunk-cost programme despite regional and industrial commitments
A technical decommissioning institution with funding, specialist staff and a safe plan for fuel, sodium and waste
Where else it has been used: Monju travels as a lesson in stopping: adaptive capacity includes an orderly exit from a prestige technology when safety and management evidence no longer support restart, plus a plan that preserves knowledge while removing the hazard.
Limits: The decision to end the programme and its planned duration do not quantify total cost, opportunity cost, safety performance or the amount of equipment actually removed. The government account describes the policy and management failures and the successor decommissioning plan; it is not an independent assessment of the nuclear programme's full social effects.
๐ต๐ชPerusince 2016still operatingbears on Large project delivery
Peru's Invierte.pe system tracks public investment across national, regional and local governments; by 25 July 2024 the three levels had executed S/25.939 billion of a S/67.797 billion investment budget, a reported 38.3% advance.
Peru replaced a project-by-project approval model with a multi-year investment cycle that registers, prioritises and follows projects across government levels, giving the finance ministry a common pipeline and a basis for tracking execution rather than seeing only annual spending requests.
Preconditions
A central investment registry and rules that bind national, regional and local project units
Budget and procurement data that can be reconciled to physical project progress
Political and managerial capacity to re-prioritise stalled projects instead of carrying them forward invisibly
Where else it has been used: Multi-year investment-management systems travel as a response to fragmented project pipelines. Peru's case shows their limit: a dashboard and prioritisation rule improve visibility, but delivery still depends on local preparation, procurement and maintenance capacity.
Limits: Budget execution is an administrative flow, not evidence that projects were selected well, completed on schedule, delivered value or reached underserved regions. The daily report aggregates three government levels and counts expenditure against a budget that can change during the year. A planning-and-monitoring system can make bottlenecks visible without guaranteeing that agencies resolve them.
๐ต๐ชPerusince 2017operating below its peakbears on Large project delivery
A 2023 Contralorรญa control report found the education portfolio of Peru's Reconstrucciรณn con Cambios programme at 36.6% implementation by July 2023 against a 78% milestone expected in the earlier baseline, with the corresponding target date already expired.
Peru created a dedicated reconstruction authority and hired government-to-government technical assistance, then used concurrent audit to expose procurement and schedule risks while projects were still moving. The 36.6% result shows that a new delivery wrapper did not automatically clear local preparation and contracting bottlenecks.
Preconditions
A reconstruction authority with a cross-sector mandate after a major climate disaster
Technical and procurement support that can work across many local implementing agencies
Concurrent oversight with enough access to identify delays early and a management response that can correct them
Where else it has been used: Government-to-government reconstruction models travel as a way to import programme-management capacity, but Peru's audit is the caution: external expertise can improve visibility while unresolved local land, design and procurement constraints still determine the calendar.
Limits: The comparison is for one education portfolio and one control milestone, not the whole reconstruction programme or every project type. The Contralorรญa measures implementation against a management baseline; it does not by itself explain land, procurement or design causes, nor the quality and resilience of completed works. The gap shows a delayed delivery path, not that every intervention failed.
๐จ๐ทCosta Ricasince 1949still operatingbears on Large project delivery
Costa Rica's integrated electricity system generated 89.4% of its electricity from renewable sources in 2024, while ICE reported electrical coverage reaching 98.7% of households despite one of the country's driest years on record.
A state utility planned generation and controlled the transmission backbone, combining public hydro and geothermal investment with regulated contracts for private and cooperative wind, solar and biomass while extending the distribution grid to remote households.
Preconditions
A technically capable utility with authority to plan generation and operate the national grid
Long-horizon finance and resource mapping for projects whose benefits arrive over decades
A regulatory framework that lets private and cooperative generators sell into a coordinated system
Where else it has been used: Renewable-grid programmes travel as portfolios rather than single projects. Costa Rica's experience makes the enabling bundle visible: a transmission spine, credible offtake and a utility that can coordinate public, private and cooperative assets.
Limits: The renewable share and household coverage are outcomes, not a project-by-project cost or schedule record. Hydrology, imports and thermal backup move the annual mix, and the utility report does not independently audit the full lifetime cost, reliability or distributional effects of the generation portfolio.
๐จ๐ทCosta Ricasince 1941still operatingbears on Large project delivery
Costa Rica's Caja Costarricense de Seguro Social reported health-insurance coverage of 92.92% of the national population in 2024, using household-survey and administrative records to include contributors, dependents, pensioners and state-covered households.
Costa Rica expanded a contributory social-insurance institution into a near-universal insurer-provider by covering dependents, pensioners and state-financed households through a common pool and an integrated primary-care network.
Preconditions
A statutory social-insurance body with authority to collect contributions and contract or operate care
A broad payroll base and public subsidy for households outside formal employment
A nationwide clinic and referral network that turns insurance status into a service
Where else it has been used: Universal-insurance models travel only when the financing and delivery institutions travel together. Costa Rica's coverage figure is a reminder that the hard capability is not declaring a right but maintaining the pool, clinics and cross-subsidy that make it real.
Limits: Insurance coverage is not the same as timely access, quality, financial protection or equal outcomes, and the estimate comes from a combination of survey and administrative data. It also does not show the operating cost or schedule of the facilities and workforce that make a legal entitlement usable.
๐ช๐จEcuadorsince 2013still operatingbears on Large project delivery
Quito opened the first line of its metro in December 2023 after a multi-administration build, and the operator reported more than 54 million passenger trips during 2024 with uninterrupted service.
Quito combined four multilateral lenders, a metropolitan delivery company and an external operator, then renegotiated the operating contract after a suspended launch to clarify responsibilities and restart service across 15 stations.
Preconditions
A metropolitan authority able to hold a long project across several mayoral administrations
Multilateral finance and procurement rules that can support a technically complex rail system
A contract-management team able to diagnose interface failures and renegotiate before the asset is abandoned
Where else it has been used: Urban rail travels as a governance problem as much as an engineering one. Quito shows that commissioning, operator interfaces and a credible restart plan can decide whether a nearly finished asset becomes a service.
Limits: Passenger trips and uninterrupted operation show use after opening, not whether construction met its original cost and schedule or whether the line changes access fairly across Quito. The operator's accountability report is self-reported, and a first-year demand figure cannot establish long-term maintenance or financial sustainability.
๐ต๐ฆPanamasince 2007still operatingbears on Large project delivery
After the third set of locks opened in 2016, the Panama Canal Authority handled 9,944 deep-draft transits and 423 million tonnes of cargo in fiscal year 2024, reporting B/.4.986 billion in total revenue despite climate constraints.
Panama put a strategic infrastructure asset under an autonomous public authority with its own revenue, long-term capital programme and technical workforce, then expanded locks and service rules around the needs of global shipping customers.
Preconditions
A legally protected authority able to retain revenue and plan beyond annual budget cycles
Stable access to multilateral and bond finance for a megaproject with a long construction period
Hydrological, engineering and maritime expertise that can coordinate a global customer base with local water and land constraints
Where else it has been used: A self-financing infrastructure authority travels where one asset has national strategic value, but Panama's canal shows the governance bargain: autonomy brings delivery capacity only when public accountability and water constraints remain visible.
Limits: Traffic and revenue show an operating logistics asset, not the expansion's full cost and schedule record, its distributional effects or the counterfactual value of alternative routes. The canal's own results report is not an independent audit of construction performance, and tonnage varies with world trade and vessel mix.
๐ต๐ฆPanamasince 2010still operatingbears on Large project delivery
Panama's public metro network, inaugurated in 2014 and expanded with Line 2, transported 118,351,007 passengers in the twelve-month operating period reported in its 2024 institutional memory.
118,351,007 passengers, passengers transported in the reported year, 2023-11 to 2024-10. 321,519 passengers in one day, peak daily passengers on line 1, 2024-04-30. Metro de Panamรก, Institutional Memory 2024, retrieved 2026-08-31.
The move, and what it needed
Panama built a dedicated metro company, sequenced lines around the capital's commuting corridors and combined central financing, concession contracts and an operating workforce so a new line could be integrated into a growing network rather than run as a one-off project.
Preconditions
A metropolitan transport authority able to reserve corridors and integrate fares and buses
Multi-year financing and procurement capacity for civil works, rolling stock and systems
An operator and maintenance regime that can turn an opened line into a dependable daily service
Where else it has been used: Metro systems travel as network institutions, not just tunnels. Panama's ridership makes the operational layer visible: integration, maintenance and staffing determine whether construction becomes reliable mobility.
Limits: Ridership demonstrates use and operating scale, not whether each line met its original cost or schedule, how much subsidy it requires or whether travel-time gains are shared across the metropolitan area. The operator reports the figures, and a peak day is not a reliability or maintenance measure.
๐ง๐ดBoliviasince 2014still operatingbears on Large project delivery
Bolivia's Mi Telefรฉrico network carried 78,715,401 passenger trips in 2024; its management report compared that result with a 97.7 million annual volume required for budget balance, or roughly 80% of the target while the system remained in service.
78,715,401 trips, passenger trips on the mi telefรฉrico network, 2024. 97,700,000 trips, annual passenger volume required for budget balance, 2024 target. Mi Telefรฉrico, 2024 management report, retrieved 2026-08-31.
The move, and what it needed
Bolivia treated steep urban mobility as a network problem: a state company linked multiple cable-car lines, integrated fares and stations and expanded service incrementally rather than waiting for a single road corridor to solve access across La Paz and El Alto.
Preconditions
A public authority able to coordinate land, stations, safety regulation and municipal transport connections
Capital financing and specialist operations staff for a high-altitude cable system
A fare and subsidy policy that can keep the network usable while it builds ridership and covers maintenance
Where else it has been used: Cable transit travels where topography makes roads expensive, but Mi Telefรฉrico makes the operating bargain visible: network scale can be delivered, yet financial break-even still depends on ridership, fares and continuing public support.
Limits: Passenger trips show use of the network, not whether construction stayed on schedule, whether fares cover operating and renewal costs or whether travel time and access improved for low-income residents. The budget-balance target is a management benchmark, not an independent estimate of the system's full social value or lifecycle cost.
๐ง๐ดBoliviasince 2013operating below its peakbears on Large project delivery
Bolivia's state lithium company doubled lithium-carbonate output to 2,064 tonnes in 2024 after its industrial plant began operating, but the Ministry of Economy reported that technical problems with evaporation ponds and the water-treatment plant kept production below the project's projected capacity of 15,000 tonnes a year.
Bolivia pursued a state-led value chain from brine extraction to industrial lithium carbonate, retaining control of the resource and plant while learning that evaporation chemistry, water treatment and process reliabilityโnot only mine ownershipโset the usable production ceiling.
Preconditions
Long-horizon public finance and technical institutions able to carry a complex chemical process from pilot to industrial scale
Reliable water, evaporation and quality-control systems in a fragile salt-flat environment
Independent engineering feedback and contracting discipline strong enough to surface underperformance before capacity claims harden into policy
Where else it has been used: Resource-processing projects travel poorly when geological ownership is mistaken for industrial capability. Bolivia's shortfall is useful benchmark evidence because it separates strategic ambition from demonstrated, repeatable throughput.
Limits: Output and design capacity establish a large production gap, but they do not by themselves show the plant's full economics, product quality, environmental effects, downstream contracts or whether the technical bottlenecks were later resolved. The ministry's bulletin is an official macroeconomic account, not an independent engineering or audit report.
๐ต๐พParaguaysince 1984still operatingbears on Large project delivery
Itaipรบ supplied Paraguay's grid with 20,383 GWh in 2024, 4.5% more than in 2023, while the binational plant generated 67,089 GWh in total and continued operating after four decades of service.
Paraguay sustained a technically complex binational asset through a permanent operating institution, shared dispatch and long-horizon maintenance rather than treating the dam as finished when construction ended; the national grid could then convert a cross-border project into domestic energy security.
Preconditions
A durable treaty and governance structure able to resolve operating and revenue disputes across two states
Engineering, maintenance and safety institutions that preserve performance over decades
Transmission and utility capacity able to absorb a large, hydrology-dependent supply stream
Where else it has been used: Large infrastructure lasts when the operating institution is designed as carefully as the concrete. Itaipรบ is a useful benchmark case because its delivery story is four decades of binational maintenance, not only its original construction.
Limits: Annual generation and supply show output from an operating asset, not the original cost and schedule, transmission reliability, environmental effects or the distributional consequences of the binational tariff and power-allocation rules. Hydrology strongly affects the yearly number, and the plant's own report is not an independent project audit.
๐ต๐พParaguaysince 2015dismantledbears on Large project delivery
Paraguay's planned Asunciรณn Metrobรบs corridor ended with an incomplete and unviable works package: in April 2024 the Public Prosecutor's Office said the state had paid certificates for works that did not meet the programme's purpose and alleged a G.180.03 billion patrimonial loss.
Paraguay attempted a multi-municipality bus rapid transit corridor through a public works contract, but weak design integration, contract management and escalation left paid works disconnected from a functioning service and triggered an ex-post prosecution rather than an operating system.
Preconditions
A corridor design agreed with the municipalities, operators and affected communities before civil works are certified
Contract-management capability able to test whether each payment creates a usable part of the service
A credible stop-and-correct process that can surface an unviable design before sunk infrastructure and litigation dominate the response
Where else it has been used: Failed transport projects are useful capability evidence when the failure is written as a delivery problem rather than a scandal alone. Metrobรบs shows why integration, certification and the ability to stop matter as much as the initial capital commitment.
Limits: The loss figure is an allegation in an ongoing criminal case, not a final judicial finding or a complete cost-benefit account. The notice establishes that the works were incomplete and did not serve the stated programme, but it does not measure the corridor's intended passenger benefits, the full financing cost or the later value of infrastructure retained by the state.
๐ฆ๐ทArgentinasince 2015dismantledbears on Large project delivery
Argentina's government review found that the ARSAT-3 construction programme, approved in 2015 with a 2019 launch plan, was suspended in 2016โ17; the unbuilt satellite forced two later orbital-protection operations and left the original nationwide broadband objective undelivered.
Argentina stopped an already contracted satellite mission when financing and programme continuity broke, then had to buy temporary orbital-protection services while a successor design was reconsidered; the episode exposes how a strategic project depends on continuity beyond one administration.
Preconditions
A multi-year financing and procurement plan protected from abrupt political and management turnover
Technical programme governance that can distinguish a redesign from an indefinite suspension
An orbital and service-continuity plan that prices the cost of delay before a launch slot is at risk
Where else it has been used: Strategic technology programmes travel with continuity mechanisms, not only engineering talent. ARSAT-3 is a useful reversal because it records the cost of stopping a national capability before the replacement is ready.
Limits: The review documents a suspension and its operational consequences, not the complete engineering, procurement or commercial case for a third satellite. ARSAT later reformulated the mission as SG-1, so the record describes the original ARSAT-3 programme's failure to deliver rather than claiming that Argentina abandoned satellite capability altogether.
The Philippine Department of Public Works and Highways' live transparency portal records 209,419 completed contracts and 997 terminated contracts out of 258,645 contracts from 2016 to 2026, against โฑ6.4 trillion of listed investment.
DPWH exposed a national project portfolio with contract, cost, accomplishment and completion fields, making delivery status inspectable across regions instead of leaving infrastructure performance inside agency spreadsheets and press releases.
Preconditions
A common project and contract identifier shared by national and district engineering offices
Field reporting and audit practices strong enough to keep accomplishment and termination statuses current
A public interface that lets citizens, contractors and oversight bodies compare commitments with delivered works
Where else it has been used: Project transparency travels when it is attached to contract data and corrections, not only a dashboard of completed works. The Philippines provides a useful portfolio-scale example while leaving cost and quality verification open.
Limits: Contract status and listed investment reveal portfolio scale and a public reporting mechanism, not whether projects met their original cost, schedule, quality or maintenance promises. The portal's totals are live and can change as records are uploaded or corrected; they are not an independent value-for-money audit.
The Philippine Department of Energy's 2017 annual report describes the Bataan Nuclear Power Plant as mothballed and records its reassessment as part of a proposed national nuclear programme, after the original plant failed to become an operating power asset.
The Philippines carried a completed but mothballed nuclear asset into later policy reviews instead of treating construction as equivalent to usable capability; the plant became a long-lived governance and maintenance obligation while the country reconsidered nuclear power.
Preconditions
Independent safety and technical institutions able to reassess a politically sensitive sunk asset
A transparent process for comparing rehabilitation with closure and alternative generation
Long-term funding and site stewardship for a facility that does not produce electricity
Where else it has been used: Megaprojects can fail at the conversion from physical completion to operating capability. Bataan is a useful negative case because it keeps construction, commissioning and usable service as separate tests.
Limits: The report establishes the plant's mothballed status and later reassessment, but not the full historical cost, safety case, debt burden or counterfactual electricity supply. The record concerns the original plant as a delivery outcome; later feasibility studies and nuclear-policy discussions are separate decisions, not evidence that the original project operated.
๐ฎ๐ฉIndonesiasince 2015operating below its peakbears on Large project delivery
Indonesia's public-works ministry reported in January 2024 that the Trans-Sumatra toll-road programme had 596 kilometres operating against a 2,749-kilometre planned network, with another 361 kilometres under construction and later phases still in planning.
Indonesia held a long corridor together through a state toll-road regulator, a state-owned lead developer and phased concessions, but the sequence exposed how land acquisition, feeder links and demand assumptions can stretch a national build over decades.
Preconditions
A regulator able to coordinate concessions, land acquisition and provincial interfaces
A credible financing and toll policy for low-demand early sections
Milestones that separate an operating link from a corridor still on a plan
Where else it has been used: Large transport programmes should publish the operating, construction and planned portions separately. Trans-Sumatra is a useful warning against treating an announced corridor as delivered infrastructure.
Limits: The 2,749-kilometre figure is the ministry's planned corridor, not a historical operating peak, so this status records a delivery shortfall rather than physical contraction. Kilometres open say nothing about land-acquisition cost, construction overruns, traffic demand, maintenance or whether the links provide value relative to cheaper roads; the source is a ministry progress account rather than an independent audit.
๐ป๐ณVietnamsince 2024still operatingbears on Large project delivery
Vietnam's Ministry of Transport reported that seven road projects were completed and opened in 2024, bringing the national expressway network to more than 2,021 kilometres.
The transport ministry set a national completion pipeline, combined central investment with local site and land coordination, and reported completed links as an expanding network rather than isolated provincial projects.
Preconditions
A central project pipeline that can sequence links across provinces
Local authorities able to clear land and connect feeder roads
A stable road standard and funding mechanism that keeps partial links usable
Where else it has been used: Network delivery travels when each opening is useful on its own but designed for later connection. The Vietnamese total shows reach; it does not substitute for a cost-and-schedule dataset.
Limits: The network total and projects opened are an annual ministry progress account, not a project-by-project cost or schedule audit. Kilometres aggregate roads begun in different years and do not show construction quality, traffic demand, land-acquisition cost or the distribution of benefits between regions.
๐ป๐ณVietnamsince 2010operating below its peakbears on Large project delivery
Vietnam's State Audit reported in 2019 that the Cat LinhโHa Dong metro's approved investment had risen from VND8,700 billion to VND18,000 billion and that a line originally due in November 2013 was still unfinished after repeated schedule changes.
The first urban-rail project was allowed to proceed with changing designs, slow site handover and a directly appointed foreign EPC contractor, while operating costs were omitted from the initial economic appraisal; late fixes converted a construction delay into a major cost revision.
Preconditions
An independent technical and cost review before design is split into buildable packages
Land and interface responsibilities settled before an EPC contract starts
A transparent appraisal that includes operations, maintenance and safety certification
Where else it has been used: Urban rail is a system contract, not just an elevated structure. Cat LinhโHa Dong is a negative case for copying a financing and contractor model without the institutions that can hold design, land and commissioning together.
Limits: The audit snapshot predates the line's eventual opening, so it records the delivery failure at the decision and construction stage rather than current ridership or service quality. The figures do not provide a full life-cycle cost, safety record or counterfactual transport benefit; the government news report is a summary of the State Audit findings rather than the complete audit file.
๐ฒ๐พMalaysiasince 2015still operatingbears on Large project delivery
Malaysia completed and opened the 57.7-kilometre MRT Putrajaya Line in two phases, with 36 operational stations serving a corridor of about two million people; full operations began on 16 March 2023.
Malaysia created a dedicated state delivery corporation, divided a technically complex urban railway into phases and used interchange stations and park-and-ride facilities to make the new line part of a larger metropolitan network.
Preconditions
A transport authority able to acquire land and coordinate metropolitan utilities
A dedicated project owner with authority over design, procurement and commissioning
A network plan that protects interchanges and feeder access beyond the line itself
Where else it has been used: Urban rail delivery travels through institutional ownership and integration. A line can be physically complete while its usefulness still depends on fares, feeders and interchange quality.
Limits: Length and station counts show the physical delivery, not the line's final cost, construction schedule, reliability, ridership or operating subsidy. The project page is maintained by the state-owned delivery corporation and describes the corridor's intended population and ridership rather than independently audited outcomes.
๐ฒ๐พMalaysiasince 2016dismantledbears on Large project delivery
Malaysia allowed the bilateral Kuala LumpurโSingapore high-speed rail agreement to terminate on 31 December 2020 and paid Singapore SGD102,815,576 in March 2021 for verified costs incurred on the project and for extending its suspension.
Malaysia used a treaty-level bilateral project to coordinate a cross-border railway, but a change of government, proposed scope changes and successive suspensions exhausted the agreement's final deadline; termination then converted planning work into a compensable sunk cost.
Preconditions
A bilateral governance structure with a clear change-control and termination rule
Independent cost and demand reviews before scope changes become political commitments
A way to preserve useful land and design work when a megaproject is stopped
Where else it has been used: Stopping a project can be a rational decision, but the capability test is whether the state exposes the sunk cost and preserves options. The HSR case keeps cancellation visible instead of counting an announced line as delivery.
Limits: The compensation figure records a bilateral settlement, not the full public cost of years of planning, land work or foregone connectivity. No rail construction was delivered; later proposals to revive or redesign a Malaysian line are separate projects, not evidence that this one operated. The 2016 start year (the bilateral agreement) and the earlier suspensions are not stated in the cited compensation statement; MyHSR's FAQ documents the suspension to 31 May 2020.
๐น๐ญThailandsince 1991dismantledbears on Large project delivery
Thailand approved Hopewell's 60-kilometre Bangkok Elevated Road and Train System concession in 1991, but work stalled at 14 percent completion for more than nine months during the 1997 property-market collapse and the government terminated the concession in spring 1998.
The concession bundled transport delivery with property-development revenue and depended on continuous site handover and a rising Bangkok property market. When the boom collapsed, construction stopped, the state and concessionaire lacked a credible re-scoping path, and the government terminated the agreement rather than deliver the promised integrated system.
Preconditions
A concession contract that separates transport obligations from speculative land-value revenue
Independent feasibility, demand and financial-stress tests before a bundled megaproject is signed
A termination and salvage plan that protects the public right-of-way and explains sunk costs
Where else it has been used: Hopewell is now used internationally as a warning against financing core transport with optimistic property capture. The transferable capability is disciplined project preparation and an honest fallback, not a particular rail technology.
Limits: The 14-percent figure is a contemporaneous project-progress estimate in a World Bank case box, not an audited final account of every structure built or every baht spent. Termination records a failed concession and its loss of optionality, not the performance of Bangkok's later rail network. The project combined an elevated road, MRT, railway grade separation and property-development rights, so the failure cannot be assigned to one technical component.
๐ช๐ธSpainsince 1992still operatingbears on Large project delivery
Spain expanded high-speed rail from the first MadridโSeville line in 1992 to more than 4,000 kilometres of high-speed track by 2024; the Transport Ministry reported that the service reached 70 percent of the population.
Spain treated high-speed rail as a long-horizon territorial network, retained a national infrastructure manager and sequenced corridors over multiple administrations. European funding, domestic engineering capacity and a common technical standard let each new line add value to a growing national system.
Preconditions
A stable national investment pipeline that survives electoral cycles
An infrastructure manager able to plan track, signalling, stations and maintenance as one network
Demand, finance and environmental appraisal strong enough to choose where extension is justified
Where else it has been used: Spain's network is a reference case for countries weighing high-speed rail, but the transferable capability is continuity of planning and network integration, not the kilometre target alone.
Limits: Track length and population reach show a sustained national build-out, not whether each corridor earns its operating cost, how much construction exceeded initial budgets or whether travel shifted from cars and planes. The ministry's statement uses a rounded lower bound for the network and a population denominator rather than land-area coverage. Lines and services opened over three decades under different governments, so the record is not a single-project causal test.
๐ช๐ธSpainsince 2008dismantledbears on Large project delivery
Spain extinguished the Castor underground gas-storage concession in October 2014 after seismic events halted operation; the state transferred administration to Enagรกs and the decree recognised โฌ1.350729 billion in compensation for the former concessionaire, with recovery structured over 30 years.
The state granted a private concession for strategic gas storage, then halted injection when induced seismicity made operation unacceptable and transferred administration to the system operator. The exit preserved the physical asset and safety responsibility but socialised a large compensation claim, showing how a rapid stop can still leave a long fiscal tail.
Preconditions
Independent seismic monitoring and a regulator able to suspend operation on safety grounds
A concession contract that defines compensation, decommissioning and liability before construction
A system operator capable of securing and maintaining a closed facility while its final fate is decided
Where else it has been used: Castor travels as a caution about subsurface megaprojects and asymmetric risk: stopping unsafe operation is necessary, but the public needs an honest liability and decommissioning plan before approving the investment.
Limits: The compensation and recovery period show the fiscal and legal consequences of closing the facility, not the full construction cost, the number of earthquakes caused or the eventual cost of sealing the wells. The concession was extinguished and the facility placed in hibernation, with later court decisions changing parts of the payment mechanism. This is evidence of a project that could not be operated safely, not a claim that every underground storage project has the same risk.
๐ต๐ฑPolandsince 2018dismantledbears on Large project delivery
Poland began the 1,000-MW Ostroลฤka C coal-block project in 2018 without full financing, suspended construction in 2020 and demolished the built elements in 2021; the Supreme Audit Office later found that the unfinished investment consumed more than PLN1.5 billion.
Poland committed to a large coal unit before its financing, fuel assumptions and climate-policy fit were fully re-tested. When the economics and policy context changed, the state-owned investors suspended the build and converted the site toward gas, but only after construction and contractual obligations had made the exit costly.
Preconditions
Independent feasibility and financing checks refreshed when a project is materially delayed
Board and shareholder oversight able to stop a state-owned project before sunk costs compound
A site-reuse and worker-transition plan that preserves options after cancellation
Where else it has been used: Ostroลฤka travels as a warning about energy megaproject lock-in: changing course is rational when assumptions fail, but review gates must come before construction and contract settlement, not after.
Limits: The audit's cost figure covers the investment and its close-out as assessed by NIK, not every wider system cost or the value of the later gas-fired replacement. The project was a state-owned-company investment rather than a conventional central-government budget line, and the audit is a retrospective accountability finding. Abandonment can prevent further losses, but it does not recover the sunk expenditure or prove that every original energy-security objective was mistaken.
๐ธ๐ชSwedensince 2010still operatingbears on Large project delivery
Stockholm delivered Nya Karolinska Solna through a 30-year publicโprivate partnership: the official inquiry reports SEK 22.8 billion for construction and equipment, a projected SEK 57 billion total through 2040 including operation and maintenance, and an external comparison that found construction 53% more expensive per square metre than Scandinavian hospitals.
Stockholm bundled design, construction, financing, operations and maintenance into a long PPP contract, transferring construction risk and locking in a lifecycle payment model. The inquiry shows the trade-off: a complex hospital was delivered and remains in use, but the long contract and specification changes made the full cost harder to compare and challenge.
Preconditions
A public buyer able to specify a complex clinical programme before contracting
Independent financial, clinical and risk scrutiny that can test a PPP against conventional procurement
Governance that keeps service requirements and change control aligned over a multi-decade contract
Where else it has been used: Nya Karolinska travels as a cautionary PPP pattern: lifecycle contracts can bundle accountability, but benchmarkers need to separate build cost, financing, maintenance and service outcomes before calling a project efficient.
Limits: The inquiry distinguishes the 22.8 billion construction-and-equipment investment from the 57 billion projection through 2040, which includes operations, maintenance, financing and additions; those figures are not a like-for-like cash budget. The 53% comparison is an external estimate with explanations including higher standards, site conditions, schedule and PPP risk transfer, so it does not isolate one causal failure or measure clinical performance.
๐ธ๐ชSwedensince 1970dismantledbears on Large project delivery
Sweden permanently shut Barsebรคck 1 in 1999 and Barsebรคck 2 in 2005; the national nuclear-safety report lists each reactor at 615 MWe and records the site as undergoing decommissioning, a 1,230-MWe asset retirement completed by political decision rather than construction failure.
Sweden converted a political decision into a regulated, staged retirement: operators removed fuel, kept the site under care and maintenance, and moved toward dismantling under the nuclear-safety authority's oversight. The capability is visible in the exit path and accountability after closure, not only in the original build.
Preconditions
A regulator with authority to supervise shutdown, fuel removal and decommissioning
Long-lived financing and liability arrangements for an asset that no longer produces revenue
A replacement-power and local-transition plan that makes a politically chosen closure workable
Where else it has been used: Barsebรคck travels as a durable-exit pattern: a benchmark should count the institutional work of retiring infrastructure, while keeping the energy trade-offs separate from the fact of closure.
Limits: Retiring a nuclear asset is evidence about a state's ability to close and decommission a major system, not about whether the original plant was safe, economical or replaceable without wider grid effects. Capacity is nameplate output rather than annual generation, and the record does not quantify replacement power, compensation or the final decommissioning cost.
๐ฎ๐ชIrelandsince 2016operating below its peakbears on Large project delivery
Ireland's New Children's Hospital project saw its approved budget rise from โฌ1.73 billion to โฌ2.24 billion by February 2024 after an independent review recorded โฌ450 million of earlier escalation and weaknesses in set-up, planning, execution and governance.
Ireland kept a politically necessary hospital project moving after cost escalation, but the state had to reset the budget and strengthen assurance rather than re-tender. The case makes the value of early challenge visible: once design, contracts and site work are locked, correcting an underestimated scope becomes more expensive than revising the approval.
Preconditions
An independent gateway review that tests scope, budget, risk and delivery capability before construction
A sponsor with enough clinical and commercial expertise to challenge designers and contractors
Change-control rules that expose scope growth and schedule slippage to ministers and the public
Where else it has been used: The hospital travels as a megaproject caution: a reset can preserve a needed public asset, but only an early assurance function can prevent a politically protected project from consuming its contingency before the building opens.
Limits: The approved budget includes the main hospital, satellite centres, commissioning, information systems and service integration, so it is not a pure construction-cost comparison. The 2019 review attributes escalation to several interacting factors, and the project was still being delivered when the 2024 budget was approved; the record therefore shows an eroded cost and governance promise, not a completed outcome evaluation.
๐ณ๐ฎNicaraguasince 2007still operatingbears on Large project delivery
Nicaragua's Ministry of Energy and Mines reported national electricity coverage of 99.575% in December 2024 after public electrification projects brought power to 732,427 homes and 3,853,005 people.
Nicaragua combined public transmission and distribution investment with rural electrification projects, tracking household coverage as a national delivery target. Extending the grid into remote departments and reporting the remaining gap made a dispersed infrastructure task legible as one programme.
Preconditions
A utility and ministry able to plan transmission, distribution and last-mile connections across departments
Capital budgets and procurement arrangements that can reach remote and low-density settlements
A household-level coverage register and maintenance capacity so new connections remain usable after construction
Where else it has been used: Near-universal connection targets travel only with last-mile finance and a maintenance system; a high connection percentage without reliable service can overstate the capability delivered.
Limits: The ministry's coverage and home counts show connection reach, not continuity, affordability, voltage quality, reliability, losses, maintenance or regional inequality. They are government administrative estimates, and the long-run change from the 2006 baseline does not isolate which projects, utilities or financing streams produced the increase.
๐ป๐ชVenezuelasince 2011still operatingbears on Large project delivery
Venezuela's Gran Misiรณn Vivienda Venezuela reported delivery of its five-millionth home on 7 June 2024, with the programme saying that community self-construction supplied 75% of its construction in the preceding four years.
Venezuela used a central housing mission to aggregate land, materials and public construction while shifting more of the build into community self-construction. A national target and local delivery groups linked a large capital programme to household-level participation.
Preconditions
A central land, finance and materials pipeline that can allocate housing projects across regions
Local organisations able to manage community construction and verify beneficiaries
A housing registry and inspection process that can track units from allocation through occupancy and maintenance
Where else it has been used: Community self-build can extend a state housing programme when it is backed by material supply, technical supervision and a beneficiary registry; without those controls, a headline unit count can conceal unfinished or unoccupied stock.
Limits: The cumulative total is reported by Venezuela's state broadcaster and does not establish independent verification, construction quality, occupancy, maintenance, location, tenure security, cost or whether every unit was newly built rather than rehabilitated. The 75% self-construction figure is a programme claim, and the source supplies no audited project-level cost or schedule data.
๐ต๐นPortugalsince 2021still operatingbears on Large project delivery
Portugal's PRR barometer recorded โฌ21.9 billion across 121 investments and 43 reforms, with โฌ14.43 billion paid to beneficiaries by 27 August 2026.
Portugal bound a large cross-government investment programme to a reform-and-milestone schedule, central monitoring and payment requests, while leaving implementation to ministries, public bodies, firms and local beneficiaries. The financing contract creates a common delivery clock across otherwise separate projects.
Preconditions
A central authority able to consolidate milestones and submit evidence for payment
Project owners with procurement, co-financing and reporting capacity
A transparent public dashboard that exposes slippage and reallocations before the deadline
Where else it has been used: Milestone-funded national plans travel when the payment gate is credible and project owners can report comparable evidence; a large envelope without a delivery spine becomes a catalogue of announcements.
Limits: Funds paid, investment counts and reform counts are programme inputs and milestones, not completed infrastructure, cost and schedule performance or long-run outcomes. The plan is EU-funded and centrally coordinated but executed through many ministries and beneficiaries; the barometer's aggregate totals cannot attribute results to one delivery unit or show which projects will remain useful after the funding window closes.
๐ต๐นPortugalsince 2010dismantledbears on Large project delivery
Portugal's 2011 Council of Ministers formally abandoned the previous LisbonโMadrid high-speed rail project, replacing its passenger priority with a freight connection through Sines, Lisbon, Setรบbal, Poceirรฃo and Madrid.
Under fiscal pressure, the government stopped a politically committed passenger megaproject by resolution and redirected the corridor toward freight interoperability, changing the objective before further construction commitments accumulated.
Preconditions
A cabinet with authority to revoke or redirect a concession-backed transport priority
A planning and appraisal process capable of naming the financial risk of the existing pipeline
An alternative corridor objective that can preserve part of the strategic value while dropping the original scope
Where else it has been used: Project reversals travel when governments can expose sunk costs and change scope openly; abandoning a line without a credible replacement only defers the delivery problem.
Limits: The number is categorical: one formally abandoned national project. The resolution records a policy reversal, not the project's final cost, the amount already spent or a completed line; the replacement freight priority and a later revival of high-speed planning mean this is evidence of stop-and-reframe capacity, not proof that Portugal solved rail delivery.
๐ฎ๐ณIndiasince 2003operating below its peakbears on Large project delivery
India and Delhi hosted the Commonwealth Games in October 2010, but the Comptroller and Auditor General recorded the total estimate rising from โน1,200 crore in the May 2003 bid to โน18,532.31 crore by October 2010.
A multi-agency event was delivered against an immovable deadline, but scope changes, late procurement and fragmented governance allowed the budget estimate to expand more than fifteen-fold between the bid and the event.
Preconditions
A host agreement that bound the Union and Delhi governments to the event and its liabilities
A portfolio of venues and city works dependent on many implementing agencies
Independent audit access to plans, estimates, procurement files and delivery evidence
Where else it has been used: Large-event delivery travels badly when an immovable deadline rewards late scope acceptance; an early integrated budget and clear decision rights matter more than the event brand.
Limits: The Games were delivered, so this is not a failed event; it documents erosion of the original cost promise. The CAG compares estimates assembled at different stages and scopes, and the October 2010 figure is not a single audited final-cost total. The audit identifies fragmented planning, late decisions and weak controls but does not attribute every rupee of the increase to one cause.
South Africa's National Research Foundation and SARAO completed the 64-dish MeerKAT radio telescope in March 2018 on schedule and on budget, then began science operations as a precursor to the Square Kilometre Array.
A national research agency led a staged telescope build, using prototype arrays, local engineering and university and industry partners to test components before integrating the full 64-dish facility in the Karoo.
Preconditions
A long-lived public research institution able to hold a technically demanding project across administrations
Prototype infrastructure and specialist engineering capacity before full-scale construction
A remote site, fibre and power systems, and a scientific user community ready to operate the facility
Where else it has been used: Research infrastructure travels when prototypes, procurement and user operations are designed as one delivery chain; a completed instrument without a funded operating model becomes a monument rather than capability.
Limits: SARAO's institutional account states that the array was on schedule and on budget but does not publish an audited cost or baseline schedule in this record. Sixty-four dishes prove physical completion, not scientific productivity, local industrial spillovers or the performance of the larger SKA programme into which MeerKAT is being integrated.
๐ฒ๐ฝMexicosince 2019still operatingbears on Large project delivery
Mexico's state telecom operator CFE Telecomunicaciones e Internet para Todos installed 113,683 free public internet sites by September 2024, reaching 81.2 percent of its stated 140,000-site 2025 goal.
A state-owned carrier combined satellite, fibre, copper and mobile contracts with public access points, using a federal operator to reach places commercial networks had not served.
Preconditions
A public operator able to aggregate several access technologies under one programme
Procurement and spectrum arrangements that reach isolated localities
Public facilities where a connection can be installed and maintained
Where else it has been used: Universal-access builds travel when a state carrier can fill commercial gaps, but the last fifth of a numerical rollout target is often the hardest and most expensive part.
Limits: The management report is self-reported and counts installed sites, not uptime, geographic service quality, affordability after the free offer or actual use. The 81.2 percent figure is calculated from the report's installed count and stated goal, and the programme was still short of that goal at the reference date.
๐ฒ๐ฝMexicosince 2014dismantledbears on Large project delivery
Mexico cancelled the partially built Texcoco airport after construction and contracts had begun; the government's 2019 cancellation paper estimated a 100 billion peso cancellation cost, including 60 billion pesos already spent and 40 billion in non-recoverable commitments.
100,000,000,000 Mexican pesos, government estimate of cancellation cost, 2019-04-26. 60,000,000,000 Mexican pesos, construction expenditure already incurred at cancellation, 2019-04-26. Grupo Aeroportuario de la Ciudad de Mรฉxico, Government of Mexico, retrieved 2026-09-02.
The move, and what it needed
A new administration stopped a politically contested megaproject, terminated contracts and redirected the airport strategy to a different site, accepting sunk costs in exchange for a different delivery path.
Preconditions
Executive authority over the state airport holding company and the federal investment plan
A contracting and bond framework that permits negotiated termination
An alternative site and delivery institution available before the old project was abandoned
Where else it has been used: Stopping a megaproject travels as a governance option only when the government can disclose the sunk-cost trade-off and still offer a credible replacement; otherwise cancellation becomes delay with a bill attached.
Limits: The 100 billion peso figure is the government's own estimate and excludes or treats differently the financing and legal costs that later audits and congressional analyses debated. Cancellation is evidence of a project running backwards and of a political decision to change course, not evidence that the replacement airport delivered better value or schedule performance.
๐น๐ทTurkeysince 2004still operatingbears on Large project delivery
Tรผrkiye opened the Marmaray rail tunnel between Europe and Asia in October 2013 and the transport ministry reported 196 million passenger journeys in 2024, with 1.226 billion cumulative journeys since opening.
The state assembled a cross-Bosphorus rail link, integrated it with existing suburban lines and then built ridership through frequent service rather than treating the tunnel as a stand-alone monument.
Preconditions
A metropolitan rail network on both sides of the crossing
Engineering and procurement capacity for a deep immersed-tube tunnel
An operating authority able to integrate timetables, fares and maintenance
Where else it has been used: A megaproject becomes capability when the operating system is funded and connected to everyday routes; a tunnel that opens without network integration will not produce this usage.
Limits: Passenger journeys show that the completed line is used at scale, not whether construction met its original budget or schedule. The ministry's cumulative figure is a programme-reported total, and ridership also reflects Istanbul's population, fares and the network of connecting lines rather than the tunnel alone.
๐น๐ทTurkeysince 2011still operatingbears on Large project delivery
Tรผrkiye's Ministry of National Education reported that the FATฤฐH project had installed 620,000 interactive classroom boards by 2023 and signed a five-year TรRKSAT service contract covering about 2,000 schools.
620,000 boards, interactive boards installed under fatiฬh, 2023. 2,000 schools, schools covered by the continuing service contract, 2023. Ministry of National Education, Government of Tรผrkiye, retrieved 2026-09-03.
The move, and what it needed
The education ministry treated classroom hardware, connectivity and content as one procurement programme, then used a national telecom operator for continuing access rather than ending delivery at installation.
Preconditions
A central school inventory and procurement authority
Electricity and connectivity in ordinary classrooms
A service provider able to maintain equipment across thousands of schools
Where else it has been used: Education technology projects travel when maintenance and teacher practice are funded alongside hardware; a board installed without a service contract quickly becomes furniture.
Limits: Equipment counts show installation, not whether teachers use the boards or whether learning improved. The ministry reports the hardware and contract itself, and the article does not give a national uptime, maintenance-cost or student-outcome series; the 2,000-school contract is a subset of the wider rollout.
๐น๐ทTurkeysince 2017dismantledbears on Large project delivery
Tรผrkiye signed the 1,000 MWe YEKA RES-1 wind contract in February 2018 and completed its factory and R&D centre in 2019, but the Energy Ministry records that the contract was cancelled under a 2024 law.
The state tied a renewable auction to domestic manufacturing and research obligations, then later used a statutory change to cancel the contract despite preparatory industrial facilities having been completed.
Preconditions
A ministry able to bundle generation rights with local-content obligations
Private consortia willing to build manufacturing and R&D capacity ahead of generation
Legislative authority able to reopen or cancel a signed concession
Where else it has been used: Industrial-policy tenders travel only when contract stability is part of the bargain; a local factory does not rescue a project whose legal foundation can be rewritten midstream.
Limits: The ministry page establishes the contract, capacity, factory and R&D milestones and the later cancellation, but it does not publish the cancellation cost, foregone generation or the reasons for the legislative change. Counting two completed facilities is a direct count of the page's milestones, not an independent audit of their operation; cancellation is evidence about delivery durability rather than renewable output.
๐จ๐ญSwitzerlandsince 1998delivered and closedbears on Large project delivery
Switzerland built the 57 km Gotthard Base Tunnel under the voter-approved NEAT programme and opened it to commercial traffic on 11 December 2016, on the date set, at a final cost forecast of CHF 9.541 billion against a 1998 cost basis of CHF 6.323 billion, both in 1998 prices.
Voters approved the programme and a ring-fenced rail fund, parliament set a total credit and a standing supervisory delegation that reports every year, and a single-purpose federal subsidiary, AlpTransit Gotthard, built it under that oversight.
Preconditions
Direct democracy that ties a ring-fenced fund to one programme
A standing parliamentary delegation that publishes cost forecasts every year
Tunnelling and rail engineering capacity already in the country
Where else it has been used: The Brenner and Mont Cenis base tunnels copy the engineering but run across two states with EU co-financing, and both have moved their opening dates repeatedly.
Limits: The 51 percent is computed from the NAD's Table 2 (6,323 to 9,541 million francs, 1998 prices) and is not a headline the publisher states. That column mixes true overrun with scope added after 1998, including extra safety works and the Sedrun geology, so a like-for-like overrun would be lower. The cost basis is in real 1998 francs; the nominal bill, about CHF 12.2 billion with inflation, VAT and interest according to the Federal Office of Transport, is larger. On schedule means against the timetable fixed late in the project, which the NAD calls on time, not against the opening years implied by the 1992 vote. The started year is the 1998 financing vote; main construction contracts followed. The source is a PDF and the figures sit in Table 2 and section 1.8.
๐ฐ๐ชKenyasince 2014still operatingbears on Large project delivery
Kenya Railways began building the Standard Gauge Railway in financial year 2014/15, completed the 472 km Mombasa to Nairobi Phase I in 2017, and opened the 120 km Phase 2A to Naivasha in September 2019 against a September 2017 start and an 18-month plan, about six months late; the line hauled 7.33 million tonnes of freight in 2025.
18 months, planned construction period of sgr phase 2a (sept 2017 start, ksh 155.5 billion), against which the line opened in september 2019, about six months (a third) late, 2017. 7,332 thousand tonnes, freight hauled on the sgr, 2025. Kenya National Bureau of Statistics, Economic Survey 2018, retrieved 2026-10-01.
The move, and what it needed
The Treasury and Kenya Railways contracted a Chinese state contractor on a design-build-supply basis with Exim Bank finance, which put construction risk and speed on the contractor and left financing, land acquisition and traffic risk with the state.
Preconditions
A bilateral lender willing to finance most of the cost against sovereign guarantees
A state able to acquire a 472 km corridor of land quickly
A port whose cargo the state could steer onto rail to fill the line
Limits: No official source found publishes the outturn cost of either phase against its budget, so the record carries a cost plan (KSh 155.5 billion for Phase 2A) and no cost performance. That is the half of the construct that matters most for a Chinese-financed turnkey contract. The schedule overrun of about six months on Phase 2A, or a third of the 18-month plan, is computed: the plan is from KNBS, and the opening is from the Ministry of Roads and Transport Strategic Plan 2023-2027 draft ('completed Phase 2A ... and was launched for operation in September 2019', https://www.transport.go.ke/sites/default/files/DRAFT%20FINAL%20MINISTERIAL%20STRATEGIC%20PLAN%202023-27%20FEB%202024%20.pdf). Phase I's schedule against its plan cannot be checked from an official copy. The National Assembly's 2014 Public Investments Committee report giving a December 2016 completion date and a USD 3.804 billion contract price was found only on an NGO mirror. The parliamentary repository timed out, so neither figure is used. The railway was meant to run to Malaba and into Uganda, and the Naivasha to Malaba section remained unbuilt for years. The same ministry plan reports rail carrying 19 percent of port throughput in 2021 against a 50 percent target for 2030. Freight tonnage is published by KNBS from Kenya Railways returns. Debt service and fiscal cost are not shown at all.
๐จ๐ฆCanadasince 1998dismantledbears on Large project delivery
Canada began registering all firearms under the 1995 Firearms Act from December 1998. The registry's first information system, CFIS I, had an initial estimate of $13 to 20 million and had cost $190 million by March 2005. Its successor CFIS II rose from $32 million to at least $87 million and was halted. In 2012 Parliament ended long-gun registration and ordered the destruction of about 7.5 million non-restricted registration records, which the RCMP deleted from CFIS in October 2012.
The Department of Justice, and later the Canada Firearms Centre, built a universal national registry through contracted IT development while the regulations it had to implement were still changing, and reported the rising cost to ministers before Parliament. In 2012 Parliament ended the long-gun part by statute and had the RCMP destroy its records.
Preconditions
Legislation and regulations settled before the system is specified, so requirements do not move during the build
Cost reporting to Parliament that covers the full programme, not only what is appropriated each year
A political settlement on registration durable enough to survive a change of government, or a design that does not depend on one
Limits: The cost figure is a Commons committee's summary of the Auditor General's 2006 follow-up. The Auditor General's own 2002 and 2006 chapters no longer open at their published OAG URLs, so the record rests on Parliament's restatement of them. The $190 million covers CFIS I, the information system, and is not the cost of the whole programme. The often quoted figure of roughly $1 billion for the programme could not be verified at an official URL today and is not used. The baseline is a range of $13 to 20 million given at an early stage, so any overrun ratio, roughly ten times, depends on which end is chosen and on how far the scope of that early estimate matched the system that was built. The committee itself attributes part of the growth to things outside the project team's control: requirements kept changing, and Bill C-10 and its regulations were delayed by more than two years. The dismantling in 2012 was a political decision by a new parliamentary majority. It was not a collapse caused by the delivery, so the status records a policy reversal rather than a technical failure. Licensing and the registry of restricted and prohibited firearms kept running, Quebec residents' records were held back pending litigation, and public agencies still register. The 7.5 million records show the scale the registry reached and say nothing about their accuracy, which the 2006 audit questioned. Like can-phoenix, this is a public IT system rather than physical infrastructure.
๐ฎ๐ฑIsraelsince 2015still operatingbears on Large project delivery
Israel's government approved the Red Line in December 2010 at an estimated NIS 10.7 billion, and NTA, the state-owned company delivering it, began the main construction works in 2015. By October 2021 the cost estimate had reached NIS 18.71 billion, about 75% above the original, and the completion date had been postponed six times, by about five years in total. The 24 km line opened in August 2023, and in December 2025 the State Comptroller put its construction cost at NIS 18.8 billion.
The government set the budget and target date by resolution in 2010 and handed delivery to NTA, a state-owned company. NTA revised the cost estimate at least eight times and moved the opening date six times, most of them before main construction began, and still brought the line into service in 2023.
Preconditions
A dedicated state delivery company with a signed framework agreement setting its accountability to the transport and finance ministries, which the audit found missing
Contingency budgets set by the transport projects procedure from the first estimate, rather than costs left out of early estimates
Land clearance and right of way secured before the main contracts are tendered
Limits: The 75% overrun is the audit's figure for the October 2021 estimate and includes indexation. Net of price-index linkage and a VAT cut, the audit puts the real increase at about NIS 7.46 billion, or 69.7%. The audit was finished before the line opened, so the overrun is measured on an estimate rather than a final account. The December 2025 audit's NIS 18.8 billion construction cost comes from a security audit and is not a reconciled final cost. The five-year delay counts slippage to the then target of 2022. The line actually opened in August 2023, about six years after the original 2017 date, so the figure understates the final schedule slip. Most postponements came before main construction began in 2015, so a hostile reader would call them planning slippage rather than construction delay. The project is national in its funding, its government resolution and its state-owned delivery company, but it serves one metropolitan area, and one line is a single case and not a national delivery record. The audit's per-kilometre comparisons with other cities are not used here. Ridership is not established: the 300,000 passengers a day in the 2025 audit is NTA's forecast. The audit also found the framework agreement between NTA and the finance and transport ministries still unsigned more than 20 years after the company was founded.
๐จ๐ณChinasince 2003still operatingbears on Large project delivery
China's railway authorities built a high-speed rail network that the National Railway Administration put at about 47,000 km in operation in December 2024, from a programme that began in 2003; a World Bank review of the 27 lines in operation at the end of 2013 found a weighted average approved unit cost of RMB 129m per km for 350 km/h lines, which it put at no more than two-thirds of costs elsewhere.
The former Ministry of Railways and its successors standardised designs, viaduct and station types and procurement across a national programme, then built continuously for two decades, so each line reused the previous line's templates, supply chain and crews.
Preconditions
A single national rail authority with power to plan corridors, acquire land and finance lines through state-owned debt
A continuous build pipeline long enough for standard designs and a domestic supply chain to pay off
Tolerance for debt and for low-traffic lines whose cost is carried off the general budget
Where else it has been used: Standardisation and a continuous pipeline are what the World Bank note offers as transferable; other countries that build one line at a time keep the high unit costs the note compares against.
Limits: Neither number is the overrun this gap asks for. The World Bank unit cost is built from estimated cost at project approval, not out-turn cost, so it cannot show whether lines came in over or under budget; the note itself warns that the data exist only in aggregate, that expenditures fall in different years and may not be comparable, and that its inputs come from China Railway and former Ministry of Railways publications. It is a 2014 staff note in a working series, covering about 10,000 km built by 2013, not the 47,000 km now operating, and the World Bank lends to Chinese railway projects. The 47,000 km figure is a rounded administrative statement by the regulator, relayed by Xinhua on the State Council portal, not a statistical table. Neither source shows the financing side: the debt carried by China State Railway Group and its predecessor, the low traffic density on many later lines, or the 2011 Wenzhou collision and the corruption cases at the former Ministry of Railways. Low unit cost also reflects cheap labour and land acquisition terms that are not institutional choices other countries can copy. No schedule-performance figure was found at an official URL.
The UAE, through the Emirates Nuclear Energy Corporation and under licences from the Federal Authority for Nuclear Regulation issued from 2010, built four APR-1400 reactors at Barakah that ENEC reported in September 2024 were all in commercial operation and providing up to 25 percent of the UAE's electricity; Unit 1 entered commercial operation on 6 April 2021, about four years after the 2017 date the regulator recorded in its 2012 construction licence review.
The UAE federal government set a nuclear policy in 2008, created an independent regulator (FANR) and a single owner-developer (ENEC), and bought a proven reactor design with a turnkey contract, so the state managed one prime contractor against one reference plant rather than a first-of-a-kind design.
Preconditions
A sovereign balance sheet able to finance a multi-unit plant over a decade of construction and delay
A regulator built from scratch with foreign technical support, credible enough to hold the schedule back when licensing required it
A vendor with a recent reference plant and an intact domestic supply chain willing to export it
Where else it has been used: The turnkey, reference-plant model is what newcomer states such as Saudi Arabia and Poland cite; the four-year first-unit slip shows the regulator and operator readiness, not the construction, set the pace.
Limits: The 25 percent is the operator reporting on itself, relayed by the Abu Dhabi government's media office, and it is phrased as "up to": it is a capacity claim, not a measured share of national generation in a given year, and no statistical series was opened to confirm it. The schedule figure is computed by us from two documents, a 2012 regulator statement that the first unit was planned for 2017 and the 2021 commercial-operation release; it is a slip of roughly four years for Unit 1 only, and the later units' original dates were not found at an official URL. The plant's construction cost, and any cost overrun, are not published in any source opened, so the cost half of the construct is missing. The reactors were designed and built by a KEPCO-led Korean consortium on the Shin-Kori 3 and 4 reference design, so the UAE's institutional contribution is commissioning, licensing, financing and operating, not engineering. ENEC is an Abu Dhabi entity and the plant sits in one emirate, though it supplies the national grid under federal regulation. The operator's own boast of "a 40 per cent improvement in schedule" from Unit 1 to Unit 4 measures operational readiness to commercial operation, not construction, and is not used here.
๐ญ๐ณHondurassince 2016still operatingbears on Large project delivery
Honduras's Infrastructure Secretariat concessioned the Palmerola international airport to Palmerola International Airport S.A., with a USD 87.1 million reference investment. Works began on 5 December 2016 and operations on 11 December 2021, about 60 months after works began, against a 61-month contracted works period (start of operation is not necessarily completion of works).
Honduras's Infrastructure Secretariat let a 34-and-a-half-year concession with a fixed works period, had Congress approve it and a trust carry it, and left SAPP and a supervising consortium to hold the concessionaire to the schedule.
Preconditions
A PPP law and a regulator able to publish and enforce contract milestones
An existing site and runway that removed most land acquisition and airside risk
Private bank financing that reached financial close (August 2018) before the main works
Limits: The schedule figure is ours, computed from two dates on the regulator's fact sheet: about 60 months against a 61-month contracted period. The sheet does not say whether the works period was amended, extended or suspended, or whether 'start of operation' means full completion. The only cost figure is a reference investment set at signature. No final cost, cost overrun or state contribution is published there, so this record shows schedule performance only and is silent on cost, which is half of what the indicator asks. The terminal was built at an existing air base with a runway, which makes it an easier build than a greenfield airport. SAPP is the concession's own regulator. Other SAPP sheets carry copy errors (the EEH sheet cites airport revenues), so treat field values with care. Passenger demand, the concession's financial health and whether the airport displaced Toncontรญn as intended are not shown.
๐ธ๐ปEl Salvadorsince 2008delivered and closedbears on Large project delivery
El Salvador's state hydro commission CEL financed the El Chaparral hydroelectric plant on the Torola River through a BCIE loan arranged in September 2008, with four planned years of construction, and inaugurated it as the 3 de Febrero plant on 19 October 2023 at an approximate cost of US$800 million; CEL gives the initial investment as US$207.8 million, so the final cost is about 3.8 times the initial figure, some 15 years after the 2008 loan.
CEL contracted a 65.8 MW dam, financed by BCIE loans the Assembly guaranteed, and when the original design failed it refinanced a redesign through two further BCIE loans and completed the plant under a later government. Successive administrations kept financing the asset rather than abandoning it, at the price of a cost multiple.
Preconditions
A state utility with borrowing capacity and access to a regional development bank willing to extend and enlarge loans
Geological and design studies whose adequacy is only tested once excavation begins
A legislature that keeps guaranteeing sovereign debt for a stalled project across changes of government
Where else it has been used: The case sits with other hydro and power projects in the corpus whose initial prices did not hold; it shows that an initial contract price transfers little risk when the state must pay for a redesign to finish the asset.
Limits: Both cost figures come from the executing agency reporting on its own project, and both are stated loosely: the final figure is the President's "costo aproximado" quoted in a CEL news item, not an audited account, and CEL's own 2022 notice gave the final cost as "more than $750 million" while the President's speech gave the initial cost as $200 million rather than $207.8 million. The overrun ratio of about 3.8 is our computation from two different documents, and CEL gives the initial figure with no year: it appears to be the price for a design later replaced after technical failures, so part of the gap reflects a changed scope, not only cost escalation; no figure separates the two. The cost statements are politically framed, attributing the overrun to previous governments, and the government that completed the plant is the source. No Court of Accounts final audit was retrieved. The start year rests on the date of the first BCIE loan reported by the Assembly, not on a published construction start date, and the 15-year span is computed from that date to inauguration; the four-year planned duration comes only from the President's quoted speech. The figures say nothing about the plant's output, which is 65.8 MW of installed capacity per the Assembly page.
๐ป๐ชVenezuelasince 1980operating below its peakbears on Large project delivery
WHO and UNICEF estimated Venezuela's national DTP3 coverage among one-year-olds at 50 percent in 2025, down from 87 percent in 2015, and coverage has not returned above 66 percent in any year since 2017.
Venezuela's health ministry ran a free national schedule through public clinics and held DTP3 near or above 80 percent for most of 2009 to 2016; when national procurement of vaccines and syringes failed during the economic collapse, the ministry kept the programme formally in place while coverage fell to around half of infants.
Preconditions
Hard-currency budgets and procurement able to keep vaccines and syringes in national stock
A clinic network staffed by vaccinators who stay through a crisis
Population denominators accurate enough that a fall in coverage is seen before an outbreak reveals it
Where else it has been used: Haiti shows the same erosion from stockouts and weak reporting; Venezuela adds that a fall can sit inside a denominator change for two years before an outbreak (measles in 2017 to 2019) makes it visible.
Limits: Construct fit is borrowed, not literal: DTP3 coverage is not the cost or schedule performance of an infrastructure project. It is filed here only on the bra-pni precedent, which EVIDENCE.md names as the reversal template (hti-immunization-erosion files the same kind of series under disaster_preparedness instead); a reviewer holding to the literal construct should reject it. The 87 percent peak is a single good year in a volatile series (50 in 2008, 87 in 2005 and 2015), and the WHO/UNICEF 2022-revision country profile (https://cdn.who.int/media/docs/default-source/country-profiles/immunization/2023-country-profiles/immunization_ven_2023.pdf) says the 2015 rise was 'due in part to a change in the population source', so part of the drop is a denominator artefact. Estimates rest on Venezuelan administrative reports with no nationally representative survey in the last five years; WHO/UNICEF graded recent years as having no accepted empirical data, and values are revised between rounds (2022 was 56 in the 2022 revision and 43 now). The same profile records national vaccine and syringe stockouts from 2016 to 2020, which locates the failure in supply rather than in the programme's design, but the series cannot separate economic collapse, emigration of health workers, sanctions-era procurement limits and denominator error. DTP3 is one antigen. PAHO separately documented that endemic measles transmission was re-established in Venezuela in 2018 and interrupted again in 2019 (https://journal.paho.org/en/articles/comprehensive-approach-achieving-reverification-measles-elimination-venezuela-2019-2023), which supports the erosion but is not part of this metric.
๐ญ๐นHaitisince 2008dismantledbears on Large project delivery
From September 2008 to September 2016, Haiti's Council of Ministers allocated PetroCaribe oil-credit revenue to 409 public projects, and the state fuel monetisation bureau BMPAD disbursed US$1,605,905,287.84 (92.4% of the revised budget), but the Superior Court of Auditors (CSCCA) found cost and schedule overruns across the audited works, including a Cap-Haรฏtien street rehabilitation that ran 18% over budget and took 21 months against an eight-month contract term.
Successive Haitian governments turned Venezuelan oil credit into a parallel capital budget, approving projects by Council-of-Ministers resolution under emergency law and having ministries such as MTPTC and MPCE award contracts directly. Spending reached US$1.6 billion, but without needs analysis, cost estimates or contract supervision, many works overran, stalled or could not be verified, and the money stopped when the oil credit ended.
Preconditions
A budget channel that runs through normal appropriation and procurement rather than ministerial resolutions under emergency powers
Project preparation, costing and supervision capacity in line ministries before money is committed
An independent audit body able to follow disbursements to physical completion while projects are live, not years later
Where else it has been used: Resource-backed off-budget investment funds in other oil-credit recipients show the same pattern: rapid disbursement, weak preparation and audits that arrive after the financing has gone.
Limits: The headline number is money disbursed, not works delivered: the CSCCA itself says missing reception certificates often stop it concluding whether a contract was finished or abandoned, and it documents abandoned works such as the Miragoรขne market and bus station, stopped at 35.27% progress, a figure from report 1 (31 January 2019). The court publishes no aggregate cost or schedule overrun for the 409 projects. The 18% overrun is one audited road project picked as the cost figure, with no claim that it is typical; other audited projects show overruns of 9%, 10%, 33% and 54% against contract (reports 1 and 3), and some came in under budget. Report 1 gives the Cap-Haรฏtien contract an 8-month term but a contractual end date of 14 May 2011 (start 18 March 2010), so the schedule overrun is 21 months against 8 or against about 14 depending on which line of the court's own sheet is read. The audit covers a sample of projects across three reports of roughly 1,000 pages each, every figure has to be read out of a PDF, and contractors contested several findings in report 3's annex. The 'dismantled' status rests on the audit ledger: the last resolution is dated 28 September 2016 and Venezuelan PetroCaribe supply then collapsed; no single Haitian text formally closing the programme was located. Projects were executed in a post-earthquake state of emergency that suspended normal procurement rules, which the court says enabled direct awards.
Under the Federal-Aid Highway Act of 1956, the federal government and the states built the Interstate System. In nominal dollars, the estimated federal cost to complete it rose from $34 billion in the 1958 Interstate Cost Estimate to $114 billion in the final estimate of 1991.
114 USD billion, estimated federal cost to complete the interstate system, final (1991) interstate cost estimate, nominal, 1991. 34 USD billion, estimated federal cost to complete the interstate system, first (1958) interstate cost estimate, nominal, 1958. US Federal Highway Administration, Highway History: How many Interstate programs were there?, retrieved 2026-10-01.
The move, and what it needed
Congress set a 90 percent federal match paid from a dedicated Highway Trust Fund fed by fuel taxes, and the Bureau of Public Roads (later FHWA) apportioned it to state highway departments by repeated cost-to-complete estimates, which let the states build to one national standard on a pay-as-you-go basis.
Preconditions
A dedicated user-tax revenue stream that kept construction off the general budget
State highway departments with existing engineering capacity to design and let contracts
A legislated national network map and design standard fixed before money flowed
Where else it has been used: The trust-fund-plus-match model was copied in transit and airport programs; it travels less well where fuel-tax revenue is small or where no subnational agencies can deliver.
Limits: Both figures are in dollars not adjusted for inflation, so the 235 percent rise we compute from them, (114-34)/34, is not a real cost overrun. FHWA itself attributes the growth to inflation, mileage added by Congress in 1968 and later, 'more safety and more attention to the environment' in later programs and 'changes in the definition of completion', as well as to the Interstate Program 'evolving'. The page gives no real-terms figure, no schedule baseline and no per-mile comparison, so it cannot separate execution failure from scope change. The 1958 ICE was a cost-to-complete estimate used to apportion funds, not a contract budget. The source gives no schedule baseline; the final estimate came 35 years after the Act. The source says later Interstate Programs have not ended, so 'concluded' refers to the end of Interstate Construction apportionments under ISTEA. The figures are federal share only.
๐จ๐บCubasince 2019operating below its peakbears on Large project delivery
Cuba's national housing programme, run by the Ministry of Construction under the housing policy presented in 2018 to close a deficit of 929,695 homes in ten years, completed 20,232 homes in 2022 and 5,493 in 2025, of which the state built 1,791.
The Ministry of Construction sets an annual national housing plan, builds part of it through state enterprises and counts the rest through subsidised self-build by households, with municipal governments responsible for local materials.
Preconditions
A domestic supply of cement, steel and blocks that the state controls and can allocate
Budget room for state construction and household subsidies
Municipal materials production able to supply self-builders when national supply fails
Limits: The count is of homes finished, not of the housing deficit, quality, cost or who received them. Two thirds of the 2025 total (3,702 of 5,493) is self-build by households ('poblaciรณn', 'esfuerzo propio'), which the programme subsidises and counts toward its plan but does not build, and the state-built part fell from 2,756 in 2024 to 1,791. The 2022 figure is from the 2024 edition of the same ONEI publication (https://web.archive.org/web/20260621201114/https://www.onei.gob.cu/sites/default/files/publicaciones/2025-05/construccion-en-cuba-indicadores-seleccionados-2024_0.pdf, table 5), which gives 32,874 for 2020, 18,645 for 2021, 20,232 for 2022, 16,065 for 2023 and 7,427 for 2024. In the 2025 edition's table 5, headed 2021 to 2025, the province rows sum to 18,645, 20,232, 16,065, 7,427 and 5,493, but the national total row is shifted one year back in its first three columns (32,874, 18,645, 20,232) and drops 16,065, so the national row is wrong there and the provinces are right; Granma (27 February 2024) independently reports 16,065 homes for 2023, 65 percent of the plan, from the ministry's annual review. 2022 is not the peak: 2020 was higher, and the series before 2020 is not in these editions. The 2018 policy and the 929,695 deficit are from Cubadebate's report of the policy's presentation (21 November 2018), state media rather than a statistical release, and the started year is the first full year under that policy. The collapse coincides with the 2021 to 2025 economic crisis, the loss of construction materials (ONEI reports corrugated steel bar output down 80 percent in 2025) and 40,191 stalled self-build homes, so the record documents a delivery capacity that ran backwards without separating the programme's management from the economy around it. onei.gob.cu refused connections on 2 October 2026, so both editions are Internet Archive copies; the PDF text layer uses a shifted font encoding and the tables were read after decoding it.
The Dominican state electricity holding CDEEE awarded the Punta Catalina coal plant in November 2013 and agreed a final contract price of US$1,945.0 million after the award, including the unloading port; after construction delays and a March 2020 dispute settlement with the contractor, the value carried into the plant's 2022 public trust was US$2,340.54 million, about 20 percent above the contract price.
The government declared a generation emergency by decree, had CDEEE tender a 2-unit coal plant as a single EPC contract with a foreign consortium, and financed milestone payments from the budget and export credit; when payments and works fell behind, it settled the disputes in 2020 at a higher value and later placed the plant in a public trust run by Banreservas' fiduciary.
Preconditions
An emergency decree that let one state company run the procurement outside normal planning
A single EPC contractor carrying design and construction risk, with milestone pricing open to dispute
An external audit body able to publish a final cost, which here came only years after commissioning
Where else it has been used: Large single-contractor energy builds tendered under emergency powers show the same sequence elsewhere in the region: a fixed headline price at award, delays tied to payment disputes, and a settlement that resets the price before any audit.
Limits: The 20 percent overrun is computed here from two documents by different bodies, not published as a figure. The presidency itself calls the US$2,340.54 million a reference number, "no es definitivo, ni como valor del activo ni como monto final de inversiรณn", pending a financial audit, so it is a floor on cost, not the final cost. The US$1,945.0 million comes from a commission appointed by the president who built the plant, hosted by the Consejo Econรณmico y Social, whose chair led it; it was written in 2017 at about 70 percent completion. The same report records that CDEEE was paying milestones against the higher offered price of US$2,040.7 million, so the baseline itself is disputed. No schedule figure with a planned and actual date was found in an opened official source; the commission only puts the cost of delay at about US$25 million a month. The award followed a tender later investigated for Odebrecht bribery, which the overrun figure does not capture.
๐ซ๐ทFrancesince 2016still operatingbears on Large project delivery
The Sociรฉtรฉ du Grand Paris, the state's public builder, renamed Sociรฉtรฉ des grands projets by the law of 27 December 2023, has been building the Grand Paris Express automatic metro since works began in June 2016. The project's estimated total cost rose from 22.6 billion euros in March 2013 to 36.1 billion euros from October 2021, about 60 percent above the 2013 estimate.
The state created a dedicated public builder with its own earmarked taxes and the right to borrow on bond markets, so that a 200 km metro could be built in parallel rather than paid for line by line. The Senate notes this borrowing moved full delivery from 2075 to 2030, at the price of cost estimates that were revised upward as studies deepened and the Olympic deadline compressed the schedule.
Preconditions
A legal vehicle with earmarked revenue and borrowing power, kept off the annual budget
Enough in-house engineering staff to supervise thousands of contractors, which the Senate says was missing until after 2018
A political commitment to the whole network that survives changes of government, here confirmed in 2013 and again in 2018
Where else it has been used: The builder is now preparing France's regional express rail networks (SERM) using the same model. Those projects have no Olympic deadline and no Paris land values to borrow against.
Limits: The 60 percent overrun is our calculation from two Senate figures. The 2013 baseline is stated in 2012 euros. The 2026 page does not state the price basis of the 36.1 billion figure, but the 2020 Senate report (r20-044) gives the preceding estimate as '35,6 milliards d'euros aux conditions รฉconomiques de 2012', which points to constant 2012 euros, so the overrun is mostly real rather than inflation. Both figures are estimates to completion and not an audited outturn, and the network is still unfinished. The Senate reports that in February 2025 the builder announced new delays to line 15 south and the first sections of lines 16 and 17, so the cost figure, unchanged since 2021, may rise again. The 36.1 billion excludes the further 3.5 billion euros, and 700 million more in 2024, that the builder pays toward other รle-de-France transport projects. The source does not quantify schedule slippage against the 2013 or 2018 timetables. The Cour des comptes reports on the project, the primary audit source, did not open from this machine, so the figures come from parliament's reading of the builder's answers to its questionnaires.
๐ช๐นEthiopiasince 2015delivered and closedbears on Large project delivery
The Ethiopian Roads Administration built the 57.4 km Batu/Zeway-Arsi Negele section (Lot 3) of the 202 km Modjo-Hawassa expressway under a project that became effective in 2015. The civil works were substantially complete by mid-2026, but only after cumulative extensions of 57 months to the project's closing date.
The Ethiopian Roads Administration split a 202 km toll expressway into lots with different financiers: Bank-financed, government-financed and Chinese EXIM-financed. It carried the World Bank lot through material shortages, contract suspensions and repeated restructurings until the civil works were done.
Preconditions
A roads authority able to manage several international contracts and financiers on one corridor
Foreign-exchange and fiscal headroom to absorb price escalation
Land acquisition and compensation processes that do not stall works, which here still leave grievances before the courts
Limits: This is one 57 km road section, not a national programme, and the schedule figure is a slip in the World Bank financing's closing date, not a construction schedule measured against a contractual completion date. The 57 months run from the original 31 December 2021 closing to 30 September 2026, and part of that time came before Lot 3 works began: the paper says phase 2 started about two years after phase 1. The World Bank documents name cost overruns from material price inflation but give no figure, and the financing was partly cancelled (about 105 million dollars cumulatively by June 2026, including 9.31 million on the additional-financing credit) and topped up with 37 million in additional financing, so no cost-performance number can be drawn from them. The status 'concluded' is provisional. The road had been provisionally taken over by the roads authority, but the documents tie public opening to installing tolling and transport systems, which were moved to another project, and no source seen here confirms the section is open to traffic. Lots 1 and 2 have been open since September 2021, while Lot 4 is financed by China EXIM Bank and is not covered. All figures are the World Bank's own supervision reporting on a project it finances.
๐บ๐พUruguaysince 2013operating below its peakbears on Large project delivery
Uruguay's transport ministry signed a FOCEM-funded contract in November 2013 to rehabilitate 422 km of the Pintado to Rivera line within 36 months, cut the scope to the 310 km between Paso de los Toros and Rivera in January 2017, took the last provisional handover in May 2019, and by June 2025 reported the line in a degraded state with freight limited to 10 to 25 km/h against its set maximum of 40 km/h and no freight carried in recent years.
The transport ministry used a regional cohesion fund to finance a capital rehabilitation, contracted a state-owned rail company to deliver it, and then let the maintenance contract lapse in 2021 without a successor, so the rebuilt track degraded within four years.
Preconditions
Grant capital for construction that carried no matching budget line for upkeep
A railway operator without freight demand on the corridor to justify ongoing maintenance
A competing national rail priority, the Ferrocarril Central, that pulled scope and attention away mid-contract
Where else it has been used: Donor-funded rehabilitations that lose their maintenance regime are a familiar pattern in regional infrastructure funds. Here the cut in scope to make way for the Ferrocarril Central shows a smaller project giving way to a larger one built on the same corridor.
Limits: The source is a scanned ministerial reply to a parliamentary information request, so the quoted text comes from the PDF's OCR layer (it names the contractor "CEU" where the contracting firm was CFU), and checking it means reading the scan. The two speeds measure the asset's condition, not cost or schedule, which is what the construct asks for. The schedule slip is our computation: 36 months from a November 2013 signature against a last provisional handover in May 2019, about two and a half years late, and on a scope that had been cut from 422 to 310 km in January 2017, so a like-for-like slip cannot be stated. The final cost of the works was not found at an official URL: only the 2013 budget of US$75 million and a CND contract figure of 1,462 million pesos, which cannot be compared without exchange-rate and indexation assumptions, so no cost overrun is claimed. The 10 km/h figure is the bottom of a range of speed restrictions and does not apply along the whole line. The absence of freight reflects demand and operator choices as well as track condition, since the pulp traffic that justified the Ferrocarril Central runs south of Paso de los Toros. This is a different delivery from the 1988 to 1990 contraction recorded in ury-rail-network, but both concern AFE's network, and the record leans on the same ministry describing its own asset.
๐ฐ๐ทSouth Koreasince 1992delivered and closedbears on Large project delivery
Korea built the Seoul to Busan high-speed line planned in 1990 at 5,846.2 billion won for opening in early 1999; by 1993 the Ministry of Transportation re-estimated the cost at 12,174.3 billion won, the first phase opened in April 2004 and the whole line opened in 2010.
The government committed to a TGV-based Seoul to Busan line on a rough estimate derived from conventional rail costs, discovered within a year of breaking ground that the cost had doubled, then cut scope, created a dedicated construction authority and passed a 1996 promotion act to keep the build going until it opened in stages.
Preconditions
A state able to keep financing a megaproject through a doubling of cost and the 1997 financial crisis
A dedicated construction authority and imported core technology (TGV) with a transfer agreement
A corridor dense enough that demand justified finishing the line despite the overrun
Where else it has been used: Later Korean lines (Honam, Suseo to Pyeongtaek) were built by the same authority with domestic rolling stock; the Gyeongbu record is the standard example of optimism bias in a first-of-a-kind national project.
Limits: The 108% increase is computed from two published figures, and it compares a 1993 re-estimate with the 1990 plan, not the final outturn with the plan; the price basis of the re-estimate is not stated, so part of the gap may be inflation. The final outturn cost was not taken from an opened source. The 1993 plan cut scope (underground Seoul, Daejeon and Daegu sections replaced with upgraded existing alignments) to save 1.43 trillion won, so the delivered line is not the planned one, and phase one ran on electrified conventional track south of Daegu. The archive commentary is written by an outside researcher; the figures are read from a scanned-document PDF. The record documents overrun on one megaproject, not Korea's general project delivery performance.
๐ต๐ชPerusince 2015dismantledbears on Large project delivery
Peru's Gasoducto Sur Peruano concession, signed in July 2014 for a US$7.328 billion gas pipeline due in commercial operation by March 2019, stood at 37.32% overall progress against 76.49% scheduled in December 2016, according to the energy regulator Osinergmin, and the Ministry of Energy and Mines terminated the concession in January 2017 after the consortium failed to reach financial close.
Peru's energy ministry awarded a single private consortium a design-build-finance-operate concession for a strategic southern gas pipeline, backed by a guaranteed-revenue charge on electricity bills, and then terminated it when the consortium could not close financing; the regulator was left administering the half-built assets.
Preconditions
A concession framework that ties construction milestones to financial close and lets the grantor terminate cleanly
Bidders whose financing does not collapse when a sponsor is caught in a corruption case
A fallback plan, with budget, for custody and re-tendering of stranded assets
Where else it has been used: Large energy concessions with guaranteed revenue travel across Latin America; the GSP case is the cautionary version, where a single sponsor's legal collapse took the financing and the schedule with it.
Limits: Both progress figures are the concessionaire's own report as relayed by the regulator (the table cites 'Fuente: Reporte de Concesionario Gasoducto Sur Peruano SA'); Osinergmin supervised but did not independently measure overall progress. The figure is a weighted composite of permits, engineering, logistics and construction, so 37% overstates what was built: construction and assembly stood at 10.37% against 63.78% planned. The record does not show the final cost to the state, the arbitration outcome, the value of stranded assets, or how far the collapse stemmed from the Odebrecht corruption case and lost bank financing rather than from Peruvian institutions' project management. It is a concession (design, finance, build, operate), so the schedule failure is shared between the grantor and a private consortium. The numbers sit inside PDFs, not a statistical API. Later governments have repeatedly proposed reviving the pipeline under new names; this record concerns the 2014 concession, which ended by decision.
๐ต๐ชPerusince 2014still operatingbears on Large project delivery
Petroperรบ's Talara Refinery Modernisation Project, begun with the May 2014 EPC contract, reached full operation in 2023 after Peru's Contralorรญa found its estimated investment had risen from US$1.3 billion to more than US$5 billion by October 2020, about 400%, and a cumulative schedule slippage of more than three years in the project's execution schedule, traced to the auxiliary-units contracting.
Peru legislated a state-owned refiner's modernisation, let the company contract a single EPC builder and then improvise the auxiliary works, and the Contralorรญa's audit traced repeated procurement reversals to a schedule slip of over three years and a several-fold cost rise before the plant reached full operation.
Preconditions
A state enterprise with procurement rules and project-management capacity matched to a multi-billion dollar build
A cost baseline updated before the investment decision, so scope changes are not read as overrun after the fact
An audit office able to review a megaproject while it is still under construction
Where else it has been used: National-oil-company refinery megaprojects (Dos Bocas in Mexico, Abreu e Lima in Brazil) show the same pattern of baseline estimates that bear little relation to final cost; Talara is the Peruvian instance with an audit trail.
Limits: The 400% figure is a conditional statement ('estarรญa significando') attributed to the Comptroller General in a press release, comparing a feasibility-study estimate with an October 2020 cost, not an audited final cost; part of the rise reflects scope changes (Flexicoking, auxiliary units) rather than pure overrun, and the press release does not separate them. The press release's own figures (US$1.3bn to over US$5bn) imply an increase of about 285%, or roughly 3.85 times the estimate, so the stated 400% is the Comptroller's rounding and not the arithmetic; the metric keeps 400 because it is the published figure. The three-year slippage is the audit's figure for the auxiliary-units component over its 2014 to 2018 review window, not a measured end-to-end delay. Petroperรบ is a state-owned company, not a ministry: the record judges the state as project owner, and Petroperรบ's subsequent financial distress and government bailouts are not shown here. 'Operating' rests on a ministry press release declaring 100% operability in September 2023, not on independent utilisation data. The evidence is of a delivery that was completed late and far over cost, not a success.
๐ฎ๐ฉIndonesiasince 2015still operatingbears on Large project delivery
Indonesia's Ministry of Energy and Mineral Resources launched the 35,000 MW generation programme in May 2015 to be completed by 2019; by December 2022, 415 units with 16,596 MW, 47 percent of the programme, had reached commercial operation.
16,596 MW (47 percent of 35,000 MW), capacity at commercial operation, share of the 35,000 mw programme, 2022-12. 35,000 MW, original programme target completed by 2019, 2019 target, restated by the minister on 2018-03-06. Kementerian Energi dan Sumber Daya Mineral, press release 040.Pers/04/SJI/2023, retrieved 2026-10-01.
The move, and what it needed
The Ministry of Energy and Mineral Resources set a single national capacity target, signed PPAs and EPC letters of intent with private producers and PLN at launch, and issued pricing and licensing rules to pull IPP finance into it; when demand undershot, the minister rephased the target instead of holding the deadline.
Preconditions
A single-buyer utility (PLN) able to sign long-term power purchase agreements that lenders accept
Land acquisition and one-stop licensing rules that can move dozens of plant sites in parallel
Demand forecasts honest enough that a capacity target does not lock in surplus take-or-pay contracts
Where else it has been used: Headline megawatt targets travel easily between governments; the Indonesian case shows that the target is the cheap part, and that a rephasing announced as on target can hide a schedule miss unless the original deadline stays on the record.
Limits: This is a schedule record, not a cost record: no official source located gives the programme's budgeted against actual cost, so the cost half of the construct is missing. The 47 percent is the ministry reporting on its own flagship. It is a capacity count at commercial operation, not a measure of whether plants run, at what availability, or at what tariff burden on PLN. The target moved: launched as a need for 35,000 MW over 2015 to 2019, it was restated in March 2018 as about 20,000 MW by 2019 with the rest in 2024 to 2025, on the argument that demand grew more slowly than the 7 percent growth assumption. Even the restated 2019 milestone had not been reached by December 2022. The ministry attributes the shortfall partly to over-forecast demand, so part of the slip may be a deliberate deferral rather than a delivery failure, and a reviewer can read the low completion as prudence. Most capacity (about 70 percent of units) is built by private IPPs under PLN purchase agreements, so delivery is joint and not attributable to the state alone. The latest official progress figure found is December 2022; nothing later was verified, and the share may be higher now.
๐ง๐ดBoliviasince 2016operating below its peakbears on Large project delivery
Bolivia's Empresa Siderรบrgica del Mutรบn, whose turnkey contract with Sinosteel was awarded in 2016 and whose complex the President inaugurated in February 2025, reported in its 2026 initial accountability review that the project stood at 92% physical progress against 100% accumulated financial execution, that its own supervision found the complex had never entered operation, and that it had served notice of intent to terminate the contract in January 2026.
Bolivia's Ministry of Mining had its state steel company sign a single turnkey design-build-operate contract with a Chinese state contractor, financed 85% by a China Eximbank credit, and accepted provisional and final receptions of each plant before the complex had run as a whole; the next administration's site supervision rejected those receptions.
Preconditions
A tied export credit that pays out against contractor milestones rather than against verified operation
A state owner company with thin in-house engineering supervision relative to the contractor
A political calendar that rewarded an inauguration date over a commissioning test
Where else it has been used: The lithium carbonate plant (bol-lithium-plant) shows the same state-led industrialisation model delivering a plant that runs well below design capacity.
Limits: The source is a slide presentation, not an audit: the figures are read off slides whose PDF text extracts garbled, and a reviewer will need to open slide 11 to check them. The review was written by the company's management under the government that took office in late 2025, which has an interest in blaming the previous administration and the contractor; Sinosteel answered the termination notice on 11 February 2026 and the dispute is unresolved, and the external audit and Contralorรญa opinion the review schedules had not been published at retrieval. The 100% financial execution covers the EXIMBANK and Treasury tranche only; an additional FINPRO credit stood at 55.79%. No official baseline schedule was opened, so the record carries no percentage schedule overrun, and the 2020 suspension under the interim government is part of the delay. The contract amendment from USD 422.6 million to USD 443.98 million on slide 13 is legible only as chart labels and is not used as the metric. The status 'eroded' records the withdrawal of a completion the state had declared in 2025, not the loss of an operating peak, because the plant never produced at scale; a reviewer may prefer to read it as a failed delivery.
๐ป๐ณVietnamsince 2024still operatingbears on Large project delivery
Vietnam built the 519-kilometre Quang Trach to Pho Noi 500kV circuit-3 transmission line across nine provinces after construction began on 18 January 2024, and the Government reported it in operation on 29 August 2024 'after six months of construction', meeting the Prime Minister's deadline of completion before September, at an investment of over VND22.3 trillion.
Vietnam's Prime Minister made the line a dated national priority with his own progress reviews, Vietnam Electricity's transmission corporation split it into four component projects built in parallel, and the nine provinces on the route were held responsible for site clearance at the same time as construction.
Preconditions
A state-owned grid company able to mobilise contractors and equipment across four sections at once
Provincial governments that answer to the centre and can clear land on a central deadline
An acute supply crisis, the 2023 northern power cuts, that justified bypassing normal sequencing
Limits: The duration is the Government's own framing: 18 January to 27 to 29 August 2024 is over seven months of construction, not six, and design, investment approval and land clearance before January are not counted. The deadline it met was set during construction by the Prime Minister, so 'on schedule' measures against a political target, not an original baseline, and no approved budget was published to compare with the outturn: the Government's own articles give the investment as VND22 trillion on 14 August and over VND22.3 trillion on 29 August. A single line built under emergency priority after the 2023 northern power shortages shows what the state can do when it concentrates resources; it does not show routine project performance. Construction quality, safety, compensation paid to displaced households and the line's actual utilisation are not in the source, and the outlet is the Government's news portal, not an audit.
๐ช๐จEcuadorsince 2009still operatingbears on Large project delivery
Ecuador contracted the 1,500 MW Coca Codo Sinclair plant in October 2009 for USD 1,979.7 million with a 60-month term ending 5 October 2014; the plant entered commercial operation in November 2016, and a National Assembly oversight report found it had cost USD 2,712.5 million by August 2021, about 37 percent above the contract price, excluding a USD 599 million transmission line.
Ecuador's government declared the plant a national priority, signed a turnkey EPC contract with Sinohydro financed by China Eximbank, and let the contractor set implementation costs, which delivered a 1,500 MW plant two years late and well above the contract price.
Preconditions
Bilateral resource-backed finance that made a large single project fundable quickly
A political decision to size the plant at 1,500 MW before definitive engineering studies existed
Weak owner-side engineering and supervision capacity relative to an open EPC contract
Where else it has been used: The turnkey, lender-tied model was repeated across Ecuador's other emblematic hydro plants and in other Chinese-financed projects in the region; the Coca Codo record is the case most cited when that model's owner-side risks are discussed.
Limits: The 37 percent overrun and the roughly 25-month delay are our arithmetic on two figures in the report, not a figure the report states; the report itself compares against the USD 987 million planning estimate, which would make the overrun far larger, so the baseline choice matters. The source is a parliamentary oversight committee report produced in a politically adversarial investigation, written partly as narrative with annexes, inside a 346-page PDF; it is not an audited cost statement, and the 'USD 2.712.490.418,04 millones' wording mixes units. The report also documents thousands of cracks in the turbine distributors, regressive erosion of the Coca river threatening the intake, and net output of 37,927 GWh from April 2016 to May 2022 against about 62,934 GWh expected, so the plant operates well below design; none of that is in the cost metric. The construction contract was signed before financing was firm and outside the public procurement law. Started is the contract signature year, not the start of physical works.
๐ต๐พParaguaysince 2020delivered and closedbears on Large project delivery
Paraguay's Ministry of Public Works began building the Hรฉroes del Chaco cable-stayed bridge, the first road link from Asunciรณn to the Chaco, in 2020 on an estimated 36-month schedule; at the end of 2023 it reported 90% physical progress against 100% planned, and the bridge was inaugurated on 3 March 2024 and recorded as concluded with provisional acceptance in the 2024 balance.
MOPC tendered the bridge through national public tender MOPC No 156/2019, awarded it to Consorcio Union, financed it from the general budget, and tracked planned against physical progress each year in its public balance.
Preconditions
A ministry able to finance a multi-year river crossing without a multilateral lender's supervision
Domestic contractors with enough balance sheet and imported expertise to take on a cable-stayed span
An annual public reporting format that records planned against actual progress, so slippage is visible before inauguration
Where else it has been used: The transferable part is the plain planned-versus-actual table in the annual balance, which lets a reader see both the early lead and the late slip.
Limits: All figures are the ministry's own reports; no audit-office or independent value-for-money review was opened. The schedule figure is a progress-against-plan percentage in a year-end table, which says the works were behind their own plan at end-2023 but not by how many months; the official sources opened give the construction start only as a year, so a month-precise overrun cannot be computed (a 2020 start plus 36 months gives completion within 2023, against an inauguration on 3 March 2024, i.e. at least two months late, and the 2021 balance showed the works ahead of plan, 43.25% against 32.30%). No total contract cost, original award value or cost overrun appears in the balances opened: the 2024 table carries only that year's spending (G. 98,099,704,513). Press reports describing the bridge as delayed and more expensive were leads only and are not the source.
๐ฉ๐ชGermanysince 2022still operatingbears on Large project delivery
After Russia's invasion of Ukraine the federal government chartered floating storage and regasification units, commissioned the first at Wilhelmshaven in December 2022, and by February 2023 expected to spend about 10 billion euros renting and running five units, against 2.94 billion euros of federal funding made available for four in May 2022.
The economy ministry chartered ready-built floating terminals through the utilities RWE and Uniper instead of building onshore plants, and Parliament passed an acceleration act letting authorities skip environmental assessment and relax procurement for these projects.
Preconditions
A global charter market with floating regasification ships available at short notice
A shock severe enough for Parliament to suspend permitting and procurement rules
A federal budget able to commit billions before the full cost was known
Where else it has been used: Finland, Italy and the Netherlands also chartered floating units in 2022. The German case is the one cited for permitting speed, and also for the overcapacity and cost debate that followed.
Limits: The two cost figures are not like for like and must not be read as an overrun percentage: the May 2022 figure was funding made available for four units, the February 2023 figure is an expected spend for five units over charter terms the government puts at five to 15 years, and the government says a large part will flow back through landing fees. The charter contracts are classified (VS-Vertraulich), so no outside reader can check either number. The schedule claim, decision in late February 2022 to commissioning in December 2022, is computed from two documents rather than published as one figure, and it covers only the first unit; the sources verified here give no commissioning dates for the others, and plans for Stade and a second Wilhelmshaven unit were only targets for end 2023. Part of the speed was bought by suspending process: the government's answer confirms that under the LNG Acceleration Act no environmental impact assessment was carried out at Wilhelmshaven or Brunsbรผttel. Nothing here shows how fully the terminals have been used, which is the main criticism of the build-out.
๐ฆ๐ทArgentinasince 1981still operatingbears on Large project delivery
Argentina's state nuclear operator connected Atucha II to the grid in June 2014, almost 33 years after construction began in July 1981, and began commercial operation in May 2016, after a build halted in 1994 and restarted in 2006.
The state nuclear company NA-SA and the atomic energy commission CNEA restarted a German-designed reactor abandoned mid-build, under a 2006 national reactivation plan, and finished it with Argentine engineering after the original vendor's nuclear business had left the field.
Preconditions
A standing nuclear institution (CNEA) that kept engineers and know-how through the suspension
State finance able to fund a restart once the fiscal crisis passed
A design close enough to Atucha I that domestic teams could complete and license it
Where else it has been used: Restarting a long-suspended nuclear build has rarely worked elsewhere; Brazil's Angra 3 is the comparable case and remains unfinished, which is why Atucha II is cited as proof of completion rather than of efficient delivery.
Limits: The 32.9 years is computed from two month-precision dates in an IAEA table, and the national regulator dates the start to 1982, so the figure could be a year shorter. No cost figure was found at an official source, so the record carries schedule only, and the indicator asks for cost too. Most of the delay is a 12-year suspension (1994 to 2006) caused by fiscal crisis and policy, not slow building, so the number mixes a stopped project with a slow one. The IAEA data are reported to the IAEA by the national authorities. The plant's energy availability factor of about 53 percent over 2015 to 2025 shows that completion did not bring reliable output, and the record should not be read as a success story beyond the fact of delivery.
๐ต๐ฑPolandsince 2019still operatingbears on Large project delivery
Poland's Maritime Office in Gdynia signed the first construction contract for the Vistula Spit waterway on 4 October 2019 and opened the canal on 17 September 2022. The Supreme Audit Office found that by April 2023 the programme's value had risen 125 percent, from PLN 880 million to over PLN 1,984 million, and that its completion deadline had slipped two years to the end of 2024.
The Council of Ministers established a multi-year programme and the Maritime Office in Gdynia contracted and built it. The minister responsible raised the budget twice and extended the deadline without telling the cabinet that the project no longer passed its own cost-benefit test. A strategic objective, sea access that bypassed Russian-controlled waters, overrode the economic appraisal.
Preconditions
A strategic or security rationale strong enough to survive a failed cost-benefit test
A multi-year programme instrument whose budget the responsible minister can revise
An audit office with the mandate to publish findings after the fact and to notify suspected crime and breaches of public finance discipline
Where else it has been used: The pattern of a politically fixed opening date and a budget revised to fit is common to prestige infrastructure. What sets this case apart is NIK's documentation that the threshold breach was known before the first contract was signed.
Limits: This is a delivery with a large documented overrun, and it should be read as evidence of poor cost and schedule performance, not as a success. The 125 percent is NIK's figure for the programme budget as revised up to April 2023. Against the further revision to PLN 2,127.5 million in May 2023 the overrun is about 142 percent, a figure computed here and not published by NIK. Final outturn cost after 2024 was not checked. NIK found that the cost-benefit threshold was crossed at a 20 percent cost increase, and that the minister knew of a more than 100 percent rise before the first contract was signed. The project is therefore uneconomic on the state's own appraisal, and NIK said it was considering notifications of suspected crime and of breaches of public finance discipline. Only the canal and lock are open. NIK states that the sea-going vessels the programme was designed for cannot call at Elblฤ g until port works there are finished, so the canal's stated purpose was not delivered on the date the first metric describes. The ministry's count of more than 1,770 vessels in the first year is mostly small craft and is not used. The programme was set up by a Council of Ministers resolution of 24 May 2016. 'Started' uses the first construction contract in 2019, and a reviewer could set it at 2016.
๐ฎ๐ฑIsraelsince 2000still operatingbears on Large project delivery
Israel Railways began electrifying the national network in 2000 and opened the electric express line to Jerusalem in September 2018, but by May 2020 only about 64 of 480 planned route kilometres were electrified, and the forecast completion date had moved from 2019, set in 2013, to the end of 2025, on an estimated cost of ILS 12.1 billion.
A state-owned rail company ran a network-wide electrification programme alongside a new flagship line and opened the line first, diverting cars from other lines to do it. Completion dates were reset repeatedly while the State Comptroller ran follow-up audits.
Preconditions
One state company carrying planning, procurement and operations, so a slip in one stream delays all of them
Repeated external audit that publishes the forecast dates, which is what makes the slippage visible
An operator willing to open a line before it had its own rolling stock, by diverting cars from other lines
Limits: This is a delivery still in progress, not a finished one. At the audit date it had reached 13.3 percent of the planned route, and no official source that opened on the retrieval date gave a later figure, so whether the 2025 date held is not shown. The audit gives an estimated cost (ILS 12.1 billion) but no original budget, so the cost overrun cannot be computed from this source; the schedule slip from 2019 to 2025 is the only performance figure. The completion dates are Israel Railways' own forecasts, which the Comptroller reports. The audit also found that the Jerusalem line opened without its own rolling stock and ran with passengers in test mode for about a year, cancelling 2.6 percent of trips against 0.5 percent on other lines. A hostile reviewer would say the record shows a slipping project, not a delivered one.
๐จ๐ฆCanadasince 2015delivered and closedbears on Large project delivery
Infrastructure Canada delivered the Samuel De Champlain Bridge corridor in Montreal as a public-private partnership signed in 2015, opening the bridge in July 2019 against a contractual in-service date of December 1, 2018, at a total project cost the department reports as C$4.2 billion, after a C$235 million settlement in 2018 that paid for delay-recovery measures and oversized-part transport issues.
Infrastructure Canada replaced a failing federal bridge through a design-build-finance-operate-maintain contract with liquidated damages for late opening, then bought back schedule with a negotiated settlement when transport restrictions and disputes threatened the date.
Preconditions
A federal asset owner with an urgent safety case (the old bridge at the end of its safe life) that justified a fixed in-service date.
A mature domestic P3 market and an owner's engineer able to price and monitor a 30-year contract.
Budget room to pay a nine-figure settlement rather than litigate through the delay.
Limits: No published figure states the outturn against the baseline, so overrun has to be pieced together by the reader. The value-for-money report gives a 2015 total of C$4.239 billion and a contractual in-service date of December 1, 2018; the department's own story page gives a rounded C$4.2 billion total and a July 2019 opening; the 2018 settlement adds C$235 million and moved the delivery date to December 21, 2018, which was itself missed. The C$4.2 billion therefore cannot include the settlement and appears to restate the original envelope, so the true outturn is at least about 5.5% above it (a computed figure, not published) and the bridge opened about seven months after the contract date. Costs also bundle a 30-year operation and maintenance period and financing, so the number is not a construction cost. All figures are the department's own; the Auditor General's 2018 doubts about the schedule are reported only in the press here. The canada.ca news release refused a scripted curl request and was read through a browser-like fetch. This is one megaproject, not a national portfolio.
๐ฏ๐ตJapansince 2012delivered and closedbears on Large project delivery
Japan's Railway Construction, Transport and Technology Agency built the 125 km Kanazawa to Tsuruga extension of the Hokuriku Shinkansen from 2012; its authorised cost rose from 1.1858 trillion yen in 2017 to 1.6779 trillion yen after a second increase of 265.8 billion yen in March 2021, completion slipped one year to the end of fiscal 2023, and the line opened on 16 March 2024.
16,779 100 million yen (oku-yen), authorised project cost after the march 2021 revision, 2021-03. 2,658 100 million yen (oku-yen), cost increase approved in the march 2021 revision, 2021-03. Board of Audit of Japan, fiscal 2020 audit report, chapter 4 item 11, retrieved 2026-10-01.
The move, and what it needed
The transport ministry authorised a state construction agency to build the line on a politically accelerated deadline; when the agency disclosed a second overrun late, the ministry set up an external verification committee, downgraded the agency's performance rating and issued a statutory improvement order, and the Board of Audit examined the causes in its annual report.
Preconditions
A specialised state construction agency with decades of high-speed rail experience
Independent audit and ministerial evaluation that can publish an overrun against a named project
Funding arrangements (lease fees, existing-line sale proceeds, local contributions) able to absorb an increase without stopping the works
Where else it has been used: Japan's verification committee and audit report resemble the UK's National Audit Office reviews of HS2; the difference is that the Japanese line opened within one year of its revised date while HS2 was cut back.
Limits: This is a delivery with a large overrun, filed because the construct asks for cost and schedule performance, not as a success. The overrun against the 2017 authorised cost is about 41 percent (1.6779 against 1.1858 trillion yen), a figure computed here, not published; against the 2012 plan it cannot be computed from this source. The final outturn cost after the March 2021 revision is not in the audit report, and the opening date comes from the delivery agency's own page. The one-year slip is measured against a deadline that the government had itself pulled three years forward in 2015, so part of the 'delay' is a political acceleration that did not hold. Some of the increase reflects new seismic standards and a consumption-tax rise, which an auditor might count as scope or price change rather than poor management; the audit office and the ministry's own verification committee nonetheless found weak reporting from site to headquarters and inadequate ministry oversight, and the ministry issued a formal business-improvement order. One project is not a national track record.
๐ธ๐ชSwedensince 1999still operatingbears on Large project delivery
Sweden built the Botniabanan railway between Nyland on the ร ngermanรคlven and Umeรฅ through the state and municipally owned company Botniabanan AB, starting construction in 1999 and opening it in August 2010, four years after the original 2006 schedule, at a final cost of almost SEK 16 billion in 2010 prices, which the national audit office calculated as 73 percent above the original estimate.
Parliament chose to build the line through a dedicated company, Botniabanan AB, owned by the state and the municipalities along the route and loan-financed so it could build quickly, while Banverket kept the physical planning; the state then rents the line back from the company.
Preconditions
Regional and municipal co-owners willing to share financing and risk
A planning and environmental permitting process that can be forecast, which is where the schedule broke
An early estimate that covered the connecting lines the project needed to work, which the audit found was missing
Limits: The 73 percent is Riksrevisionen's own recalculation in constant 2010 prices; the government's communication in the same document gives SEK 15 billion in current prices, about 3 billion more than at the 1997 decision, so the size of the overrun depends on the price basis chosen. The 140 percent counts follow-on investments on ร dalsbanan and other connecting lines that the audit office judged necessary, a boundary Trafikverket disputed. The four-year delay is computed from two sentences, not published as a figure, and much of it came from environmental permitting under the 1999 Environmental Code rather than construction. The start year rests on a 1999/2000 parliamentary motion saying construction had just begun, because the audit itself dates only the 1997 decision. One railway line is a single project, so this is not a portfolio-level overrun rate, and the audit predates a decade of operation, so whether ridership and freight benefits arrived is not shown.
The Department for Transport and HS2 Ltd began building High Speed Two in 2020. The government cancelled everything beyond Birmingham in 2023. By May 2026 the London to Birmingham railway alone was estimated at ยฃ87.7 billion to ยฃ102.7 billion, more than double the 2020 Phase 1 estimate, with services from Old Oak Common expected between 2036 and 2039 against a 2012 plan to complete the full network by 2033.
100 % overrun (more than, excluding inflation), cost increase over the 2020 phase 1 estimate, low end of the may 2026 range, 2026-05. 2033 year, year dft expected the full london, manchester and leeds network to be complete at the 2012 approval, 2012. National Audit Office (UK), High Speed Two reset, HC 52, Session 2026-27, 29 June 2026, retrieved 2026-10-01.
The move, and what it needed
The Department for Transport approved a national high-speed line, set up an arm's-length company to build it, and committed to construction contracts in 2020 before designs and costs were stable. It then responded to the cost growth by cutting the northern phases instead of the scope of Phase 1.
Preconditions
A sponsoring department able to commit tens of billions over decades without a firm cost baseline
Construction contracts let before a reliable total cost estimate existed, which HS2 Ltd began renegotiating in 2025 to better incentivise cost-effective delivery
Repeated political resets in 2020, 2023 and 2025 that changed scope faster than the client organisation could re-estimate
Where else it has been used: The NAO's 2026 report records the government's 2025 changes to how it manages mega-projects and a separate Euston delivery company set up in April 2026, both framed as lessons from HS2.
Limits: No HS2 train has run. This is a record of a build in progress, not a delivered railway, and the erosion is to planned scope: the Manchester and Leeds legs were cancelled before they were built, so nothing in service was lost. The May 2026 estimate is in a mixed price base and the NAO says it carries a high level of uncertainty until a baseline is set in spring 2027. The over 100% figure compares a London to Birmingham estimate with the 2020 Phase 1 estimate, which excluded inflation and was itself a rebaseline after the 2019 Oakervee review, so the overrun against the 2012 approval is larger and is not stated as one number by the NAO. The 2033 completion year is for a network that no longer exists, so it is a schedule reference, not a like-for-like comparison. Both figures are read from PDFs, with text-extraction spacing artefacts cleaned in the quotes. Spend to March 2026 was ยฃ46.8 billion in nominal prices, including the cancelled Phase 2.
The Department for Transport and Transport for London built Crossrail through Crossrail Ltd from 2009. The NAO put its forecast cost at 28% above the 2010 budget of ยฃ14.8 billion and its opening at up to three and a half years past the original December 2018 date. The central section opened in 2022 and carried 257.4 million journeys between April 2025 and March 2026.
The Department for Transport and Transport for London set up a dedicated delivery company, Crossrail Ltd, with private delivery partners, and when it missed its 2018 opening they replaced its leadership, added ยฃ825 million and more, and handed the finished asset to Transport for London to operate.
Preconditions
Joint sponsorship by a national department and a city transport authority with its own revenue base
Dedicated local funding streams, including a business rate supplement, alongside central grant
An operator ready to take over and run the line once commissioned
Limits: The 28% is the NAO's May 2021 forecast, made before the line opened; this record has not verified the final outturn cost from an audit source, and the NAO itself notes that later funding packages kept moving. The schedule figure is also a 2021 forecast of 'at least 3 years' late; the opening in 2022 is confirmed only by the regulator's statement that the central section opened that year. The cost is nominal, so part of the 28% is inflation over a decade. Crossrail is a London line co-sponsored and part funded by Transport for London and the Greater London Authority, including a business rate supplement, so it is not purely a national delivery, and its success in carrying passengers says nothing about the governance failures the NAO documented in 2018 to 2019. The journey count is a regulator's estimate from ticketing data.
๐ฎ๐ณIndiasince 2008still operatingbears on Large project delivery
Indian Railways, through the Dedicated Freight Corridor Corporation, built two freight-only corridors approved by Cabinet in 2008 at โน28,181 crore and reassessed in 2015 at โน81,459 crore, and by March 2025 had commissioned 2,741 of their 2,843 route kilometres.
The Ministry of Railways created a dedicated special-purpose company in 2006 to build freight-only lines alongside the trunk network, financed largely by World Bank and JICA loans, and kept building through large cost revisions until almost the whole length was commissioned.
Preconditions
A state railway able to set up and capitalise a separate construction company
Multilateral and bilateral lenders willing to finance most of the cost on long tenors
State governments acquiring land along long linear routes, which the CAG found was a main source of delay
Limits: The 189% is computed here from two CAG figures; the CAG does not publish it as a percentage. The two estimates are not like for like: the scope changed between them, route lengths were revised upward, the Sonnagar to Dankuni section was added in 2008 and is not part of the corridors now reported as 2,843 km, and both figures are nominal across seven years. The 2015 figure is itself old, and no later audited completion cost is used here. The CAG found avoidable price escalation of โน2,233.81 crore from delays, contract awards 13 to 25 months late, and a target of March 2017 for the first Eastern section, so the schedule slipped by years, but the report states no single schedule overrun figure. The commissioning figure is a ministry press release, self-reported and unaudited. The PIB page refused an automated fetch tool (HTTP 403) but opened with a standard browser user agent.
๐ฒ๐ฝMexicosince 2019still operatingbears on Large project delivery
Pemex built the Olmeca refinery at Dos Bocas, authorized in December 2018 and fully operational from February 2025, under an authorized budget of US$15,963.0 million, 79 percent above the US$8,918.5 million Class IV estimate stated when its board authorized the FEL II phase in July 2020.
The federal government chose a greenfield refinery as a political flagship and had Pemex deliver it through a dedicated subsidiary, PTI Infraestructura de Desarrollo, funded through equity contributions from Pemex Industrial Transformation (2020 to 2023). The board re-authorized the scope and budget mid-construction in August 2022.
Preconditions
A state oil company able to absorb repeated capital contributions outside the normal investment appraisal
Political commitment strong enough to keep funding through cost rebaselining
A site and licensor package secured before front-end engineering was complete
Limits: Pemex is the state oil company reporting on its own project. The filing is made under US securities law and carries legal liability, but it is not an audit, and Mexico's audit office (ASF) is not reachable from outside Mexico to check against. The publisher is a state enterprise, not a ministry or statistics office, so a strict reading of the publisher rule could exclude it. The 79 percent overrun is our computation from two figures that do not measure the same thing. The Class IV estimate excludes external works, escalation, contingency and administrative costs and carries a stated range of +30 to -25 percent, while the later budget includes VAT. A share of the gap is therefore a change of definition and not an overrun. The government's original public figure, around US$8 billion, appears only in press reports and is not used here. The schedule is shown only by the issuer's own changing expectations. The FY2025 filing (https://www.sec.gov/Archives/edgar/data/0000932782/000093278226000006/ck0000932782-20251231.htm) says the refinery "has been fully operational since February 2025". No filing states an original completion date to measure the slip against. Full operation is a claim about start-up, not about utilisation or margins, which the FY2025 filing excludes from its refining margin series.
๐ฆ๐บAustraliasince 2018operating below its peakbears on Large project delivery
The Australian Government began building Inland Rail, a Melbourne to Brisbane freight railway, in 2018 through the Australian Rail Track Corporation, which passed delivery to its subsidiary Inland Rail Pty Ltd in May 2024; a verification report delivered to the infrastructure department in November 2025 reported the company's estimate at completion of A$45.6 billion (P90, Target scenario, estimate current as of December 2024) with completions finalised in May 2036, and the 2026-27 Budget then consolidated the project at Parkes and returned $4.4 billion in equity.
The Commonwealth committed to a 1,700 km freight line through a government-owned track company and started construction before the route ends, approvals and designs were settled; when successive reviews put the cost far above the budget, it moved delivery into a dedicated subsidiary, commissioned independent cost assurance and then cut the line back to the sections already under way.
Preconditions
A government-owned rail company able to take Commonwealth equity and build without a separate procurement for each section
Construction started on sections whose environmental approvals and designs were immature
A budget process willing to recognise a reversal and return equity rather than keep funding the full scope
Where else it has been used: The pattern matches gbr-hs2 in this corpus: a national line cut back to its first sections after its cost estimate tripled, with the corridor preserved for later.
Limits: Both numbers are forecasts, not outturn: they are Inland Rail Pty Ltd's own December 2024 estimate as reviewed by a consultancy (ACIL Allen) for the department, and the report itself says IRPL's core assumptions have not been met and the cost is likely to increase with further delay. The faster Stretch scenario is A$44.7 billion with completions by June 2035. Both figures are in Table 1.1 on PDF page 12; the 2018 start and the 2015 business case are on page 8. The truncation is not in this report: it comes from Budget Paper No. 2, 2026-27 (https://budget.gov.au/content/bp2/download/bp2_2026-27.pdf, PDF page 139, printed page 124), which says the government will 'consolidate the Inland Rail project at Parkes and preserve the remaining rail corridor and intermodal sites for future investment, with $4.4 billion in equity to be returned to the Budget'. The 2015 business case cost and the estimates in the 2023 Schott review are published on infrastructure.gov.au, which did not answer from the machine used for this record, so no original baseline is given here and the overrun against it is not stated. The erosion is to planned scope: Beveridge to Parkes is still being built and nothing in service was lost, and no section of the full line has opened to double-stacked through traffic.
๐ซ๐ฎFinlandsince 2009delivered and closedbears on Large project delivery
The Finnish state, with the City of Vantaa paying part of the cost, began building the Kehรคrata (Ring Rail Line) through Helsinki Airport in Vantaa in 2009; construction costs came in above estimate, a glycol leak found at the airport took a year to resolve and reset the schedule, passenger service began on 1 July 2015, and the construction cost finally came to 770 million euros.
The state rail administration and the City of Vantaa cofunded an airport loop tied to new housing zoned along the line; when the cost forecast rose before construction, the ministry renegotiated the cost split and kept the start date, then absorbed a further overrun from ground contamination at the airport.
Preconditions
A municipality willing to share capital cost in exchange for land-use value along the line
A state budget able to absorb an overrun on a single project without stopping it
Airport ground conditions that were not fully surveyed before tunnelling
Limits: The 770 million euro figure is stated by the delivery agency in an anniversary news article ('Hankkeen rakennuskustannukset olivat lopulta 770 miljoonaa euroa'), not taken from an audited final account, and the page flags itself as an old article. The page says only that costs were higher than estimated; the size of the overrun comes from a different source: the Ministry of Transport and Communications release of 10 March 2008 (https://lvm.fi/-/keharadan-kustannukset-ylittyvat-770641) put the forecast at 590 million euros, 75 million more than the estimate before it, so the final cost is about 30 percent above the 2008 forecast and about 50 percent above the earlier one. Those percentages are computed here, and neither source states the price basis, so part of the gap is inflation over 2008 to 2015. The originally planned opening date is not on either page, so the schedule slip is given only as the one year the agency attributes to the glycol problem. The state and the City of Vantaa shared the cost (73 and 27 percent under the original agreement, per the 2008 release). The same article says population growth along the line has been slower than forecast, especially in Kivistรถ, and that not all of the line's targets had been met by 2020. One line is not a national delivery record.
๐ฌ๐นGuatemalasince 1977delivered and closedbears on Large project delivery
Guatemala's state electricity institute INDE built the 300 MW Chixoy hydroelectric project under contracts signed in 1977 and 1978; after the power-tunnel lining collapsed in late 1983 it entered commercial operation in November 1985, 41 months later than originally planned, at a final cost of US$823.1 million excluding finance charges against an appraisal estimate of US$372.7 million, an increase of 120 percent.
The state utility chose a single-stage scheme with a 26 km power tunnel through poorly understood geology, contracted the works on winning bids before the seismic redesign that followed the 1976 earthquake was settled, and financed the doubling of cost with supplementary loans and local funds.
Preconditions
A state utility with a monopoly on generation and access to multilateral loans
Cost estimates built from bid prices with contingencies sized for ordinary geology
Lenders willing to add supplementary financing rather than stop a half-built dam
Limits: The final cost is INDE's own figure as reported in the World Bank's completion report, which says INDE's data 'reflect shortcomings' and could not reconcile cost increases with physical quantities (para. 4.26, PDF page 32). Both costs and the 120 percent increase are in the table under para. 4.04 (PDF page 25), in para. 4.24 (PDF pages 31 to 32) and in Table 3 (PDF page 71); the 41-month delay against the original plan is in para. 4.04 (PDF page 24) and the summary's 'about 3.5 years' on PDF page 8; para. 3.18 (PDF page 19) records that the bids implied commercial operation by October 1981 while the appraisal team judged it unlikely before June 1982; the November 1985 start of operation is in para. 2.08 (PDF page 13). The text is OCR of a scanned 1991 document. The appraisal estimate is dated to the March 1977 appraisal; its exact price basis is not stated, and price increases account for US$141.1 million of the overrun by INDE's breakdown. The power tunnel's lining remained structurally unreliable after repair, so 'concluded' means built and operating, not sound. The project was co-financed by the World Bank, the IDB and others, though INDE owned and built it. The PCR records that the resettlement programme was judged flawed in conception; the 1980 to 1982 killings of Rรญo Negro residents displaced by the reservoir, documented elsewhere, are outside what this cost figure measures and are the larger failure of this project. This is a 1977 to 1985 build and says little about Guatemala's delivery capacity now.
๐ณ๐ฌNigeriasince 2018operating below its peakbears on Large project delivery
Nigeria's Federal Ministry of Works awarded the rehabilitation of the Abuja-Kaduna-Zaria-Kano dual carriageway to Julius Berger on 20 December 2017 at an initial N155.7 billion with a 36-month completion period; after repeated variations and augmentations, the Federal Executive Council cut the contract sum under Section I from N797.3 billion to N740.8 billion on 23 September 2024, and the ministry issued a 14-day termination notice on 4 November 2024 after the contractor rejected the reviewed sum and stopped work.
The federal government let one large contract for a 375 km corridor, paid for it through repeated variations and augmentations, and when a new minister demanded a lower fixed price the contractor refused and the ministry terminated, split the work among other contractors and took direct control.
Preconditions
Award and variation approvals routed through the Federal Executive Council rather than a fixed-price baseline
A naira devaluation and inflation shock that made earlier contract sums unenforceable
Political will to terminate a dominant contractor on a flagship route
Limits: The source is the ministry's own statement in a dispute with the contractor, so it carries the government's side. Figures are nominal naira across years of high inflation and a large devaluation, so most of the roughly fivefold growth may be price change rather than poor delivery; the ministry also re-scoped Section I to concrete pavement and gave 38 km of it to Dangote, so the 2024 sum buys a different road from the 2017 one. The termination release frames the N797.3 billion review as concerning 'Section I' while the N155.7 billion award covered all three sections; a ministry release of 7 October 2024 (https://www.fmw.gov.ng/index.php/read/3233) describes the same N740 billion as the total contract sum within Julius Berger's scope, with N391 billion already paid, so the two sums may or may not cover identical scope. What ended is the 2017 contract, not the road: Sections II and III were partly completed and handed over before 2023, and a ministry update dated 9 September 2026 (https://www.fmw.gov.ng/index.php/read/3425, also shown as a banner on the source page) says the government took over the project, finished 118 km of lot 1, has lot 2 over 60 percent complete and schedules completion for December 2026. Earlier ministry pages (https://fmw.gov.ng/index.php/read/2352) give a 36-month completion period from 21 May 2018, so the road is at least five years late; a 2018 ministry release (https://fmw.gov.ng/index.php/read/1299) states the award sum as N155.47 billion, slightly different from the N155.75 billion here. No audit or Bureau of Public Procurement document was opened.
๐ต๐ฐPakistansince 2008operating below its peakbears on Large project delivery
WAPDA's subsidiary NJHPC's 969 MW Neelum-Jhelum project, under construction from January 2008 and substantially completed in 2018, had cost Rs 423,446 million by June 2023 against Rs 84,502 million in the 2002 revised PC-I, nine years late; it peaked at 4,940.67 GWh in 2020 and was forced into complete shutdown by a head race tunnel failure on 2 April 2024 (the auditor counts 1,797.704 million units lost in FY2023-24), after an earlier tail race tunnel collapse in July 2022.
WAPDA started tunnelling in difficult Himalayan geology before design issues were settled, and the planning ministry re-approved the cost four times rather than stopping or re-scoping the project, so the bill rose with each revision and the tunnels later failed where weak rock had been flagged.
Preconditions
A state utility able to borrow and add a consumer surcharge to keep financing a project as costs rose
A project appraisal system that approves revised PC-Is after the fact instead of gating on final design
A strategic water-rights deadline with India that pushed construction to start before geology and design were resolved
Limits: This is a delivery that ran backwards twice: badly over cost and schedule while it was being built, then lost its output after it was finished. The overrun is measured against the 2002 revision, which is the auditor's choice of baseline; against the 1989 original (Rs 15,012 million) it would be far larger, and against the 2018 fourth revision (Rs 506,808 million) the project came in under. Part of the rupee overrun is exchange-rate depreciation and interest during construction, which a cost-performance reading should not charge entirely to delivery. The plant never reached its design output of 5,150 GWh even at its 2020 peak, partly because India's upstream Kishanganga project and a higher environmental flow reduced water, which management cites and the auditor rejects. The overrun percentage (about 400 percent) is our computation from two published figures, and the Rs 338,944 million is the auditor's own subtraction. The 2024 shutdown is documented in a second AGP report; the expected restart date circulating in the press is not from this source and is not claimed here. Management disputes several findings and never convened the Departmental Accounts Committee, so the report records the auditor's view without a settled reply.
๐ต๐ฐPakistansince 2012delivered and closedbears on Large project delivery
WAPDA built a 1,410 MW powerhouse on Tarbela Dam's fourth tunnel, effective April 2012 and commissioned in 2018; the World Bank recorded it as completed on time and below its estimated cost, with a capital cost expected at US$749 million.
WAPDA added generating capacity to an existing dam's spare tunnel under World Bank procurement and supervision, so the geological and resettlement risks that sink greenfield hydropower were mostly absent.
Preconditions
An existing large dam with an unused tunnel and an operator that had run it for decades
Concessional finance with lender-imposed procurement, supervision and an independent panel of experts
A power deficit large enough that the plant was dispatched at once
Where else it has been used: Pakistan and the World Bank repeated the model at the same dam with the fifth extension (T5HP).
Limits: The lender is grading its own project: 'on time and in less than estimated cost' is the World Bank's sentence in a restructuring paper, not an audit, and the paper gives no baseline figure, so the size of the saving cannot be checked from this source. The cost is an expectation in 2019, before every work was closed; the intakes for tunnels 3 and 4 were still being finished, and the loan closing date moved from 2018 to 2024 because the same project was extended to build T5HP. That fifth extension, financed under the same project number, is rated Moderately Satisfactory in the 2023 ISR and was due for completion in 2025; T4's record is not the record of the whole project. The World Bank financed about 90 percent of the civil and electro-mechanical works, so the case shows WAPDA delivering under a lender's procurement and supervision regime, which is a precondition rather than proof of domestic capability. Building a powerhouse on an existing dam and tunnel is much less risky than a greenfield dam, which is part of why it came in on budget.
๐ฎ๐นItalysince 2003still operatingbears on Large project delivery
Works on MOSE, Italy's system of 78 mobile gates at Venice's three lagoon inlets, began in 2003. All the gates were first raised on 10 July 2020, and 6,689.07 million euros had been appropriated by mid-2022 with the works about 95 percent complete and still unfinished past their planned 2021 completion.
The state handed design and construction of the lagoon barrier to a single concessionaire consortium under a special law for Venice, and when delays and judicial proceedings stalled it, it appointed extraordinary commissioners to finish and operate the gates provisionally.
Preconditions
A special legal regime and dedicated national funding for Venice dating from the 1980s
A single-concessionaire model that concentrated design, construction and control in one body
A visible, recurring flood threat that kept funding flowing through repeated delays
Limits: This is a delivery record with a bad cost and schedule history, not a success story: the barrier works, but late and expensive. The source gives total appropriations, not a baseline estimate, so a percentage overrun cannot be computed from it; the Court's report names a 'progressive increase in costs' without a starting figure here, and the original estimate quoted in the press is not used. Appropriated funds are not spending, and the total includes related works (Piano Europa, decarbonisation, San Marco basilica defences). The 95 percent completion figure is from June 2022 and the 2021 completion date had already passed; the record does not say when, or whether, the works were finished after that. The Court lists well-known judicial proceedings, a progressive increase in costs, technical difficulties, environmental questions and Covid-19 among the complexities of the build. The Court reports the gates have repeatedly protected the city, but the source gives no count of closures, so the record does not claim one.
๐ง๐ฉBangladeshsince 2009delivered and closedbears on Large project delivery
Bangladesh's Bridge Authority completed the 6.15 km Padma road-rail bridge with government funds. It opened to traffic on 26 June 2022 at a reported actual cost of Tk 30,770 crore, six and a half years after the 31 December 2015 completion date in the 2010 Asian Development Bank appraisal, which had costed the project at $2,915 million.
After its external co-financing package changed, the government kept the Bridge Authority as executing agency and paid for the bridge from its own budget. The Authority revised the project document as costs rose and opened the road deck in 2022, with rail running on the lower deck since.
Preconditions
A fiscal position able to absorb a multi-billion-dollar project without concessional lenders
A specialised bridge authority with a design and supervision record from the earlier Jamuna Bridge
Political commitment strong enough to carry repeated cost and schedule revisions
Where else it has been used: Self-financing a flagship after donors withdraw is rare. Where it has been tried, the usual result is the one here: the asset was delivered, and its cost and schedule were judged against shifting baselines.
Limits: Neither source publishes an overrun percentage. Converting the ADB appraisal at the RRP's own rate of Tk 70.11 per dollar gives about Tk 20,440 crore, so the actual cost is about 51 percent higher. That figure is our computation, and it mixes a 2010 dollar estimate with a taka outturn after more than a decade of taka depreciation and inflation, so much of the gap is currency and price level, not overrun. The ADB estimate includes $99 million of financing charges and appraised a co-financed project. The bridge was built with government funds, as the Bridge Authority page states, after that financing structure changed, so the two figures do not describe the same contract package. Against the government's own revised estimates on the same page, the outturn is about 5 percent below the Tk 32,306 crore figure and about 2 percent above the Tk 30,193 crore figure, which is why the cost record depends on which baseline is chosen. The Bridge Authority page is itself inconsistent: it lists two different estimates under the third revised project document. The schedule slip is clearer: an appraised completion of December 2015 against road opening in June 2022 and project closure in June 2024. The started year is the project start the Bridge Authority states, before main construction contracts were let.
๐ฒ๐ฆMoroccosince 2014still operatingbears on Large project delivery
Morocco's solar agency MASEN had the Noor-Ouarzazate I, II and III concentrated solar plants (510 MW) built, grid-connected and fully operational by October 2018, at a completed capital cost of US$1,962 million, about 7 percent below the appraisal estimate of US$2,093 million. Noor II and III then generated an average of 739 GWh a year in 2019 to 2023, against the 1,144 GWh expected at appraisal.
MASEN, the state agency set up for the Moroccan Solar Plan, tendered the plants to private developers under power purchase agreements and sat between them and the national utility ONEE. A World Bank-financed cost mitigation mechanism covered the gap between the solar tariff and what the utility paid.
Preconditions
Concessional finance pooled from the World Bank, KfW, AfDB, EIB, the EU and AFD, coordinated by one national agency
A state buyer willing to absorb a tariff above the cost of grid power for the life of the contracts
A site with high direct irradiation where the state delivered land, access, water and grid evacuation before the developers arrived
Where else it has been used: Morocco tried the same structure at Noor-Midelt. Construction had not started by June 2024, and Noor-Midelt II was redesigned as photovoltaic plus battery storage, because PV and battery costs fell while CSP costs did not.
Limits: The lender graded its own project here. The World Bank financed it, and the report rates the outcome 'Moderately Unsatisfactory', with efficacy and efficiency both 'Modest'. The cost figure is the Bank's economic-analysis number for the Ouarzazate plants. A footnote says it nets a 14 percent saving at Noor I against an overrun at Noor III. On schedule the report gives no figure, only that Noor II and III were built 'with a delay of just a few months'. Cost was on budget, but output was not. Generation ran about 35 percent below the appraisal expectation. Noor II missed its peak-hour share, at 32 percent against a 41 percent target. The Noor III tower plant had two prolonged outages caused by cracks in its molten salt tank: September 2021 to December 2022, and April 2024 to at least October 2024, when the report says it was still offline. The second phase financed under the same operation, Noor-Midelt, had not started construction by the June 2024 closing date. The project therefore delivered 510 MW against a 1,310 MW target. The economic rate of return was negative without carbon benefits. Generation-mix forecasts did not materialise either: coal's share rose to almost 70 percent by 2020. The capital came largely from concessional international lenders, with about US$4.1 billion disbursed from all financiers. A copy of this design would need that finance. All figures are read from a 40-page PDF.
๐บ๐ฟUzbekistansince 2013delivered and closedbears on Large project delivery
Uzbekistan's government and its state railway UTY began building the 124 km Pap-Angren line with a 19 km tunnel in mid-2013 and finished the works within nine months of the World Bank loan's 2015 effectiveness, with the main infrastructure costing US$1,475.68 million against an estimate of US$1,438.75 million, about 2.6 percent over, and benefits flowing about 18 months earlier than appraisal assumed.
The Cabinet of Ministers decreed the line and put it on a fast track, with the state railway UTY building the track with its own crews and equipment while design proceeded, and a Chinese contractor driving the tunnel in parallel.
Preconditions
A state railway that had already built 640 km of new line with its own workforce
Central authority able to order design and construction to run at once, and to compel the railway to share data
Concessional tunnel financing from China Exim Bank
Limits: The overrun of 2.6 percent is our computation from two figures in the ICR's cost table. The baseline is weak: construction ran simultaneously with detailed design and cost estimation, and by the Bank's appraisal in late 2014 the works were already very advanced, so the estimate was set when much of the cost was known, which flatters the overrun. The ICR gives no planned and actual completion dates for the line itself, only that works finished within nine months of loan effectiveness and benefits began about 18 months early. The 19 km tunnel was built by China Rail and Tunnel Company with China Exim Bank financing, so not all of the delivery was Uzbek institutional capacity, though the rail line was built by UTY's own workforce. The World Bank's own components went badly: the signaling contract was cancelled for misprocurement, US$72.87 million of the loan was cancelled, resettlement compliance was downgraded to moderately unsatisfactory, and UTY declined a second user survey. Passenger demand came in lower than projected. The source is the World Bank evaluating a project it part-financed (about 12 percent of cost), read from a PDF.
Between October 2008 and June 2020 the Ministry of Infrastructure's Cellule Infrastructure reopened 2,752 of a targeted 2,762 km of high-priority earth roads across five provinces and maintained 3,335 km, finishing 3.6 percent over the financing-agreement total and 28 months behind the combined original schedule.
The Ministry of Infrastructure's project unit (Cellule Infrastructure) bought the cheapest passable standard, spot-improved earth roads, through output-based contracts and the Office des Routes, so that mileage, not finish quality, set the pace and kept civil works inside budget.
Preconditions
Concessional donor finance covering most of the cost and a donor willing to extend the closing date four times
A road maintenance fund created in 2009 to keep reopened sections passable
Enough security on the selected corridors for contractors to work
Limits: The source is the World Bank's own completion report on a project it financed, and the PDF has to be read (table 3 and paragraph 50, pages 19-20). IDA carried 61.6 percent of the money and DFID, through World Bank-administered trust funds, 24.2 percent; the government's own Road Maintenance Fund put in 14.2 percent, so the delivery is Congolese institutions spending donor money under Bank supervision. The 3.6 percent total overrun is measured against financing agreements that were enlarged twice, so it does not compare with the 2008 appraisal cost (US$123 million for 1,800 km); the road works came in 22.6 percent under budget while institution building ran 1,656 percent and the environmental and social programme 235 percent over. The 28-month delay includes a 12-month suspension of disbursements in 2017-18 for non-compliance with environmental and social obligations, including gender-based violence, after an Inspection Panel investigation. The roads are low-standard reopened earth roads, about one fifth of the 15,800 km priority network; the ICR says ten kilometres are unaccounted for, rates M&E quality Modest, and says road-sector management and the maintenance fund's allocation remain opaque. Nothing in the report shows whether the roads stayed open after June 2020.
๐บ๐ฌUgandasince 2013delivered and closedbears on Large project delivery
Uganda began building the 600 MW Karuma hydropower station in August 2013 against a 60-month plan. By the end of June 2023 the Auditor General reported it 54 months late at 99.9 percent physical progress, and UEGCL records commercial operation from 12 June 2024.
The Ministry of Energy and Mineral Development and UEGCL let the dam as a single turnkey EPC contract to a foreign state contractor, financed mostly by an export-credit loan. That handed design and construction to the contractor, and land, resettlement and transmission stayed with Ugandan agencies, where much of the delay arose.
Preconditions
An export-credit lender willing to finance most of the cost against a sovereign guarantee
A state land acquisition and resettlement process able to keep pace with an EPC contractor, which it did not
Transmission lines built in step to evacuate the power
Limits: The record carries schedule performance only. No official baseline-against-outturn cost figure was found: UEGCL's page states a project cost of US$1.398 billion with no original budget beside it, so cost overrun is unknown. The Auditor General names USD 30 million of interest claims on delayed payments, which is one cost of the delay and not the overrun. The 54-month figure is measured at June 2023, when the plant was not yet in commercial operation. Commercial operation on 12 June 2024 implies a total overrun near 66 months, more than double the plan, but that figure is our computation from two sources. The 60-month plan is the Auditor General's statement of the plan, and the contract itself was not checked. The same report finds that the community development projects promised with the dam were largely unfunded (UGX 3 billion of 38.3 billion received) and that resettlement lagged, so the delivery is incomplete beyond the powerhouse. The record says nothing about how the plant has performed since commissioning.
๐น๐ฟTanzaniasince 2018delivered and closedbears on Large project delivery
TANESCO contracted the 2,115 MW Julius Nyerere hydropower dam on the Rufiji in December 2018 for TZS 6.56 trillion over 42 months to June 2022. The contractor completed it on 31 March 2025, and claims settled in the contractor's favour raised the cost to TZS 7.44 trillion.
The government had TANESCO let the whole dam as one EPC contract to an Egyptian joint venture and financed it from the domestic budget rather than concessional loans. That kept the project moving and kept it in national hands, but left contract management and the transmission build with TANESCO, where the audited losses arose.
Preconditions
A treasury able to fund a multi-trillion-shilling project from domestic revenue over several budget years
A utility with the contract-management capacity to enforce delay damages and resist claims, which the Auditor General found lacking
Transmission lines built in step with generation
Limits: The sources are Swahili PDFs. The quotes are from extracted text with spacing artifacts removed, and the English in the claim is our translation. Neither overrun is published as a percentage. The cost overrun of about 13.4 percent (6.56 to 7.44 trillion) and the schedule overrun of about 33.5 months against a 42-month plan (roughly 80 percent) are our computations from the Auditor General's dates and figures. The cost figure counts only the USD 400 million of claims the dispute board awarded. Price adjustment, financing costs, transmission, resettlement and supervision are not shown, so the true overrun is likely larger. The Auditor General adds that the contractor was granted two ex-gratia extensions to 14 January 2024, that TZS 327.03 billion of liquidated damages had not been deducted by November 2025, and that the contractor was released from TZS 262.34 billion of community projects. At completion, three of nine turbines had not been tested because the reservoir was too low, and the full-capacity test was not done on time because TANESCO lacked transmission capacity. Completion of the civil contract therefore does not mean the plant delivers 2,115 MW. The record says nothing about the dam's environmental cost in the Selous (Nyerere) protected area.
๐ท๐ดRomaniasince 2009still operatingbears on Large project delivery
Romania's national road administration has built the 582 km A1 motorway in sections, and construction of the sections the auditors timed began in 2009, and in November 2025 the European Court of Auditors found its latest cost estimate 11 percent below the original estimate in real terms, while one EU co-funded section had run 79 months, or 263 percent of its planned time, late.
Romania's national road company split a 582 km corridor into many small procurement lots and financed most of them with EU cohesion and connecting-Europe funds, which kept unit prices visible and auditable but multiplied permits and interfaces.
Preconditions
EU co-financing with audit and use-it-or-lose-it deadlines that discipline cost reporting
A national road agency able to tender and supervise dozens of separate lots
A permitting regime that, here, required a building permit every 7 km and slowed delivery
Limits: The cost figure is the auditors' aggregate of many sections planned and priced in different years, reindexed to 2019 values from an original estimate whose own date the auditors do not give, and it excludes the Bucharest-Pitesti section; the 2020 report warned that the unfinished Pitesti-Sibiu and Deva-Lugoj stretches left the figures subject to significant change. A real-terms fall can come from cheaper design, deferred scope or optimistic pricing of unbuilt sections, and the report does not say which. The auditors did not compute a schedule delay for the A1 as a whole because its timeline varies by section, and the 2020 report took 2030 as the original opening date only because the TEN-T regulation sets it, so the motorway's schedule performance cannot be read from the cost row. The 2026 update was a desk review of figures supplied by the Commission, national authorities and project promoters, with no site visits. The 79 months is one EU co-funded action, not the whole road, and it compares the Lugoj-Deva lots 2-4 action's planned end of May 2016 with an end date of December 2022 that was still a forecast in 2020. The 2020 report also found โฌ3.7 million of EU co-funding wasted, a finished Lugoj-Deva stretch left unused and 800 metres of built motorway demolished, and one building permit needed for every 7 km.
๐ฐ๐ฟKazakhstansince 2009delivered and closedbears on Large project delivery
Kazakhstan's Committee for Roads upgraded 1,281 km of the Western Europe-Western China corridor and connecting roads under a US$2.5 billion programme between 2009 and 2024, for an actual cost of US$2,468.6 million against US$2,500 million appraised, but over 15 years against a planned four and a half.
Kazakhstan's Committee for Roads borrowed from the World Bank to rebuild its stretch of the Western Europe-Western China corridor to dual carriageway, procured it in large international civil works packages, and used the savings from low bids to extend the scope across three oblasts.
Preconditions
Sovereign borrowing capacity backed by oil revenue
A road agency able to run international competitive bidding under lender rules
A corridor with transit demand that neighbours and lenders also wanted built
Limits: The cost line looks on budget only because the scope was moved to fit the money: bid prices came in low and the tenge depreciated, and the savings bought 219 km of extra road, so the total says nothing about cost per km against plan. The schedule figure is the honest one, and it is poor: the closing date was extended four times, the first three to absorb added scope and the last for COVID-19 and the termination of several civil works contracts, and a 42 km Otar-Uzynagash section was still unfinished at closing after two contractors were terminated for non-performance. The source is the lender's own completion report; the World Bank financed 85 percent of the cost, so this is a borrowed-money build supervised by the lender, though executed and repaid by Kazakhstan. Traffic came in about a third below forecast. It is one corridor programme, not a portfolio, so it cannot stand for Kazakhstan's average project performance.
๐น๐ณTunisiasince 2010still operatingbears on Large project delivery
Tunisia's transport ministry and the Societe du RFR built the first two lines of the Tunis rapid rail network, about 18 km plus 28 trains, financed from December 2010; the EIB's approximate total cost for the same scheme rose from EUR 550 million at the 2009 appraisal to EUR 653 million at the 2017 appraisal, Line E opened in March 2023 and Line D on 25 January 2025, against a 2015 project study that scheduled construction of the Line E viaduct for 2014 to 2017.
The Ministry of Transport created a dedicated project company, the Societe du RFR, assembled a 40 percent state and 60 percent multi-donor financing package, and built the first two lines through civil works contracts while SNCFT took the operating role; when the Bardo municipality stopped the works, the company went to the administrative court and the cabinet chose a design variant in December 2023.
Preconditions
Concessional lenders (EIB, AFD, KfW, EU) willing to stay in for more than a decade and to suspend and restart disbursement rather than exit
A rail right of way partly owned by SNCFT, so most of the route did not need fresh land
An administrative court able to overrule a municipality that blocked works on a national project
Where else it has been used: The same multi-lender package financed Line A in 2013 and is meant to carry the remaining lines of the planned 85 km network; the Tunis case is usually retold as a delay story, and the part that drops out is that the lines did open.
Limits: The cost figures are the EIB's approximate appraisal totals, not an audited out-turn: the 18.7 percent rise from 550 to 653 million EUR is our computation across two appraisals whose scope wording differs slightly (the 2017 sheet names a maintenance facility and 28 trains, the 2010 sheet only rolling stock), and in January 2025 the French Treasury still quoted a total of 550 million EUR including rolling stock, so the publishers do not agree on the out-turn. The cost is in euros and says nothing about dinar costs after the currency's depreciation. The record has no official original line-opening date: the only official schedule is a 2015 impact study placing the Line E viaduct's construction in 2014 to 2017, so the roughly six-year slip to the March 2023 Line E opening is a comparison of one structure's plan against the line's opening, not a published overrun. The EIB suspended disbursements in 2016 over construction companies' failures to implement safeguards (ESDS 18 July 2017), and works at Bardo were halted by the municipality until an administrative court annulled the stop on 29 December 2023 (EIB register 236943929), so part of the delay is litigation rather than delivery. About 60 percent of financing came from foreign lenders and French and Korean firms supplied engineering and trains. This is a city network, not a national programme, and it records poor delivery rather than good: it is evidence of the construct's weak end.
๐ฑ๐ฐSri Lankasince 2003delivered and closedbears on Large project delivery
Sri Lanka's Road Development Authority built the country's first access-controlled expressway, about 126 km from Kottawa to Godagama. Sections I to III opened on 26 November 2011 and section IV on 15 March 2014. The project took 13 years against a planned six. Its total cost rose from US$295.9 million at the 1999 appraisal to US$906.5 million at completion in 2014, partly because it was widened from two to four lanes.
The Road Development Authority ran one project management unit for an expressway financed in sections by ADB, JICA, the Nordic Development Fund and later China Exim. When prices rose and the design changed, it went back to the lenders for a supplementary loan instead of cutting the scope.
Preconditions
Several concessional lenders willing to finance one corridor in parallel and to top up after a cost overrun
Land acquisition and resettlement capacity, whose absence caused most of the early delay
A design finalized before contract award, which this project lacked and which ADB names as its main lesson
Where else it has been used: The same Road Development Authority model was reused for the Colombo-Katunayake Expressway (2013), the Outer Circular Highway and the Southern Expressway extension to Hambantota (2020), mostly with Chinese loans instead of multilateral ones.
Limits: The cost growth is not a like-for-like overrun. In 2008 the scope was widened from a two-lane road to four lanes, and against the 2008 re-appraisal of $757.7 million the overrun is about 20 percent, not about 206 percent. Both ratios are our computation from the report's figures. The cost table is sourced to the borrower's own completion report (Government of Sri Lanka, 2011), and ADB, as lender, judges a project it financed. The PCR's cost-by-financier table puts China Exim at $159.48 million within the $906.47 million total, so the section it financed is included. The schedule figure (13 years against six) runs from the 1999 approval and includes four years of design and land acquisition before construction.
๐ณ๐ตNepalsince 2001operating below its peakbears on Large project delivery
Nepal's government built the Melamchi River diversion tunnel and headworks to pipe an additional 170 million litres a day into the Kathmandu Valley, closing the project in June 2021, more than 14 years after its original March 2007 closing date; the 2021 flood then damaged the headworks, and water has since been diverted for about eight months a year.
The Ministry of Physical Planning and Works and later the Ministry of Water Supply, with ADB as lead financier, tied a trans-basin water diversion to a reform of Kathmandu's water utility in a single loan. The infrastructure was delivered and the institutional reforms largely were not.
Preconditions
Concessional finance from several lenders willing to stay through repeated contractor failure and the exit of three co-financiers
An executing agency able to cancel and re-tender a failed tunnel contract in mountainous, flood-prone terrain
A utility (KUKL) whose tariff and water-levy reforms the project assumed but could not compel
Where else it has been used: ADB's own lesson is that bundling a major physical build with a governance overhaul in one operation was over-ambitious; the remaining reforms were moved into separate pipeline programmes.
Limits: The 14 years is the gap between two loan closing dates, which is a lender's administrative milestone, not a measured completion date for the tunnel; the source gives no physical completion schedule against a baseline. The cost figure is not usable as an overrun: appraisal was $464 million and completion $345.3 million, but the 2008 scope change and the exit of the World Bank, NORAD and Sida mean the two totals price different projects, and ADB says separately that a 2014 additional loan covered tunnel cost overruns without stating the overrun in percent. The erosion is stated in words and one current number: the source publishes no figure for the year-round flow before the 2021 flood, so the secondMetric holds the post-flood level, not a measured peak. The flood is an exogenous shock, though ADB attributes the vulnerability to design that did not fit the terrain. The validation rests on the project completion report written by the lender and the executing agency, and ADB was itself a party to the project's design. The delay mixes national causes (slow decisions, leadership turnover, poor contractor management) with contractor failure, the cancelled and re-bid contracts of nonperforming contractors.
๐ฟ๐ผZimbabwesince 2020still operatingbears on Large project delivery
Zimbabwe's government, funding the works from the budget, set out in its 2021 Budget to finish upgrading the roughly 580 km Harare-Beitbridge highway by 2022; it had opened 355 km to traffic by October 2022, and in July 2025 the Finance Ministry still reported 88 km outstanding, more than two and a half years past the 2022 target.
The Ministry of Transport split the corridor into sections and paid local contractors from the national budget, opening the road section by section as money arrived.
Preconditions
Budget space for a large capital programme during high inflation and currency redenomination
Domestic contractors able to build trunk-road sections to regional standards
A political commitment to a flagship corridor strong enough to survive successive missed targets
Limits: This is a schedule record with no cost figure: the ministry publishes allocations in a currency that was redenominated twice in the period (ZWL to ZiG) and gives no baseline budget for the whole road, so no cost overrun can be computed. The schedule slip of more than two and a half years (end of 2022 to July 2025) is our arithmetic from targets and progress the same ministry wrote in successive budget documents. It is not an audited figure, and the ministry itself kept moving the target (2022, then 2023, then 2025). The documents give the road's length as 560 km in 2020 and 580 km from 2022 onward without explaining the change. Kilometres 'opened to traffic' are self-reported and say nothing about pavement quality. The start year is that of the earliest official progress note, July 2020. Delivery used several local contractors under a national emergency programme, so this is evidence about a fiscally funded, multi-contractor build, not a single tender.
๐ฟ๐ฒZambiasince 2016operating below its peakbears on Large project delivery
Zambia's government contracted Shandong Dejian Group for a fixed-price US$200 million national programme of solar powered milling plants, financed in 2016 and due 24 months after the March 2016 deposit. The Auditor General found it still incomplete at 31 March 2022, 38 months after the scheduled completion date, and found 555 of the completed plants non-functional at audit in January 2024.
The Ministry of Finance borrowed from China Development Bank to buy a turnkey national programme of off-grid milling plants from one Chinese contractor, and appointed the Zambia Co-operative Federation to run them through local co-operatives. No public body was funded to maintain the equipment, so delivery slipped for years and a third of the completed plants stopped working.
Preconditions
A maintenance and spare-parts budget for battery-based equipment spread across rural districts
An implementing agent with the technical staff to accept, commission and service the plants
Contract management able to enforce the fixed-price completion clause against the supplier
Limits: The Auditor General never publishes the peak number of working plants. The secondMetric is the contracted count, not the peak installed or working count. The 555 non-functional plants are counted among the 1,553 milling plants, about 36 percent; that share is our inference. The audit inspected selected sites, so the 555 figure may rest partly on progress reports rather than visits to every plant. The report is internally inconsistent: the loan paragraph describes 1,000 plants for US$200 million, while the finding counts 1,583 contracted units. It also dates the deposit (March 2016) before the China Development Bank loan (July 2016). The section gives no final cost and no completion date, so neither a cost overrun nor a final schedule overrun can be computed. 38 months is the delay at 31 March 2022 against the revised January 2019 date; against the original 6 March 2018 date it is about 49 months. The plants are distributed small units rather than a single structure, which makes the record a weaker fit for major public infrastructure. The Zambia Co-operative Federation, the implementing agent, is a federation and not a ministry. The audit also found K15.5 million in lease fees that were never paid into the Treasury. Nothing here shows how many households used the mills, or whether the non-functional plants were later repaired.
๐ฒ๐ฑMalisince 2014delivered and closedbears on Large project delivery
Mali's water assets company SOMAPEP-SA built the reservoirs, transmission mains and 1,016 km of distribution networks that brought water services to 803,000 people in Bamako and six other towns, alongside the Kabala treatment plant that came into service in June 2019. Of the US$130 million approved, US$119.6 million was spent, and the project closed in December 2022 against an original closing date of December 2018.
Mali's government split urban water into an asset-holding company (SOMAPEP-SA) that builds and a utility (SOMAGEP-SA) that operates, under a regulator (CREE). SOMAPEP-SA then used an IDA credit to build the transmission and distribution side of an eleven-donor Kabala programme, while other donors financed the treatment plant.
Preconditions
A standing asset company and utility able to procure and supervise works through two coups and a pandemic
Donor co-financing coordinated through one programme, so the plant and the networks were built in step
Tariffs set high enough to keep the system running, which Mali did not do
Limits: The source is the World Bank's self-evaluation of its own credit. Spending US$119.6 million of an approved US$130 million is not a cost underrun against a fixed scope. The approved amount grew from US$80 million to US$130 million in 2016, when six towns were added. The report says savings were used to raise targets, and it publishes no unit cost or contract-level overrun. The four-year schedule slip, from a December 2018 to a December 2022 closing, is our reading of the key dates. Three of those years were added with the 2016 scope expansion and one year in 2021 for COVID-19 and the disbursement suspensions that followed the 2020 and 2021 coups, so most of the slip was planned. Targets outside Bamako were not met because distribution networks were finished late. The Kabala treatment plant itself was financed by AFD, the EU and the EIB, and its cost and schedule are not in this record. The project was designed to make the urban water sector financially sustainable, and that objective failed: tariffs were frozen for eight years, and the indicator was removed in 2021. The project was IDA-financed throughout. The national contribution was implementation by SOMAPEP-SA and SOMAGEP-SA, not money.
๐ง๐ซBurkina Fasosince 2001delivered and closedbears on Large project delivery
Between 2001 and 2007, ONEA, Burkina Faso's national water utility, delivered the Ziga project through a dedicated project unit. The project built the Ziga earth dam and its transmission main to Ouagadougou, plus 56,000 subsidised connections and 400 standpipes. Eleven external financiers and the government paid US$269.37 million, 131 percent of the US$205.88 million estimated at appraisal, and the project closed on its original date of 31 December 2007.
The government gave ONEA a mandate by decree in 1996 to set up a dedicated Ziga project unit. That unit ran one investment programme for eleven external financiers and the government, with joint annual supervision meetings, while the government adjusted tariffs and paid its own water bills so the utility stayed solvent.
Preconditions
A national utility already financially sound enough to borrow and to recover costs through tariffs
Donor coordination under one project unit and one supervision committee rather than parallel projects
Political backing for tariff adjustments and for the state paying its own utility bills
Where else it has been used: The ONEA model, a public utility under a performance contract with its own project units, became a West African reference case for urban water reform. Later Burkina Faso programmes extended it to sanitation and secondary towns.
Limits: The 31 percent overrun is in US dollars, and much of it is exchange rate. The ICR's own note puts the dollar at 640 CFA francs at appraisal and 506 at closing. Converted at those two rates, the appraisal total of about CFAF 132 to 134 billion and the actual total of about CFAF 136 billion are within a few percent of each other. Spending happened between the two dates, so neither conversion is exact, and the ICR prints both 640 and 650 for the appraisal rate. The real cost overrun in local currency is therefore unknown and probably small, and the dollar figure overstates it. The schedule figure only covers the 2001 to 2007 financing window. The project was conceived in the early 1990s and launched in 1994, so on-time closing says nothing about the decade of preparation before it. The ICR is the World Bank grading an operation it financed and supervised, and it relies on the borrower's reporting. The people-served figure is an estimate the ICR computes at 10 people per connection and 120 per standpipe, not a count. Donors provided almost all of the money, and the borrower's share was about US$8.8 million, so the record shows execution by ONEA, not national financing. The data are from 2007 and say nothing about Burkina Faso's delivery of large projects since then.
๐ง๐ฎBurundisince 2020delivered and closedbears on Large project delivery
The Government of Burundi, through REGIDESO and a World Bank project unit, built the 49.5 MW Jiji (32.5 MW) and Mulembwe (17 MW) run-of-river plants with their transmission lines and substations. The project was approved in 2014 at US$270.3 million with an original closing date of December 2019. Construction began only in February 2020, after the 2015 political crisis. The World Bank put the estimated cost at US$319.1 million in November 2024. Jiji was inaugurated in June 2025 and Mulembwe in June 2026.
Burundi bundled two run-of-river plants, transmission and utility reform into one donor-pooled project under a World Bank-supervised implementation unit. It held the scheme together through a political crisis and two additional financings until both plants were commissioned.
Preconditions
Concessional co-financiers willing to top up a stalled grant rather than cancel it
An owner's engineer and contractors able to work in a fragile state with fuel and foreign-exchange shortages
A utility (REGIDESO) able to absorb and operate new generation capacity
Limits: The cost figure is the World Bank's estimate in November 2024, before either plant was commissioned. At that date it reported a financing gap of US$41.7 million still to close. The 18 percent is the Bank's own ratio in its cost table, set against the 2014 appraisal. It excludes a further estimated US$19 million lost to currency depreciation on the 2014 pledges. The schedule slip is not published as a figure. The original closing date of 31 December 2019 against inauguration in June 2025 (Jiji) and June 2026 (Mulembwe) is our reading of a 6.5-year slip. Much of it comes from the 2015 to 2020 political crisis, which delayed hiring the owner's engineer to 2017 and the contractor's mobilisation to 2020, so it is not purely an execution failure. The project was financed and supervised by IDA, the EU, the EIB and the AfDB. Delivery was shared, and the Bank reports weak contract management in the project unit. Completion rests on press releases: the Presidency's June 2025 release for Jiji, and the EIB's June 2026 release for Mulembwe. Neither release shows that the plants deliver the planned 235 GWh a year.
๐ฒ๐ฌMadagascarsince 2004delivered and closedbears on Large project delivery
Madagascar's road authority and transport agencies delivered road, port, airport and Northern Railway works under the Transport Infrastructure Investment Project from 2004, cutting travel time on the RN44 from 24 to 6 hours, with the closing date extended four years from 30 June 2008 to 30 June 2012 and IDA-financed costs of US$169.39 million against US$150.00 million at appraisal plus US$15.60 million of additional financing added in 2007.
Madagascar's road authority and port and rail agencies ran a multi-modal investment programme on a lender credit, re-scoping it repeatedly as priorities and the political situation changed, and finished four years late.
Preconditions
A road authority and transport agencies able to procure civil works
Lender flexibility to restructure and add financing
Political continuity through implementation, which Madagascar did not have
Limits: The costs are in Annex 1 of the World Bank's own completion report, which rated the outcome Moderately Satisfactory. They are IDA disbursements only (the borrower's own contribution is shown as US$1.11 million against 17.42 appraised), and the 112.9 percent of appraisal is almost entirely the US$15.60 million additional financing of 2007 for the Northern Railway component, which was not in the appraisal; against the revised estimate including that financing, the outturn is 102.3 percent, so the record shows no meaningful cost overrun and should not be read as one. Port and airport components spent a fraction of their estimates and the RN5a upgrade was dropped at the January 2006 restructuring, so scope shrank in some places while it grew in others. The RN44 travel-time result is the report's own indicator comment. The closing date moved through five restructurings, and the 2009 political crisis froze disbursements during implementation. The project is a portfolio of works, not one build, so no single project's baseline and outturn are shown. Plain-text file, no PDF page numbers.
๐ฒ๐ผMalawisince 2012still operatingbears on Large project delivery
Malawi's electricity utility ESCOM completed and energised all 10 substations and seven of nine planned distribution lines under the Energy Sector Support Project approved in 2011, closing on 15 October 2018 against an original closing date of 30 October 2016, having disbursed US$72.1 million of US$84.7 million.
Malawi's state utility reinforced the grid in its load centres with a package of substations and lines on concessional credit, and finished the works itself after its supervising consultant ran out of budget.
Preconditions
A utility with in-house engineers able to finish works without a consultant
Concessional finance with room to extend closing dates
Demand growth that kept the works a priority
Limits: The World Bank's completion report, which rated the outcome Satisfactory, is the source; the substation and line counts are in paras. 103 and around 583 of the text file and the dates and amounts in the data sheet. The report gives no appraisal cost per work package, so cost performance cannot be read beyond the undisbursed balance, which reflects delay and cancellation rather than savings. The two-year extension followed procurement delays the report attributes to the political situation and inexperience with Bank procedures, and the implementation consultant stopped work in December 2017 when its budget ran out, leaving ESCOM to finish alone. At US$85 million this is a modest programme of grid reinforcement, not a single major build, so it is weak evidence on major project delivery. Plain-text file, no PDF page numbers.
๐ง๐ผBotswanasince 2009still operatingbears on Large project delivery
Botswana Power Corporation built the 600 MW Morupule B coal power station under an EPC contract appraised in 2009, taking the plant over with known defects on 4 May 2014 against a target of 31 December 2013, after which it ran at a capacity factor of 52 percent from April 2015 to January 2016 against the 80 percent projected at appraisal.
The state utility bought a large coal plant on a turnkey contract from a lowest-cost bidder, accepted it with a punch list of defects to meet a power shortfall, and then imported power and ran diesel units while the defects were chased.
Preconditions
Owner's engineering capacity to verify design and construction quality under an EPC contract
Contract terms that let the owner refuse takeover of a defective plant
A grid with enough backup supply to wait for a plant to be fixed
Limits: The World Bank's completion report rated outcome, Bank and borrower performance all Unsatisfactory. The 52 percent is the measured gross capacity factor for the whole plant in Table 2 for April 2015 to January 2016 (54 percent for April 2014 to March 2015); Table 1 gives about 38 percent from grid connection to March 2014, and paras. 91 and 96 describe availability 'hovering around 50 percent'. On cost the project did not overrun: the power-station, transmission and water component came to US$1,188.6 million against an appraisal estimate of US$1,539.6 million (Annex 1), so the failure is in schedule and performance, not budget. The report attributes the defects mainly to the Chinese EPC contractor's design and construction quality, with some operation and maintenance issues, and records three site fatalities; the plant was financed largely by an ICBC loan with a World Bank guarantee. The report does not cover the plant after 2016, so whether it later reached its design output is not shown. 'Operating' is used because the plant generates; the record documents a delivery that never reached its design performance rather than one that eroded from a peak.
๐ฑ๐ธLesothosince 2011still operatingbears on Large project delivery
Lesotho's Metolong Authority built the Metolong Dam and raw water pump station near Maseru, starting impoundment on 15 February 2014 against a contract date of 31 October 2012 and delivering 93 megalitres a day on 8 October 2014 against a contract date of 1 July 2013, with every component of the programme overrunning its cost.
Lesotho created a dedicated authority to deliver a dam, treatment works and bulk pipelines for the lowlands with many co-financiers, and delivered water about 15 months behind contract after price escalation and procurement delays.
Preconditions
A ring-fenced delivery authority for one programme
Several co-financiers able to cover cost gaps
Experience from the Lesotho Highlands Water Project to draw on
Limits: The contract and revised dates are in Table 9.1 (Annex 9, page 53 of 59) of the World Bank's completion report, which rated the outcome Moderately Satisfactory; both delays in months (15.2 and 15.5) are computed here from the dates in that table. The report states that all components overran their cost, attributing it partly to construction-price inflation from South Africa's 2010 World Cup works (para. 72), but gives no programme-wide baseline and outturn cost, and the World Bank financed only part of a programme co-financed by the Kuwait Fund, Arab and other lenders, the EIB and the government. The project closing date was extended three times for a cumulative 42 months, to 31 December 2018. Impoundment coincided with the 2015 to 2016 drought. Plain-text file of a PDF.
๐ฒ๐บMauritiussince 2010still operatingbears on Large project delivery
Mauritius's Ministry of Public Infrastructure upgraded the 12.1 km B6 Belle Rive to Quartier Militaire road under the World Bank Infrastructure Project approved in 2009, at a cost of US$33.88 million against an appraisal estimate of US$21.60 million, 157 percent of the estimate.
The infrastructure ministry procured a rural road upgrade on an estimate that was never checked against designs, then added scope under local pressure, and paid half as much again as appraised.
Preconditions
Detailed engineering designs before the cost estimate is fixed
A scope-change control that resists late stakeholder additions
A lender willing to reallocate funds from cancelled technical assistance
Limits: The costs are in Annex 1 of the World Bank's own completion report, which rated the outcome Moderately Satisfactory and the borrower Moderately Unsatisfactory. Para. 41 explains most of the increase as poor checking of the estimate at appraisal and added scope requested by stakeholders (junctions, a bridge, village works), with further increases from weather and flooding, and gives slightly different totals (about US$32 million, then about US$35 million for the contract), so the 157 percent is the annex figure and the narrative does not match it exactly. A 12 km road is a small project for the construct, chosen because it is the one Mauritian build with an official baseline and outturn found; Mauritius's larger Metro Express light rail was not assessed. The project closed on its original date, 31 December 2014. Plain-text file, no PDF page numbers.
๐ธ๐ฉSudansince 1973delivered and closedbears on Large project delivery
Sudan's government and the Rahad Corporation (RAHCO) built the Rahad irrigation scheme, 300,000 feddans east of the Rahad River, a tributary of the Blue Nile, with a main pumping station, a siphon under the Dinder River and a barrage on the Rahad River, under credits approved in March 1973; the physical works were completed in December 1982 against a planned June 1978, at a final cost of US$395.6 million against an appraisal estimate of US$125.0 million.
The government let a large pump irrigation scheme go to tender before detailed design was finished, then refinanced a doubling of cost through a second round of Arab and IDA credits and absorbed the remainder in the national budget.
Preconditions
Appraisal based on a feasibility study rather than detailed design
Several co-financiers willing to raise commitments mid-project
A construction market where contractors had better-paying bids elsewhere in the region
Limits: The report is the World Bank auditing a project it financed alongside the Kuwait Fund, the Arab Fund, the Saudi Fund and others, so the delivery was Sudanese but the money and the judgement were largely external. Much of the threefold overrun came before construction: the project was reappraised in 1975 at US$320.7 million, 156 percent above 1973, because detailed designs had not been done at appraisal and the 1973 oil shock drove inflation, so only a 23 percent overrun is measured against a design-based estimate. The final cost is itself an estimate: the report says local costs such as administration and settlement are approximate and that better information had been requested. Figures are nominal dollars at the exchange rates of the time. The schedule comparison uses the dates for completing physical components in the report's data table (June 1978 appraised, December 1982 actual); the text says the project finished more than four years late. The scheme did develop the full planned area, and the auditors re-estimated its economic return at 20 percent against 14 percent at appraisal, so a cost overrun did not mean a failed asset. The text file is an OCR of a 1984 typescript and some table cells are garbled.
๐ฌ๐ญGhanasince 2007delivered and closedbears on Large project delivery
Ghana's Department of Urban Roads, joined in 2012 by the Ministry of Local Government and Rural Development, set out in 2007 to build Sub-Saharan Africa's first bus rapid transit corridor in Accra for an estimated US$46 million of infrastructure, but after a design review raised that estimate to US$100.7 million and counterpart funds did not arrive, the BRT was dropped in the 2014 restructuring without its 9.1 km of dedicated lanes being built, and the project closed in December 2015, 35.5 months after its original December 2012 closing date.
Two government agencies split responsibility for the roads and the buses of a first-of-its-kind BRT, tendered it before the design was complete, and when bids came in at double the estimate and the treasury could not pay its share, cut the corridor and spent the remaining credit on cheaper bus lanes elsewhere.
Preconditions
A new transit mode with no domestic design or operating precedent
An appraisal without an up-to-date land use plan or transport master plan for the city
Counterpart funding that depended on the annual budget rather than a ring-fenced allocation
Limits: The source is the World Bank's own completion report on a project it financed with AFD and GEF money, and it rates the outcome Unsatisfactory, so it judges the Bank's design as well as Ghana's delivery: the report says the BRT design was not fully developed at appraisal and that bid documents were prepared during implementation. The US$100.7 million is a revised estimate, not money spent, because the BRT was cancelled before Lots 2 and 3 were contracted; what was built under Lot 1 (the Odaw Bridge expansion and a railway flyover) cost US$14.7 million against US$9.2 million estimated, a 60 percent overrun. The 35.5 months is computed from the original and actual closing dates in the report's key dates table (31 December 2012 and 15 December 2015); the text gives the extensions as 29.5 months for two of them. The remaining funds went to quality bus services on two other corridors, which were not operating at closure. What happened to bus services in Accra after 2015 is not covered here.
Cรดte d'Ivoire's Ministry of Public Works built the 70 km Yabayo-Gagnoa road and rebuilt the bridge over the N'Zi River under the Fourth Highway Project (Credit 406-IVC, approved June 1973), completing it in September 1978 against a planned June 1976, at a cost of US$24.7 million against an appraisal estimate of US$12.3 million, an overrun of 101 percent.
The public works ministry split a new road into separate contracts for earthworks, paving and aggregate, gave the first to an inexperienced state enterprise, and the later contracts stalled waiting for it, doubling cost and tripling the construction time.
Preconditions
A government push to upgrade main roads to high standards in the 1970s boom
State-owned construction enterprises given work without proven capacity
A lender that accepted the borrower's contract split against internal warnings
Limits: This is a modest project by today's standards: a 70 km road and one bridge, US$24.7 million in 1970s dollars, of which the IDA credit was US$7.5 million and the rest the government's, so it is a mid-sized public works delivery rather than a national megaproject; no larger Ivorian project with a published baseline and outturn was found. The source is the World Bank auditing its own lending. The audit attributes the overrun to the road construction and the 27-month delay mainly to the decision, accepted by the Bank, to split the works by function: a government-owned enterprise lacking managers and trained operators did the earthworks late, holding up the French-Ivorian joint venture that was to lay the base and surface. The schedule figures come from the audit's basic data sheet (June 1976 planned, September 1978 actual, a 175 percent time overrun); the text gives completion as 27 months after the project completion date. Despite the overrun, the auditors re-estimated the economic return at 24 percent against 14 percent at appraisal. The text is an OCR of a 1981 typescript.
๐ฌ๐ณGuineasince 1989delivered and closedbears on Large project delivery
Guinea's Ministry of Transport and Public Works paved the Dubreka-Kolaboui (196.5 km) and Mamou-Kankan (401 km) national roads under the Fourth Highway Project, approved in June 1988, at a latest estimated cost of US$207 million against an appraisal estimate of US$200 million, with the IDA credit closing on 31 December 1996, three years late, and the last 94 km of Mamou-Kankan still under construction in 1997.
The public works ministry packaged two long trunk-road upgrades with spot repairs and labour-based maintenance into one programme, and paid for it by assembling seven donors in parallel, each financing particular sections.
Preconditions
Donors willing to finance road sections in parallel under one sector programme
A ministry able to procure and supervise several international civil works contracts at once
Counterpart funds for local costs, which here arrived at less than half the planned level
Limits: The report is the World Bank's own completion report on a credit it made, and the IDA credit was only US$55 million of the US$207 million: parallel finance of US$143 million came from the AfDB, the French CFD, USAID, Japan's OECF, BADEA and the European Community, so the cost figure covers a multi-donor programme run through the ministry. The cost is a 'latest estimate', not an audited outturn, and the 3.5 percent overrun is small partly because the scope changed: about 500 km of cement-stabilised base and asphalt concrete on 310 km were added, so the money bought a higher standard than appraised. The three-year delay is on the credit's closing date, not on road completion dates; the report says the last 94 km of Mamou-Kankan, financed by the EC, was to be finished by the end of 1999. The government paid about US$8 million of its planned US$19.2 million contribution. The report blames delays on a slow restructuring of the ministry, poor technical studies, cumbersome procurement and late payments to contractors.
The Republic of the Congo's transport agency ATC rebuilt an 88 km section of the Brazzaville to Pointe-Noire railway through the Mayombe range, with 19 bridges, three tunnels including one of 4.6 km and 17 million m3 of earthworks, completing it in mid-1985 against an expected end of 1979 at a cost of US$399 million against an estimate of US$149 million, an overrun of 168 percent.
The state transport agency, running the railway at the same time, managed a mountain realignment with tunnels on engineering that underestimated the ground, and when bids came in high and the works slipped, ten lenders refinanced the doubling of cost rather than stop.
Preconditions
A single national rail corridor whose collapse was feared, which made stopping politically impossible
Lenders willing to add supplementary credits to a half-built line
Traffic forecasts tied to timber and manganese exports that did not materialise
Limits: The audit is the World Bank's evaluators on a project the Bank appraised and part financed: the realignment had ten financiers, including France, the European Community and the Bank's US$38 million loan and US$30 million supplementary credit, so the money was largely external and the audit judges the Bank's appraisal as much as ATC's execution; it says the Bank accepted detailed engineering based on inadequate soil, topographic and geological study and allowed physical contingencies of only 10 percent for earthworks and 15 percent for tunnels. Figures are nominal dollars over a decade of high inflation, so part of the overrun is price escalation. The 168 percent and five and a half year delay (a 157 percent time overrun) are the audit's own computations. Traffic, expected to double, stayed flat, and the audit puts the economic return at zero or negative and calls the project unsustainable, so the record is a delivery that was finished but did not pay off. The text is an OCR of a 1990 typescript and some words are garbled.
๐ฌ๐ฒGambiasince 2019still operatingbears on Large project delivery
The Gambia's Ministry of Transport, Works and Infrastructure contracted the US$35.7 million (D1.8 billion) Banjul roads, drainage and sewage rehabilitation in May 2019 without tender, for completion by November 2021; by August 2021 12,953 m of the 15,216 m of roads, 85 percent, were built and 63 percent of the contract price paid, the pumping and lifting stations were outstanding, and the completion date had been extended to November 2022.
The works ministry accepted a contractor's unsolicited offer to pre-finance the reconstruction of the capital's roads and drains, signed a single-source design-and-build contract, then switched to paying it from the budget in interim instalments ahead of the hardest components.
Preconditions
A procurement authority that can be bypassed by single sourcing for a presidential priority
A contractor willing to start work before a signed contract
Budget room to pay interim certificates once the pre-financing was dropped
Limits: The audit is a compliance management letter, not a completion audit: it stops at 31 August 2021, three months before the original deadline, so it shows progress and a one-year extension of the deadline to November 2022, not an outturn. Whether the project was finished, at what final cost and on what date is not shown here, and the status 'operating' rests only on the audit's finding that most roads were built and in use. The 85 percent is of a road length that an addendum had raised from 14,614 m to 15,216 m, and the audit does not report a cost increase from that addendum. The figures are in a 24-page PDF (site verification on page 23). The audit's main findings are about process, not schedule: the contract was single sourced and gazetted only in February 2021, works began on 18 March 2019 before the contract was signed on 11 May 2019, no performance security was provided, withholding tax was not deducted, and 63 percent had been paid while critical components worth US$2.47 million, 6.9 percent of the sum, were outstanding. A contractor-initiated, pre-financed offer is how the project began, so the institutional delivery is partly the contractor's proposal.
Guinea-Bissau's port authority (JAPG) rehabilitated the old commercial wharf of Bissau Port, built river ports and ferry ramps and, with a new deepwater wharf financed in parallel by four Arab funds, completed the IDA-financed Bissau Port Project on 30 June 1990 against a planned 31 December 1987, at a final cost of US$20.0 million against an appraisal estimate of US$17.36 million.
The government split its port programme into an IDA part and an Arab-funded new berth awarded to the same contractor, so when the Arab funds suspended payments over the country's arrears the contractor slowed work on both, and IDA supervision kept the project going to completion.
Preconditions
A port authority set up as a separate agency able to borrow and contract
Several lenders financing separate parts through one contractor
A government able to keep current on earlier loans, which here it was not
Limits: The report is the World Bank on its own credit, and it notes that the government sent no comments for Part II, the borrower's review, despite repeated requests; consultants acting for the borrower prepared a final report the PCR drew on, so the borrower's own view is largely missing. The cost figures cover only the IDA-financed part (wharf repair, river ports, technical assistance and supervision of the new berth); the new deepwater wharf itself was financed separately by the Kuwait Fund, the Saudi Fund, BADEA and the OPEC Fund, and its cost and schedule are not reported here, though the report blames repeated suspensions of Arab Fund disbursements, triggered by the government's late repayment of earlier loans, for most of the delay. The 14.9 percent overrun is in dollars; the report says the credit, made in SDRs, covered the work without an overrun, and scope grew with four extra ferry ramps, container storage and more technical assistance. The project is small in absolute terms, about US$20 million, but Bissau is the country's main port and the report credits the project with a modern deepwater facility where only a badly damaged wharf had existed. The text is an OCR of a 1991 typescript.
๐ธ๐ฎSloveniasince 2008delivered and closedbears on Large project delivery
Slovenia's state-owned power holding HSE and its subsidiary TEล raised the estimated cost of the 600 MW ล oลกtanj unit 6 from 1,302,747 thousand euros in the August 2011 investment programme, the ceiling written into the 2012 law guaranteeing its 440 million euro EIB loan, to 1,428,457 thousand euros in the February 2014 programme, breaching the condition on which the state guarantee was given, according to the Court of Audit.
A state-owned utility group committed to a single large lignite unit, financed it with a state-guaranteed EIB loan conditioned on a cost ceiling, and then revised the investment programme upward several times while construction ran, leaving the state as guarantor to accept the breach.
Preconditions
A state holding with a captive coal mine and a political mandate to keep the lignite region working
A state guarantee law that set a cost ceiling but no mechanism to stop the project when it was crossed
Shareholder oversight bodies that did not ask for cash-flow reports until after the overrun
Limits: Both figures are in a PDF, on PDF page 9 of the post-audit report, which restates finding 2.1.2 of the main audit report of 19 August 2015 (no. 321-1/2014/49). The main report itself is published with the investment values redacted as business secrets (marked (P)) in its OCR copy, so the post-audit report is the readable source. The comparison is between two revisions of the investment programme, not between the first approved estimate and the final outturn: the record does not show the earlier, much lower estimates of the original 2006 to 2008 investment documents, which this protocol did not read at an official source, and so understates the full growth. NIP 6 is an estimate made while the unit was being built, not a final account. The court's main finding concerns HSE's management of the group's cash flows, and it also found the assumed coal cost (2.25 euros per GJ) and electricity prices unachievable, so the unit's financial case failed on more than construction cost. The unit was built under contract by an Alstom consortium; the delivering institutions are the state-owned HSE and TEล , and the state as guarantor and owner (through AUKN, SOD and SDH), whose oversight the court also faulted. The started year is the main construction contract with Alstom, as commonly reported; it was not read at the cited source. Status is concluded because the unit was built and operates; the record says nothing about its economics after 2016 or Slovenia's coal exit plans.
๐ฑ๐ปLatviasince 2016still operatingbears on Large project delivery
The Latvian State Audit Office found in June 2024 that the estimated construction cost of Rail Baltica's main line in Latvia had risen sevenfold since 2017 to about 8.7 billion euros, that design work due by the end of 2021 was only 54 percent complete and now planned to finish by the end of 2027, and that the Riga Central Station works contracted in 2019 at 430 million euros had risen to 565 million euros.
The transport ministry approved scope changes and cost increases for stations, the Riga node and the main line itself, often without bringing a full cost picture to the Cabinet, while design guidelines kept changing and EU co-financing deadlines kept slipping.
Preconditions
A first cost estimate made before any design guidelines existed
A ministry able to approve scope changes without a single Cabinet-level change process
Financing that depends on successive EU funding calls rather than a committed national budget
Where else it has been used: The joint Baltic review found the same pattern in Estonia and Lithuania, so the overrun is a property of the joint project's governance as well as of Latvian administration.
Limits: The source is a situation review (situฤcijas izpฤte), which the audit office itself describes as a short compilation of facts with no conclusions or recommendations, not a full audit. The 8.7 billion is stated as 'aptuveni' (about) and as a sevenfold rise since 2017, so the 2017 baseline (roughly 1.2 billion) is implied, not printed; it is not used as a metric here. Much of the rise is scope and specification change (the design guidelines were revised more than 100 times after 2018, and the station node went from two tracks to four), so the multiple measures how far the 2017 estimate was from a buildable project as much as delivery performance. The project is not finished: the release says only 4.5 percent of the needed main line financing was available, and the joint Baltic review it summarises puts the whole three-country project at 23.8 billion euros in 2023 against 5.8 billion in 2017, with the first phase now due in 2030. Rail Baltica is a joint Estonian, Latvian and Lithuanian project funded about 85 percent by the EU; this record covers only the Latvian section and the Latvian ministry's management. Status is operating because the project is under construction; a decision to phase or cut the Latvian scope after June 2024 was not checked at an official source. The started year is the 2016 Cabinet approval of the project scope named in the release. The 565 million euro station figure is computed here: the release says the 2019 contract cost of 430 million euros rose by a further 135 million in 2024, and does not print 565.
๐ฆ๐นAustriasince 1998delivered and closedbears on Large project delivery
Austria's federal rail infrastructure company รBB-Infrastruktur AG forecast the Koralmbahn's total cost at 6.139 billion euros at the end of 2023, 282.4 million euros or 5 percent above the 2012 baseline, but missed both the 2018 completion date in the 2004 Koralmbahn contract and the 2023 opening date in the 2012 to 2017 framework plan, according to the Court of Audit, and the line entered passenger service on 14 December 2025.
The federal government funded a new high-speed line through the federal rail framework plan and its state-owned infrastructure company managed the build, holding total cost close to the 2012 estimate while a hard-rock tunnel lot overran in cost and time and pushed the opening back twice.
Preconditions
A rolling multi-year federal rail framework plan that funds the project year by year
A state infrastructure company with project controlling strong enough to absorb a lot overrun inside the total
A supreme audit institution that audits individual construction lots, not only totals
Limits: The cost figure is รBB-Infrastruktur's own forecast, which the court reports; it is a forecast at the end of 2023, not a final account, and the 5 percent increase is measured from 2012, when the project was already well advanced, so it says nothing about growth from the first estimates of the 1990s and 2000s, which the release does not give. The court's audit covered 2012 to 2023. The release says the tunnel lot KAT 2 alone rose 297.75 million euros and took about 31 months longer, with contractor claims above a third of the original contract sum, and that the company paid tunnel stoppages contrary to the contract. The opening date of 14 December 2025 is taken from รBB's own project page (https://www.unsereoebb.at/de/suedstrecke, 'Am 14. Dezember 2025 geht die Koralmbahn in den Fahrgastbetrieb'), not from the court. The full report PDF (https://www.rechnungshof.gv.at/rh/home/home/2025_14_Koralmbahn.pdf) was not opened; the figures come from the release. The started year is approximate and was not read at either source: preparatory works on the line began in the late 1990s and the Koralm tunnel drives later. รBB-Infrastruktur is a state-owned company, so the delivery is institutional, but the record reports cost and time only, not ridership.
๐ฉ๐ฐDenmarksince 2020still operatingbears on Large project delivery
The European Court of Auditors found in 2020 that the Fehmarn Belt fixed link, the 18 km rail and road tunnel that Denmark finances, designs and builds through its state company Femern A/S, had an initial cost estimate of 4.4 billion euros at 2008 prices (5.0 billion at 2019 values) against a latest estimate of 7.7 billion, a 54 percent increase, and that construction planned to start in 2012 for opening in 2018 had started in 2020 with opening expected in 2028.
Denmark took on the whole financing and construction risk of a cross-border tunnel through a state-owned company repaid from future tolls, under a state treaty that limits the neighbouring country to its own access lines.
Preconditions
A state company model already used for the Great Belt and รresund links, financed by state-guaranteed loans repaid from tolls
A bilateral treaty dividing responsibilities, which also left the timetable exposed to the other country's planning law
EU co-financing under the trans-European transport network
Where else it has been used: The same state-guaranteed company model built the Great Belt and รresund fixed links, so this is the third use of the Danish model; the delay this time came from the planning process on the other side of the border.
Limits: The figures are in a PDF: the timeline table is Annex II (PDF page 66) and the cost table Annex III (PDF page 67), with the 5,016 to 7,711 million euro comparison and 54 percent in the table on PDF page 36. The ECA compares the latest estimate with the first political estimate on purpose, so part of the increase reflects scope settled after 2008, and the re-indexation uses the GDP deflator, not construction prices. The ECA's 'transport flagship infrastructure' includes the Danish and German rail access lines, and Germany finances and builds its own access lines, so the 7.7 billion is not a purely Danish cost; the report states that the tunnel itself is financed, designed and built by Denmark (paragraph 23, PDF page 21) and that the fixed link is therefore a Danish project. The figures are from 2019 and 2020; the tunnel is still under construction, and later Danish estimates and the current opening date were not checked at a Danish source, so the 2028 opening is the ECA's estimate at the time. The ECA reports that the construction start slipped largely because of the German planning approval process and litigation, which is outside Danish institutions' control. The EU co-finances the project, and the ECA audits EU spending, not Femern A/S's management.
๐จ๐ฟCzechiasince 2020still operatingbears on Large project delivery
The Czech Supreme Audit Office reported in June 2026 that 41 motorway sections totalling more than 300 km were opened to traffic between 2020 and 2024, after years with almost none (no kilometre in 2014, four in 2018), but that the average cost the state paid per kilometre of motorway opened rose from nearly CZK 275 million in 2020 to CZK 505 million in 2024, an increase of nearly 84 percent.
The transport ministry and its Road and Motorway Directorate prepare motorway sections through long zoning and permitting processes and then build many of them in a short window financed by the State Fund for Transport Infrastructure and EU funds, paying whatever the market price is when the sections finally reach construction.
Preconditions
A state road directorate with a pipeline of permitted sections ready to tender
A transport fund able to raise annual spending sharply
Permitting reform that releases sections held up for years, which is also what bunches them into a high-price period
Limits: The source is the audit office's English press release; the comprehensive report it summarises, based on 13 audits completed from 2020 to 2025, was not opened. Cost per kilometre opened is an average over whatever sections happened to open in each year, so it mixes simple and complex sections (tunnels, bridges, terrain) and nominal prices across a period of high construction inflation; the office says the rise 'cannot be attributed solely' to inflation or complexity, which is a judgement, not a decomposition, and no baseline against outturn is given for any single section. The 2020 figure is stated as 'nearly' 275 million. The same release gives project-level findings that support the schedule side: preparation of the D4 section took the Ministry of Transport and the Road and Motorway Directorate 21 years and was delivered as a PPP whose payments run to 2049, and the directorate had not completed the modernisation of 20 sections of the D1 for CZK 28 billion after 10 years. The record shows delivery volume rising and unit cost rising together; it does not show whether the 2020 to 2024 output was on its planned schedule. The started year is the first year of the audit office's comparison, not the start of the motorway programme.
๐ญ๐ทCroatiasince 2018delivered and closedbears on Large project delivery
Croatia's road authority awarded the 2,440-metre Peljeลกac Bridge to China Road and Bridge Corporation in 2018 for HRK 2.081 billion excluding VAT with a 36-month completion deadline, works started on 31 July 2018, the last steel segment was installed by 28 July 2021, and the government expected the bridge and access roads to open in June 2022, in a project whose eligible costs of 420 million euros were 85 percent financed by an EU grant of 357 million euros.
The government revived a cancelled bridge as an EU cohesion-funded major project, tendered it on price with a fixed deadline, and let it to the lowest bidder, a Chinese state contractor, while EU money covered 85 percent of eligible costs.
Preconditions
An EU cohesion allocation large enough to cover most of a single national megaproject
A feasibility study commissioned by the European Commission that settled the choice of route
A road authority able to run an open international tender and enforce a fixed deadline
Limits: The source is a government news item written by the national news agency Hina at the moment the span was closed, so it carries the government's framing and gives a contract price and a deadline, not a final cost: no final account or audit of the bridge's outturn cost was read. The page puts the total project value at 'โฌ526 billion', an evident typing error for 526 million, and the EU grant figure appears as 'โฌ357' without a unit; the record uses the 420 million euro eligible cost and 357 million euro grant stated elsewhere on the same page, and the European Commission's release of 7 June 2017 (read through the Internet Archive, https://web.archive.org/web/20190715235249/https://ec.europa.eu/regional_policy/en/newsroom/news/2017/06/06-07-2017-commission-approves-eu-financing-of-the-peljesac-bridge-in-croatia) confirms 357 million euros and 85 percent. The 36-month deadline from 31 July 2018 fell at the end of July 2021, when the span was joined; the expected June 2022 opening was a forecast on the page, and the actual opening date (reported as 26 July 2022) was not read at an official source, so the record can say only that the structure was joined on its contract deadline and opened about a year later with its access roads. The project had an earlier history the page omits: a first construction contract from 2007 was cancelled by a later government, according to press reports not used as a source. The bridge was built by a Chinese state-owned contractor, so the delivery attributable to Croatia is the procurement, financing and project management.
๐ธ๐ฐSlovakiasince 2015still operatingbears on Large project delivery
Slovakia's state decommissioning company JAVYS, which the European Court of Auditors reported in 2016 as holding the Bohunice V1 decommissioning end date at 2025 with a cost estimate of 1,239 million euros, reported in its 2025 annual report that the final decontamination and demolition contract signed on 18 December 2024 runs 60 months, putting the end of decommissioning at the end of 2029, at a total estimated cost of 1,220 million euros, with 66 decommissioning projects completed by 31 December 2025.
A state company created to decommission shut-down reactors runs dozens of EU-funded projects under a detailed decommissioning plan monitored by the European Commission and the EBRD, and reschedules the end date when the final demolition contract is procured.
Preconditions
A dedicated state decommissioning company with its own nuclear licences
EU accession commitments that fund most of the cost through a ring-fenced programme
A regulator able to license each decommissioning stage separately
Where else it has been used: Lithuania's Ignalina decommissioning, funded under the same EU arrangement, slipped much further (ltu-ignalina-decommissioning).
Limits: JAVYS reports on its own programme; the figures are in a PDF, the 2029 end date and the 60-month contract of 18 December 2024 on PDF page 17, the 1,220 million euro estimate and the 735 million EU share on PDF page 19, and the 66 completed projects on PDF page 20. The baseline comes from a different publisher: the European Court of Auditors' Special Report 22/2016 (https://www.eca.europa.eu/Lists/ECADocuments/SR16_22/SR_NUCLEAR_DECOMMISSIONING_EN.pdf) says the Slovak end date 'has remained at 2025' (paragraph 64, PDF page 40) and gives estimated costs of 1,146 million euros in 2011 and 1,239 million in 2015 (Table 8, PDF page 46). The two cost figures are not on the same basis, so the record does not claim the cost fell; it claims cost was held roughly flat while completion slipped about four years. Decommissioning is a stretch for 'major public infrastructure projects': it is a large state engineering programme that builds waste facilities and demolishes a plant rather than a new asset, and it is mostly paid for by the EU (735 million euros of the 1,220 million, through the EBRD-managed BIDSF fund and the Slovak Innovation and Energy Agency). The started year is 2015, the start of the second decommissioning stage under the nuclear regulator's licence of 23 December 2014; the first stage began earlier. Whether the 2029 date holds is not known.
๐ฑ๐นLithuaniasince 2010still operatingbears on Large project delivery
The European Court of Auditors found in 2016 that Lithuania's Ignalina nuclear power plant decommissioning had pushed its end date back by a further nine years since 2011 to 2038, that the interim spent fuel storage facility was ten years behind the 2005 decommissioning plan, and that the estimated decommissioning cost had risen 15 percent from 2,930 million euros in 2011 to 3,376 million in 2015, and 67 percent since the 2010 estimate.
A state enterprise decommissions a large Soviet-designed plant under EU accession commitments, building fuel and waste storage through international contracts paid by an EU fund, and its end date moves each time a key facility is delayed.
Preconditions
EU accession-linked funding that covers most of the cost
Reliable design records of the plant being dismantled, which were missing here
Contract management capacity strong enough to resolve disputes with foreign suppliers quickly
Where else it has been used: Slovakia's Bohunice V1, funded under the same arrangement, held its end date until 2025 and then slipped to 2029 (svk-bohunice-v1).
Limits: The figures are in a PDF: the 2038 end date and the nine-year postponement in paragraph 60 (PDF page 38), the ten-year delay of the interim spent fuel storage facility in Table 6, the 67 percent rise in paragraph 73 (PDF page 44) and the 2011 and 2015 estimates in Table 8 (PDF page 46). The report is ten years old and the current end date and cost estimate were not checked at a Lithuanian source; the Lithuanian audit office's site (valstybeskontrole.lt) returned a Cloudflare challenge. The ECA audits EU spending, and the EU funds most of the programme, so the figures describe a jointly financed delivery carried out by the state enterprise Ignalina NPP. Decommissioning is a stretch for 'major public infrastructure projects': it builds waste and fuel storage facilities and dismantles a plant, and the report says delays came partly from contractual disputes with a foreign contractor and from poor information on how the Soviet-built plant was constructed. The started year is approximate: 2010 is the base year of the ECA's cost comparison, and the report does not give a start date for decommissioning itself.
๐ท๐ธSerbiasince 2010delivered and closedbears on Large project delivery
Serbia's state road company Koridori Srbije built 45.5 km of Corridor X motorway on the E-75 towards North Macedonia and the E-80 Dimitrovgrad bypass towards Bulgaria under a loan approved in July 2009, finishing at a total project cost of EUR316.2 million against an appraisal estimate of EUR318.2 million, but taking about 10 years from approval to full completion against a planned 6.5, with the loan closing in September 2019 instead of December 2015.
A dedicated state motorway company procured large lots under development bank rules, held the cost line by re-bidding failed contracts and drawing an additional loan, and accepted years of delay rather than scope cuts.
Preconditions
A single implementing company for the corridor with procurement and supervision staff
Contingencies in the appraisal estimate large enough to absorb variation orders
Government budget for expropriation, whose absence caused the early delays
Limits: The figures are the World Bank's, in a completion report on a project it financed and rated, written with data from the implementing agency. The two costs are in para. 11 and Table 2 (printed page 8) and para. 54 (page 19); the 10 years against 6.5 and the closing dates are in para. 32 (page 13) and the datasheet (page 2). Staying on budget was helped by a 2016 additional loan of EUR35 million that covered overruns (datasheet and para. 58), and the construction cost alone rose 21 percent over the initial contract prices, so 'on budget' means within an appraisal estimate that carried contingencies, not within contract prices. The Dimitrovgrad bypass contractors were terminated twice and its works finished in November 2019 against an April 2012 contract date (Table 6). The record covers only the World Bank financed sections of a longer corridor co-financed by EIB, EBRD and others, whose cost and schedule are not reported here. Traffic in 2018 was 46 percent of the appraisal forecast on the E-80. The start year is the first works contract (April 2010, Table 6).
๐บ๐ฆUkrainesince 2016delivered and closedbears on Large project delivery
Ukraine's road agency Ukravtodor rehabilitated or upgraded 156.3 km of the M-03 Kyiv to Kharkiv and M-12 corridors under a US$560 million loan approved in November 2015, after slow progress led the government in 2017 to cancel US$222.2 million of the loan, dropping three M-03 bypasses and two long-term maintenance contracts, so the loan closed at US$337.8 million in September 2022, nine months after its original date.
The national road agency delivered the widening of a corridor it had already been rebuilding with the same lender, and when bypasses stalled on land acquisition the government cancelled them rather than extend the loan.
Preconditions
An experienced implementation unit from earlier corridor loans
Land acquisition and resettlement procedures that can be completed before works are tendered
A lender willing to cancel and reallocate rather than hold undisbursed funds
Limits: The figures are the World Bank's, in a completion report on its own loan, and the loan amount is not the full cost of the works; no figure for the cost of the delivered roads against an appraisal estimate per km was used. The cancellation and its reasons (land acquisition not started at Poltava, community objections at Kopyly, a realignment at Chutove) are in paras. 23 and 26 to 27 (printed pages 10 to 12); the 156.3 km is in para. 41 (page 16); amounts and dates are on the datasheet (page 2). In 2019 disbursement was about 11 percent with over 60 percent of the period elapsed, and some lots were moved in from a parallel World Bank road project, so the delivered kilometres mix two programmes. Russia's invasion in February 2022 stopped the civil works under force majeure seven months before closing, so the final delay is not all the agency's. The record shows a scope cut, not a cost overrun, and says nothing about the separate Great Construction programme.
๐ง๐พBelarussince 2016delivered and closedbears on Large project delivery
Belarus's Ministry of Transport and Communications widened 154 km of the M6 Minsk to Grodno highway to dual carriageway under a loan approved in December 2014, against 86 km planned, after civil works for the planned sections were awarded for less than half the estimate; the corridor component cost US$242.15 million against US$259.28 million at entry, and the loan closed in June 2022 against an original June 2020.
The transport ministry tendered the corridor in lots to joint ventures of local state-owned contractors, and when bids came in far below estimate it used the savings to add 67 km of the same corridor without new financing.
Preconditions
Complete designs before effectiveness, so works could start as soon as procurement closed
A domestic road construction industry with capacity and equipment
A cost estimate high enough that competitive bids produced savings rather than claims
Limits: The figures are the World Bank's, from a completion report on its own loan, with costs reported by the borrower. The component costs are in Table 2 (printed page 9), the 154 against 86 km in para. 32 (page 14), the savings in para. 52 (page 21) and the dates on the datasheet (page 2). The low cost was helped by a government decision to exempt the works from VAT, so part of the saving is a transfer within the state, not efficiency. The actual cost excludes design, supervision and land acquisition, which the government paid separately. Works were not contracted until July 2016, more than a year after effectiveness, because of a dispute over whether state-owned Belarusian contractors were eligible, and the road safety centre was incomplete at closing. The report was written in 2022, after which World Bank engagement with Belarus stopped, so later maintenance of the road is unknown.
The Federation of Bosnia and Herzegovina's public road company PC Roads FBH delivered 76.9 km of road works, including a 35 km upgrade of the Neum to Stolac road, under a project approved in August 2016, but rehabilitated only 41.75 km of main roads against an appraisal target of 178 km, which was cut to 42.23 km in 2021, and the closing date moved from December 2021 to September 2023.
The entity road company took a multilateral loan for a list of sections whose designs did not yet exist, then cut the list to what designs, land and budget allowed and concentrated on one strategic coastal link.
Preconditions
Detailed designs and land acquisition finished before the results framework is fixed
An entity-level road agency able to borrow with state guarantee
Co-financing that can be raised when costs move
Limits: The figures are the World Bank's, from a completion report on a project it co-financed with the EIB. The 178 and 42.23 km targets are in para. 55 (printed page 19), and the 41.75 km achieved against the revised 42.23 km target is in the outcome table (Table 4, page 13), the 76.9 km and the Neum to Stolac upgrade in paras. 30 to 31 (page 11), the cost per km in para. 45 (page 17) and the dates on the datasheet (page 2). The Bank attributes the cut to designs that were only 1.0 km complete at entry, so the 178 km target may have been an error of preparation rather than a delivery that shrank; the report says rehabilitation cost per km came out more than five times the plan for the same reason. Total project cost rose from US$101.3 million to about US$105.5 million, mostly as an EIB increase. Land acquisition and COVID-19 delayed the Crnaja tunnel and led to sections being cancelled. The project covers one of the country's two entities, not Bosnia and Herzegovina as a whole.
๐ฒ๐ฉMoldovasince 2016delivered and closedbears on Large project delivery
Moldova's State Roads Administration upgraded 149.3 km of unpaved local roads to paved standard under a project approved in October 2015, after unit costs doubled from US$230,000 to about US$415,000 per km and contracts on 42.3 km were terminated, so a 2019 restructuring cut the target from 300 km to 150 km and the project closed in July 2023 instead of March 2021.
The road administration designed for a low-cost paving technique, failed to attract bidders able to build it, then redesigned for asphalt at twice the unit cost and halved the length.
Preconditions
A contractor market able to bid on the chosen technique
Bid and performance security terms small firms can meet
Maintenance budgets matched to the pavement type chosen
Limits: The figures are the World Bank's, from a completion report on its own credit, which rated it. The 300 to 150 km cut and the unit cost doubling are in para. 54 (printed page 26), the terminated contracts on 42.3 km in para. 39 (page 21), the 149.3 km in the results table (page 16) and the dates on the datasheet. The doubling followed a switch from a cheap surface treatment that local contractors could not build to conventional asphalt, so part of the cost rise bought a more durable road. These are local roads, a programme of many small contracts rather than one major build, and the US$102.5 million project is small against Moldova's national road spending.
๐ฆ๐ฑAlbaniasince 2018delivered and closedbears on Large project delivery
Albania's Albanian Development Fund rehabilitated about 70 km of regional and local roads under a US$50 million loan approved in April 2018, against 55 km planned, after civil works were awarded at up to 48 percent below estimate, with the closing date extended by 18 months to June 2024.
A development fund that has implemented donor-financed roads since 1993 ran competitive bids among local firms, and used the savings to extend the list of roads.
Preconditions
An implementing agency with long donor procurement experience
Enough local contractors to make bidding competitive
A lender willing to let savings extend scope
Limits: This is a small programme of regional road contracts, not a major national build, and it is filed because no official cost or schedule outturn for a larger Albanian project (the Arbri road concession, the Durres to Kukes motorway) was found; it is weak evidence for the construct. The figures are the World Bank's, from a completion report on its own loan. The 55 and 70 km are in para. 12 and para. 58 (printed page 16); the report attributes the savings partly to the Albanian Development Fund's own cost estimation method, which it says overestimated costs, so 'under budget' partly measures a padded estimate. The 18-month extension is on the datasheet and in para. 21; the report cites COVID-19, the 2019 earthquake and land acquisition at Pogradec. 'About 70 km' is approximate in the source, and the source is inconsistent: the summary results table reports 62 km rehabilitated against the 55 km target, while the annex results table and the text give 70 km as of October 2024. The start year is the approval and signing year (2018); the first works date is not stated on the pages read.
North Macedonia's Public Enterprise for State Roads contracted Sinohydro in November 2013 to build the 56.7 km Kicevo to Ohrid section of motorway A2 for EUR411 million by January 2018; by December 2021 five annexes had moved completion to December 2023 and raised the contract value by EUR187 million (45 percent) to EUR598 million, and the State Audit Office found works 72 percent complete by value in August 2023.
The government named a Chinese contractor by statute under a bilateral loan agreement and started construction on an old design, then absorbed the redesign of tunnels, cuttings and landslide sections through successive annexes that raised the price and moved the deadline.
Preconditions
A loan-tied contractor selection that skips competitive tendering
A basic design old enough to be out of step with ground conditions
Parliamentary approval of annexes rather than a fixed-price baseline
Limits: The figures are in a PDF in Macedonian, in the key audit matter section 5.1 (printed pages 25 to 27); the 45 percent increase is also stated in the summary (printed page 4). The road was still under construction when the audit was written, so the record shows an overrun in progress, not an outturn; later annexes and deadlines reported in the press (a further increase and a 2027 deadline) were not checked against an official document. The EUR598 million is the contract value; the audit puts total investment, including expropriation, interest and exchange differences, at about EUR575 million spent by November 2023, and financing at US$685 million in two Chinese loans. The audit attributes the overrun to a basic design drawn up between 2002 and 2010 that did not match the terrain, giving 42 variations, which is the auditor's reading. The contractor was named by a 2013 law rather than chosen by open tender, so the record also reflects a procurement choice. 'Operating' means the build is still running, not that the road is open.
๐ญ๐บHungarysince 2006delivered and closedbears on Large project delivery
Hungary's State Audit Office reported in September 2010 that the first section of Budapest metro line 4, approved by the capital in September 2004 at a cost of HUF236.5 billion with a deadline of end 2009, was forecast to cost HUF370 billion, 56.4 percent more, and to open in the second half of 2013.
The capital set a deadline in its 2004 permit before permits and land were secured, signed works contracts in 2006 with penalty deadlines it then had to relax, and redesigned stations after an architectural competition while construction was under way.
Preconditions
A financing agreement between city and state that fixed the budget before design was complete
A permit deadline set without reference to the state of technical preparation
Contracts whose penalty dates were linked, so one slip moved all of them
Limits: The figures are in a PDF in Hungarian (summary findings, printed pages 12 to 13, with a chart of planned dates and costs); they are a mid-construction forecast, not the final outturn, and the line's opening in March 2014 and its final cost are not in this source. The auditor attributes part of the increase to scope added after the permit: EU-supported municipal works around stations (HUF39.5 billion), a single-vault station redesign (HUF44.8 billion) and reserves of about 14 percent, so not all of the 56.4 percent is overrun on the same scope. The report's chart puts the opening date in the 1996 feasibility study at 2005, so the delay looks larger against that baseline. The investor was the City of Budapest through its project directorate, with central government paying a large share under a 2004 agreement, so the delivery is a joint municipal and national one. Later OLAF findings on irregularities were not used.
๐ง๐ฌBulgariasince 2008delivered and closedbears on Large project delivery
Bulgaria's Road Infrastructure Agency rehabilitated national roads under a EUR144 million project approved in June 2007, but after budget cuts in 2010 cancelled three lots on about 87 km where works had not started, reducing the length from 450 km to 307 km, and closed in June 2013, two years late, at EUR97.1 million, 67 percent of the appraisal estimate.
The road agency contracted nine regional rehabilitation lots on a loan, then, when the national budget could not cover its co-financing and payments, cancelled the lots not yet started and stretched the rest over two more years.
Preconditions
Co-financing from a national budget exposed to a fiscal shock
Contracts signed before the budget for their full duration was secured
A lender willing to restructure and extend rather than close
Limits: The figures are the World Bank's, from a completion report on its own loan, which reproduces the agency's own implementation report as an annex. The 450 to 307 km reduction and the cancelled lots are in section 1.6 (printed page 4); the costs in Annex 1 (page 19); the original and revised closing dates on the datasheet. Spending fell because scope was cut, not because works came in cheap, so the 67 percent is a measure of a shrunken programme, not of savings. The cause the report names is the government's fiscal restraint after the 2008 to 2009 crisis, with 2010 budget funds spent on 2009 obligations, and only 8 to 9 percent of the loan disbursed by mid-2009. These are rehabilitation lots on existing class I to III roads, not a new motorway, and the record says nothing about the EU-funded motorways built in the same years.
๐ช๐ฌEgyptsince 2009still operatingbears on Large project delivery
The Egyptian Electricity Holding Company built the 1,300 MW Ain Sokhna supercritical thermal power plant, with its first unit entering commercial operation on 3 March 2015 and the second on 4 June 2015, at a final cost of US$1,159.4 million against a base cost of US$2,139.6 million estimated at appraisal in 2009.
The state power holding company procured a large thermal plant in a few big packages co-financed by four lenders, priced the appraisal estimate conservatively, and redeployed the savings into grid works.
Preconditions
A state utility with an engineering department experienced in large thermal plants
Co-financing from several development lenders committed before bidding
A market for supercritical boilers and turbines competitive enough to bid below estimate
Limits: The World Bank, which lent US$600 million of the cost, wrote the report. The costs are in Table 1 and para. 23 (ICR pages 8 to 9), the commercial operation dates in para. 20 of the annex on design and implementation; the report says construction was 'initially delayed' but finished 'about when expected' and gives no original target date for commercial operation, so schedule performance rests on that judgement and on the 2015 dates. The ICR does not explain why the plant cost 46 percent less than estimated; a 2013 restructuring had already cut the estimate to US$1,655.3 million, which suggests the 2009 estimate was high rather than that delivery was unusually efficient, and the savings went into a transmission component. That transmission component ran late: it was not finished at the extended loan closing of 31 December 2018 (original closing 31 December 2015), and its final cost was not known. The plant was designed to burn heavy fuel oil as well as gas, which the report says raised its cost and lowered its efficiency against a combined-cycle plant. The status rests on the report's account of operation in 2015 to 2018, not on a current source.
๐ฎ๐ถIraqsince 2010delivered and closedbears on Large project delivery
Iraq's Ministry of Electricity set out in 2007 to restore 400 MW at Hartha power station units 2 and 3 by December 2010 under the Emergency Electricity Reconstruction Project, but when the project closed on 30 June 2015 no capacity had been restored, the works were about 70 percent complete, the cost estimate had risen from US$150 million to US$187.62 million with a US$98.97 million financing gap, and the contractor withdrew from the site.
The ministry tried to finish a rehabilitation begun under sanctions by contracting one turnkey supplier to install equipment bought a decade earlier, and when disputes and budget shortfalls stalled the works, it could neither fund a variation nor terminate.
Preconditions
Legacy equipment that tied the contract to bidders familiar with its design
A government budget able to fund variations and security costs, which was missing after the 2014 oil price fall
A contract with clear termination and dispute rules, which the report says were lacking
Limits: The World Bank wrote the report and rated both itself and the borrower unsatisfactory or moderately unsatisfactory, and it says blame for the delays was shared between the ministry and the contractor (para. 53, PDF page 29). The 0 MW and 70 percent figures are in the results table (PDF page 7); the cost table and the US$98.97 million gap are in Annex 1 (PDF page 43); the halt is para. 46 (PDF page 25: 'works have practically stopped'). The cost estimate excludes about US$19 million the government paid for contractor security and an unsettled price adjustment. Much of the failure came from the design: the project had to reuse equipment bought in 1999 under the Oil for Food Programme, which narrowed bidding and bred disputes, and the security situation from 2014 raised costs. The status is 'concluded' because the financed project closed and its result, no capacity restored, stands; it is a delivery failure, not a reversal, since nothing delivered was later lost and the report does not say the rehabilitation was terminated. Whether the ministry later completed the units with its own money was not read at any Iraqi source. The units are part of one station, a modest project by value; the record evidences delivery failure on a rehabilitation, not on a new build.
๐ฑ๐งLebanonsince 2012operating below its peakbears on Large project delivery
Lebanon's Council for Development and Reconstruction and the Beirut Mount Lebanon Water Establishment built most of a system to bring 250,000 cubic metres a day from the Awali to 1.2 million people in Greater Beirut, but when the World Bank loan closed on 30 April 2025, 106 months after the original June 2016 closing date, no additional water had been delivered, the 24/7 supply reached in a 20,000-customer pilot zone in 2018 had not been maintained, and component costs had risen from US$370 million at approval to US$561.7 million.
The reconstruction council and the regional water utility split a gravity conveyance, treatment and distribution system between them, financed the start with a World Bank loan, and depended on utility surpluses and a second loan for the rest; when the utility's revenue and the state's finances collapsed, the scheme stopped short of delivering water.
Preconditions
A utility with cash surpluses able to co-finance distribution, which disappeared with the 2019 crisis
Parliamentary ratification of loans in time, which failed for the 2018 additional financing
Contract terms that allocate contractor insolvency risk, which the report says were missing
Limits: The World Bank wrote the report and rated the outcome Highly Unsatisfactory and its own performance Moderately Unsatisfactory. The 106 months are in para. 17; the zero water delivered, the 75, 84 and 89 percent completion of tunnels, pipelines and treatment plant and the 20,000-customer pilot are in Table 2 and paras. 29 to 31 (ICR pages 8 to 9). The US$561.7 million is the project budget at closing printed in Table 3 of the report against US$370 million at appraisal, and matches the sum of the three components in Annex 3 (US$434.0, 112.7 and 15.0 million; page 31) against 294.0, 61.0 and 15.0 at approval. Lebanon's 2019 financial collapse, COVID-19, the 2020 port explosion and a contractor's insolvency all fall inside the delay, so it is not a clean measure of institutional capacity in normal times, and a US$90 million additional financing approved in 2018 lapsed because Lebanon did not make it effective. The erosion is the pilot zone's 24/7 supply, reached in 2018 and lost; that is a small part of the project, and the main failure is non-completion, not loss. A follow-on Second Greater Beirut Water Supply Project was approved on 14 January 2025 to finish the works, so the delivery may still complete. The related Bisri dam, financed under a separate project, was partly cancelled before construction (footnote in para. 8) and is not counted here.
๐ฏ๐ดJordansince 2004delivered and closedbears on Large project delivery
Jordan's Ministry of Public Works and Housing built the 40 km first phase of the Amman Ring Road under the Amman Development Corridor Project, whose estimated total cost rose from US$160.6 million at appraisal in 2004 to US$455.97 million in March 2013, with the road works alone rising from US$106.92 million to US$279.87 million, while the loan closing date moved from June 2009 to December 2012 and one of three sections was 65 percent complete at closing.
The public works ministry split a 40 km expressway around Amman into three sections financed by three lenders, acquired land late and at rising prices, and absorbed design changes during construction with additional financing rather than rescoping.
Preconditions
A land acquisition budget fixed before prices rose along an announced alignment
Three lenders with separate procurement and supervision rules on one road
A ministry able to fund overruns from the budget, which ran short in the final section
Limits: The World Bank, which financed one of three road sections (with the Arab Fund and the EIB financing the others), wrote the report and rated the outcome, its own performance and the borrower's Moderately Unsatisfactory. The costs are in Annex 1 (printed page 18): land acquisition rose from US$40.9 million to US$158.43 million and is a large part of the overrun, and the latest estimate includes US$42 million of payments due after closing to finish Section 1 and settle bills. Against the 2008 additional-financing estimate of US$343.18 million the overrun is 32 percent, so much of the growth was already recognised mid-project. The 65 percent figure for Section 1 is in the results table; the report says in para. 48 that the road remained uncertain of reaching operation 'when nearly twice the expected construction time has passed', and that the inland port and customs depot were 'gradually dropped'. The text is a scanned 2013 report with OCR errors in the data sheet. When the whole road opened was not read at an official source; 'concluded' means the financed project closed with the road mostly built.
๐ฌ๐ชGeorgiasince 2016delivered and closedbears on Large project delivery
Georgia's Roads Department completed most civil works on the Zemo Osiauri to Chumateleti and Chumateleti to Khevi sections of the East-West Highway across the Rikoti Pass by the project's closing on 30 June 2025, four and a half years after the original loan closing date of 31 December 2020, and the government's own financing rose to US$91.03 million against US$44 million planned for the enlarged project (US$24 million at approval plus US$20 million under the 2017 additional financing).
The Roads Department carried the national east-west highway over its hardest mountain pass in contracts financed by the World Bank and the EIB, and when contracts failed or were re-tendered the government covered the gap from its own budget instead of cutting scope.
Preconditions
A sequence of earlier highway projects that built the Roads Department's contract management
Lenders willing to co-finance tunnels and bridges in seismic mountain terrain
Budget room to nearly quadruple counterpart funding
Limits: The World Bank wrote the report and rated the outcome Moderately Satisfactory and efficiency Modest. The financing figures are in the data sheet and Table 2 (ICR page 4); the closing-date changes in the restructuring table (paras. 10 to 11) and para. 23. Much of the extra time and government money followed a 2017 decision to add the 11.6 km Chumateleti to Khevi link, co-financed by an EIB loan of EUR 67.47 million, so the plan the outturn is set against grew mid-project; Table 2's component 1 estimates (parent plus additional financing, US$273.48 million including EIB) against actuals (US$359.72 million) give a 32 percent overrun by my own sum, which the report does not state. The report attributes the 18-month extension of 2023 to re-tendering, COVID-19 and mountain construction, and the parallel Fourth East-West Highway project's ICR (2020) records that a Zemo Osiauri to Chumateleti lot was terminated in December 2018 and re-bid under this project. Most civil works were complete at closing, with residual works finished afterwards on government money; the date of full opening was not read at an official source.
๐ฆ๐ฒArmeniasince 2010delivered and closedbears on Large project delivery
Armenia's Ministry of Transport, Communication and Information Technologies reconstructed about 30 km of the M1 Yerevan to Ashtarak and M2 Yerevan to Artashat roads under the first tranche of the ADB-financed North-South Road Corridor programme for an implementation cost of US$68.70 million against US$68.0 million at appraisal, but completed the project on 16 June 2016 against a planned 31 December 2013, after one round of bids was rejected and the M1 section was cut from 18.4 km to about 10 km.
The transport ministry used a multitranche ADB facility to rebuild the national north-south corridor section by section, adjusting each tranche's scope to land acquisition and fiscal limits instead of extending its budget.
Preconditions
A multitranche facility that lets scope move between tranches
A project management unit that also prepares designs and tenders for later tranches
Government readiness to finance gaps when bids come in above estimate
Limits: ADB, the lender, wrote the report. The completion dates and costs are in the basic data table at the front of the PDF (pages 2 to 4); the scope changes are paras. 8 to 16. The 30-month slip from December 2013 to June 2016 is my subtraction from the two dates; the loan closed on 23 February 2018 against 30 June 2014 in the loan agreement. Cost held near appraisal only because the scope moved: the M1 section was cut to about 10 km because of land acquisition, 2 km was transferred to tranche 2 when Armenia hit fiscal limits, and 19.6 km of the M2 was added, so the outturn buys different roads; the combined tranche 1 and 2 works contract was US$10 million above the engineer's estimate and the government undertook to fund overruns (para. 14 and footnote 13). This is the first and smallest tranche of a programme to rebuild a 550 km corridor; later tranches, including the Talin to Gyumri works, were not read, and the programme's overall delivery is not evidenced here.
๐ฆ๐ฟAzerbaijansince 2009delivered and closedbears on Large project delivery
Azerbaijan's water agencies built water supply and sanitation systems in eight regional (rayon) centres under the Second National Water Supply and Sanitation Project, after a 2012 restructuring cut the scope from 21 rayons to 8 because of cost increases, while the total project cost rose from US$560 million to US$641.8 million and the loan closed on 30 June 2019 instead of 28 February 2013.
The government launched a national programme to rebuild water systems in all regional centres on rough cost estimates, financed slices of it with World Bank loans, and when real costs emerged, cut the number of towns and stretched the timetable instead of finding the money.
Preconditions
Feasibility studies before approval, which were missing here
An implementing utility not already running a dozen donor projects at once
A lender willing to refuse additional financing and force a downsizing
Limits: The World Bank wrote the report and rated the outcome Moderately Satisfactory. The financing and dates are in the data sheet (ICR pages 1 to 2); the cut from 21 to 8 rayons, the reasons and the closing extensions are in Box 1 (page 8), and footnote 7 attributes the reduction to cost increases and unanticipated investments. The approval cost and the final cost cover different scopes: the original estimates for 43 rayons across both national projects were 'not supported by thorough technical studies', so the 2008 figure was not a real baseline, and a US$150 million additional financing in 2014 paid for design changes required by new government policy (legalised dwellings, EU standards), not only for overruns. Implementation moved from AzerSu to the irrigation agency SAWMA in 2009 and then to its successor company after mis-procurement in Nakhchivan, so 'the country's institutions' here are several, and the record does not separate their performance. People served and service quality are in the report but are not the subject of this record.
๐ฐ๐ฌKyrgyzstansince 2010delivered and closedbears on Large project delivery
The Kyrgyz Ministry of Transport and Communications rehabilitated 50 km of the Osh-Batken-Isfana road corridor, against about 30 km planned at approval in 2009, for a total project cost of US$52.46 million against US$30.72 million estimated, over four years and ten months instead of three, closing on 31 December 2014 instead of 30 June 2013.
The transport ministry's implementation group rehabilitated a corridor section on a World Bank credit and, delivering on budget, was given two additional financings to extend the works along the same road.
Preconditions
An implementation unit with a track record on other donors' road projects
A lender that rewards delivery with additional financing on the same corridor
A corridor whose remaining sections were already designed
Limits: The World Bank wrote the report and rated the outcome Satisfactory. The costs are in Annex 1 (PDF page 36), the 30 km plan in para. 1.5.2, the 50 km outturn in para. 3.2.2, and the schedule in para. 1.7.2. Most of the cost rise followed added scope: two additional financings added road length and the repair of about 350 km of Osh urban roads, both paid from the road component (US$27.29 million estimated, US$50.27 million actual), so no cost per kilometre of the corridor can be computed from the report and the record cannot separate overrun from added scope. The 18-month extension also followed the added scope. This is a 50 km rehabilitation, small against the country's network, and it says nothing about larger state-financed builds; no Kyrgyz audit was read.
๐ฎ๐ทIransince 2000delivered and closedbears on Large project delivery
The Tehran Sewerage Company built interceptors, trunk mains, tunnels and a wastewater treatment plant under the first phase of the Tehran Sewerage Project at a latest estimated cost of US$359.22 million against US$341.45 million at appraisal in 2000, with the treatment works costing 189 percent of their estimate and the loan closing on 30 June 2008, two years late, before the plant and the eastern trunk main were finished.
The capital's sewerage company built the network in parallel with a design-build-operate contract for the treatment plant, letting the network race ahead on cheap domestic pipe while the plant ran late and over budget under inflation and a foreign-currency contract.
Preconditions
A domestic pipe industry able to supply at falling prices
Counterpart funds indexed for inflation, which were short here
A foreign-currency contract hedged against a weakening rial, which it was not
Limits: The World Bank, which lent US$145 million, wrote the report and rated the outcome Satisfactory. The component costs are in Annex 1 (PDF page 36); the US$28.7 million funding shortfall, high inflation (12 to 24 percent a year) and a euro-denominated plant contract as the cause of the two-year extension are in para. 50 (PDF page 20). The total stayed near the estimate only because interceptors and laterals came in at 61 percent of estimate while the trunk main (178 percent) and treatment works (189 percent) overran, so the aggregate hides the overrun on the hardest parts. The plant and the eastern trunk main were unfinished at closing, and the report says only that the company had resources to finish them; completion was not read at a current source. The figures are 2000 to 2008 and say little about Iranian delivery under sanctions since.
๐ท๐บRussiasince 2014delivered and closedbears on Large project delivery
Russia's Accounts Chamber found in June 2020 that the state company Avtodor's Central Ring Road project, whose first stage was originally to be built in 2014 to 2018 and was now scheduled for 2014 to 2021, had risen in cost by 2.6 billion roubles in 2019 to 315.9 billion roubles, and that launch complexes 1 and 5 were only 49 and 61 percent complete on 1 January 2020.
The federal road company procured a ring road around Moscow as long-term investment concessions, financed by the National Wealth Fund, budget subsidies and investors, and as works lagged the government extended deadlines and the Accounts Chamber audited the project year by year.
Preconditions
A sovereign wealth fund able to buy the state road company's bonds
Concession contracts with enforceable milestones, which the audits found weakly policed
An audit office that publishes a check of the same project each year
Limits: The live ach.gov.ru did not answer a script today, so the cited copy is the Internet Archive's of 15 September 2020. The page is the audit office's summary of its check, not the full report, and it does not state the original cost, so the record shows cost growth in one year, not the full cost overrun against an approved baseline; the schedule baseline is in the page's footnote, which gives the construction period of the first stage as 2014 to 2021, originally 2014 to 2018; the Chamber's 5 August 2015 release on the same project (archived at https://web.archive.org/web/20210717054203/https://ach.gov.ru/news/sredstva-fonda-nacionalnogo-blagosostoyaniya-napravlennye-na-stroitelstvo-ckad-ne-ispolzuyutsya-v-po-22619) says construction of complexes 1 and 5 had not started a year after their agreements were signed and that National Wealth Fund money sat unused in a bank. The road is a concession-based project of a state company with private investors (33.8 billion roubles of 186.5 billion invested by end 2019), so delivery is partly private. The 2014 start is the year of the first concession agreements cited by the 2015 release. Status 'concluded' assumes the ring opened in the years after the audit; no official source for the final opening date or final cost was read, which is the record's main weakness.
Papua New Guinea's Department of Works restored or maintained 600 km of national roads against a 640 km target under the Road Maintenance and Rehabilitation Project, closing in June 2012 five and a half years after the original December 2006 closing date, with national road works under the original loan costing US$35.62 million, 158 percent of their US$22.52 million appraisal estimate.
The works department ran a lender-financed maintenance and restoration programme on national and provincial roads, absorbed a cost surge on roads by cutting bridges, and extended the programme twice.
Preconditions
Local contractors able to bid for maintenance packages
Provincial governments with budgets to co-fund their share
A lender willing to restructure scope inside a fixed envelope
Limits: The component costs are in Annex 1 (project cost by component, original loan) and the dates and road lengths on the datasheet (sections B and F), read from the plain-text version. The appraisal figure excludes VAT and the actual includes it, so part of the 158 percent is a tax difference the table itself flags. Total project cost was on budget (US$58.21 million against US$58.78 million) because bridge components were cut, so the overrun on roads was paid for by building fewer bridges: provincial bridges spent 11 percent of their estimate. The ICR attributes the road cost rise to the construction boom from the PNG LNG project and to sealing more gravel roads than planned. The World Bank rated the outcome Satisfactory, and the programme is maintenance and rehabilitation rather than one new structure. The cost comparison covers the original IBRD loan only; a 2007 IDA additional financing scaled the programme to two more provinces and extended it, and its national road works also overran, at US$40.86 million against US$28.72 million (142 percent, Annex 1 table b), so the June 2012 closing reflects added scope as well as delay. The government paid US$2.28 million of the US$25.07 million counterpart planned at appraisal, with provinces and other donors covering part of the gap.
๐ฑ๐ฆLaossince 1997delivered and closedbears on Large project delivery
Laos's Ministry of Communication, Transport, Post and Construction upgraded about 200 km of Road No. 13 between Banlak 35 and Pakse to a bituminous standard, completed in March 2002, at US$157,785 per km against an appraisal estimate of US$209,670 per km, and the project closed on its original June 2003 date at US$63.78 million against US$68.92 million appraised.
The transport ministry upgraded a 200 km trunk road under two international contracts and ran a multi-year routine and periodic maintenance programme alongside it.
Preconditions
International competitive bidding that attracted capable contractors
A maintenance fund to protect the upgraded road
Grant finance from a second donor for bridges
Limits: The unit costs are in section 4.3 of the ICR and the totals in Annex 2 (printed page 30), read from the plain-text version; the record states the two published unit costs and does not compute a percentage. The saving is on unit cost, not on the road works line: Annex 2 shows Roads No. 13 and A-1 at US$34.58 million against US$33.45 million appraised even though the A-1 bypass was removed and financed under another project, and the project total came in under appraisal only because physical and price contingencies of US$9.96 million were not needed (actual US$63.78 million against a US$58.96 million baseline before contingencies). The Annex 3 table labels both unit costs 'Calendar 2002 US$ per km' and puts the contract cost at US$153,785, so the actual came in at 103 percent of contract; the 25 percent saving is against the appraisal estimate, which was set before the 1997 Asian crisis and may have been pessimistic. Actual construction took 45 months against 42 contracted, 7 percent longer. Annex 2 prints 88 percent beside the road works line although its own figures give 103 percent, so the table is not internally consistent. JICA separately financed 44 bridges on the road, about six of whose approaches still showed unacceptable settlement. The World Bank wrote and rated the report, and it describes a delivery finished more than 20 years ago.
๐ฒ๐ณMongoliasince 2017delivered and closedbears on Large project delivery
Mongolia's Ministry of Energy and its distribution companies rehabilitated the Baganuur and Erdenet-Bulgan distribution networks and built a 10 MWp solar plant in the Western Energy System under the Second Energy Sector Project, spending US$53.35 million against a US$55.40 million appraisal estimate but closing in August 2024, 25 months after the original September 2022 closing date.
The energy ministry split a lender-financed programme into distribution upgrades run by regional companies and one turnkey solar plant, kept costs inside the envelope and took two extensions to finish.
Preconditions
Regional distribution companies able to act as implementing units
Grant co-finance for the renewable component
A lender willing to extend rather than cancel
Limits: The cost and date figures are in section 2e of the IEG review (pages 2 to 3), read from the plain-text version; the full ICR was not opened. IEG attributes 18 of the 25 months to subprojects added in the western system at the 2022 restructuring and seven months to inefficiencies, with COVID-19 equipment delays, government changes after elections and procurement problems named as causes, so roughly seven months is slippage on the original scope. The planned US$1.0 million borrower contribution was cancelled, so the project was wholly grant financed. The project is mid-sized, about US$55 million, so it shows delivery on a modest programme rather than on Mongolia's flagship builds such as the new Ulaanbaatar airport, for which no completion figures were found. The solar plant was reported operational in 2022, ahead of the formal closing, so the schedule figure mostly concerns the distribution works. IEG rated the outcome Satisfactory.
๐ณ๐ฟNew Zealandsince 2016delivered and closedbears on Large project delivery
City Rail Link Limited, owned by the Crown and Auckland Council, achieved practical completion of Auckland's 3.45 km underground City Rail Link on 30 June 2026 against a target delivery cost of NZ$5,493 million, up from a NZ$3.4 billion estimate made in 2014, and the line opened to the public on 13 September 2026, after its company had planned completion for late 2024.
The Crown and Auckland Council set up a jointly owned delivery company, let the main stations and tunnels to an alliance, and re-baselined the budget twice as costs rose rather than cutting scope.
Preconditions
A sponsor agreement splitting cost 50:50 between national and local government
An alliance contract that shares cost risk with the builders
Political tolerance for re-baselining a flagship project
Limits: The NZ$5,493 million and the 30 June 2026 practical completion are in the 2025/26 annual report, a Word document; the 13 September 2026 opening is on the company's home page (https://www.cityraillink.co.nz/). The NZ$3.4 billion 2014 estimate and the 'late 2024' completion plan come from the company's cost page as archived on 26 January 2023 (https://web.archive.org/web/20230126164501/https://www.cityraillink.co.nz/city-rail-link-cost-timeline), which also gives the April 2019 revision to NZ$4.419 billion; the live page no longer carries them. The 62 percent increase over the 2014 estimate is our calculation and mixes price bases: the company attributes the 2019 rise to construction inflation, a larger contingency and a scope change for nine-car trains. Total spend to 30 June 2026 was NZ$5,283 million against approved funding of NZ$5,644 million, and the company reports itself within budget, which is true only against the 2019 and later baselines. The company is reporting on itself; no Auditor-General review of the outturn was opened. 'Started' is the year main works began, which the cited report does not state.
๐น๐ฏTajikistansince 2007still operatingbears on Large project delivery
The Government of Tajikistan resumed construction of the Soviet-era Rogun hydropower plant in 2007 from the state budget, signed the main dam contract in 2016 and commissioned two temporary units (400 MW) in 2018 and 2019. It invested about US$3.3 billion in 2007 to 2022, and in November 2024 the World Bank put the cost to complete the plant at US$6.29 billion, with construction completion moved to 2033 and full reservoir impoundment not before 2039, 63 years after construction first began in 1976.
The Tajik government restarted a stalled Soviet megaproject from its own budget and a Eurobond, brought two temporary turbines online early to earn revenue during construction, then used World Bank-backed safety and environmental studies to draw in a donor financing package for the rest.
Preconditions
A site with Soviet-era works, surveys and tunnels already in place
A state willing to divert a large share of GDP to one project for over a decade
Riparian water agreements and an independent dam-safety review credible enough for international lenders
Limits: No cost overrun can be computed: the appraisal document gives no original budget or baseline completion date to set the US$6.29 billion against, so the gap's unit (percent overrun) is not met. The US$6.29 billion is a forward estimate of remaining cost with contingencies, not an outturn, and adding it to the US$3.3 billion already spent is our arithmetic, not the Bank's. The schedule slip is our reading of milestone dates: construction began in 1976, the Soviet works stopped in 1992, and completion is now 2033 with impoundment in 2039. The Bank itself expects 'some further delays' to the March 2024 milestones. It attributes slow progress to financing shortfalls. The source is the lender's appraisal for its own grant, written to justify financing. The project is now financed by ten development partners, so delivery after 2024 is shared rather than purely Tajik. Spending close to 39 percent of GDP and a US$500 million Eurobond left the country at high risk of debt distress. 'Operating' means two temporary units run, not that the 3,780 MW plant is built.
๐ฉ๐ฐDenmarksince 2007still operatingbears on Large project delivery
Denmark set up Kvalitetsfonden in 2007 to fund a new hospital structure. Through it the regions are building or rebuilding 16 hospitals with a combined investment frame of about DKK 57 billion (2023 prices). Rigsrevisionen reported in November 2023 that the programme's largest overrun, Nyt Hospital Nordsjรฆlland, had its budget raised by DKK 2.3 billion, or 44 percent, and was expected six years later than planned.
Parliament created a state fund that gave each region a fixed, inflation-indexed grant frame per hospital and made the regions carry any overrun. The health ministry supervised changes against the grant conditions, and Rigsrevisionen audited the projects repeatedly while they were being built.
Preconditions
A 2007 structural reform that created five regions large enough to act as client for a multi-billion-krone building
A national decision to concentrate acute care in fewer, larger hospitals, which fixed the scope in advance
An audit office and a parliamentary audit committee that report on projects while they are still under construction
Where else it has been used: Norway's hospital trusts and the Netherlands' hospital rebuild programmes used fixed regional frames in a similar way. In the Danish case the largest single-site projects still ran over.
Limits: The headline 44 percent belongs to one project, the worst in the programme, so it is not the programme's cost performance. The report gives no programme-wide overrun in its text. Figure 2 of the same report lists original and latest budgets in 2009 prices for 15 of the 16 projects, leaving out Nyt Hospital Bispebjerg, whose budget had not been published. Summed by hand, the 15 come to about DKK 43.9 billion against frames of DKK 39.2 billion, roughly 12 percent over. That is our calculation from chart labels, not a published figure. Several projects held their frame partly by cutting scope through savings catalogues, which Rigsrevisionen criticised in Beretning 11/2019 and its follow-ups. A frame that holds after cuts is not a delivery on time and on specification. 2007 is the year the fund was set up. Grants were confirmed project by project afterwards (2013 for Nordsjรฆlland), so construction started years after 2007. The programme is unfinished: Rigsrevisionen's January 2025 follow-up still counted seven projects under construction, six of them under heightened ministry supervision, so final outturns are not known. The figure is a budget forecast in 2023 prices and is reported as a region's estimate, not an audited final account.
๐จ๐ฟCzechiasince 2013still operatingbears on Large project delivery
Czechia's Road and Motorway Directorate (ลSD) rebuilt 160.8 km of the D1 motorway between Miroลกovice and Kรฝvalka in 20 sections from April 2013, reopening the last sections in October 2021, three years later than the 2012 schedule, with costs projected at CZK 29.844 billion including the ล mejkalka bridge against CZK 28.208 billion planned.
The Road and Motorway Directorate rebuilt the country's main Prague to Brno motorway section by section with traffic still running. The Ministry of Transport approved 20 separate section projects instead of one programme budget, and EU cohesion funds paid CZK 14.0 billion of the cost.
Preconditions
EU Operational Programme Transport co-financing for two budget periods, which paid half the cost
A road agency able to tender and supervise many concurrent section contracts, and a ministry that never set a binding overall schedule or budget
A permit appeal process that averaged 444 days per contested section, which set the pace more than construction did
Limits: The cost figure looks good only because of the baseline it is compared with. The 2012 plan of CZK 28.208 billion was never approved: the audit office says the Ministry of Transport rejected the summary investment plan and approved 20 separate section projects, with no binding overall budget or schedule. The roughly 6 percent overrun is our computation from two published totals, in nominal crowns, with no inflation adjustment. The final cost is unknown. The audit office says the figure is not final, several sections had not been financially settled, and the ล mejkalka bridge cost is a forecast for work planned from 2022 to 2026. Against an earlier baseline the project looks worse. The audit office's 2017 release (audit 16/06) gave the expected total as CZK 20.5 billion and found unit costs up 73 percent, from CZK 74 to 128 million per km, between the 2013 sections and those begun in 2015 and 2016. The three-year slip is measured to the August 2021 schedule, which ลSD revised 73 times. In November 2022 eight sections still lacked final approval, and nine bridges and three overpasses were in poor or very poor condition. The audit says the modernisation is not complete, so a later finish date would lengthen the slip. This is a rebuild of an existing motorway kept open to traffic, not a new build. The audit looked closely at only five sections, worth CZK 6.792 billion, and a cyber attack on ลSD's systems limited what auditors could see. The figures are ลSD's own, reviewed by the auditor, not independently measured.
Papua New Guinea's Independent Public Business Corporation, using ADB and OPEC Fund loans, built the first phase of the Lae tidal basin, a new berth and container terminal at the port of Lae, between April 2012 and December 2014. It was about four years later than planned and cost $310.4 million, against a $154 million project at the 2007 approval.
The government had its state enterprise holding company, IPBC, borrow concessionally from ADB and the OPEC Fund. IPBC procured a single design-and-build contract for reclamation, berth and terminal, and the Morobe Provincial Administration ran the livelihood programme for displaced settlers and the Labu community.
Preconditions
Multilateral lenders willing to add $89 million of financing in 2011 when geotechnical findings broke the original design
A resettlement plan acceptable to settler households on government land and to the Labu villages
A state port company with tariff revenue to service the debt, which the validation doubted
Limits: The headline 7 percent overrun is against a 2011 revised estimate that had already absorbed $137.4 million of design changes found by later geotechnical work. Against the $154 million package approved in 2007 the increase is about 100 percent. That is our computation and is not like for like, because the scope grew. The validation reports civil works costs up 116 percent and involuntary resettlement costs up 3,590 percent against appraisal. Figures are nominal US dollars. The evaluator is the lender's own evaluation department, validating a completion report the lender wrote. It rated the project less than successful overall, less than effective, less than efficient and less than likely sustainable. The validation attributes the four-year delay to the combination of post-approval design changes and resettlement: 482 settler households on government marshland and a landing site used by three Labu villages, for which a supplementary resettlement plan acceptable to the affected communities had to be prepared. The contractors were foreign: a Korean design-build consultant and China Harbour Engineering. The national institutions' role was procurement, finance and resettlement. A post-completion audit found siltation back to 11 m below chart datum, from the dredged 14 m, at the basin's western entrance. The source says it appeared to have stabilised and does not report lost berth capacity. We have not treated it as an eroded delivery, but a reviewer could. The validation finds handling capacity of 2.0 million revenue tons a year, above the 1.4 million target, so the cost and delay bought a working asset. One project says nothing about PNG's record on other major works.
๐ฌ๐ทGreecesince 2008delivered and closedbears on Large project delivery
Greece procured 744 km of motorway as three PPP concessions that commenced in 2008; after works were suspended in 2011 and the contracts were renegotiated in 2013 and 2015, the European Court of Auditors found the three projects delayed by four years on average, their combined cost per km up 36 percent, from 9.1 to 12.4 million euro, and the planned road length cut to 486 km.
The Greek state bundled several hundred kilometres of motorway into three design-build-finance-operate concessions, transferred traffic risk to the concessionaires, and paid toll-backed private finance plus EU co-financing; when traffic collapsed the state renegotiated, deferred sections and paid about 1.2 billion euro more rather than stop the works.
Preconditions
Access to EU cohesion co-financing large enough to fill a renegotiated financing gap
A parliament willing to ratify concession contracts in the absence of a general concession law
Contracts signed before archaeology, permits and land expropriation were settled, which is what turned the crisis into public liability
Where else it has been used: Spain's A-1 and C-25 concessions, audited in the same report, were renegotiated soon after signature at a public cost of about 300 million euro, a smaller version of the same pattern.
Limits: This is a record of poor cost and schedule performance, not of good delivery, and it is filed because the gap asks for exactly that figure. The 36 percent is a cost per km, and it mixes two movements: the absolute cost of the Central motorway fell (2,375 to 1,594 million euro) while its length was cut by 55 percent, and Olympia's length was cut by 45 percent, so per-km cost rises partly because fixed and financial costs were spread over fewer kilometres. It is not a like-for-like overrun on a fixed scope. Table 3 of the report shows the combined total project cost of the three concessions falling from 6,743 million euro (2,375 + 2,825 + 1,543) to 6,004 million euro (1,594 + 2,619 + 1,791) as length fell from 744 to 486 km, so the 36 percent per-km increase coincides with an absolute cost about 11 percent lower on a 35 percent smaller scope. The ECA computed the figures from information provided by the Greek authorities and audited only three of the five second-wave concessions (Aegean and Nea Odos are not covered). The completion dates in Table 2 are labelled estimated/actual as of the audit, so the four-year average includes forecasts for the Central and Olympia sections due in August 2017. The ECA attributes much of the cost to the 2009 financial crisis, under which traffic fell around 50 percent below the lenders' worst case, and to the public partner's poor preparation (archaeology, environmental permits, expropriation): the record cannot separate those causes. The deferred northern and southern sections of E-65 and the deferred Olympia stretches were taken out of these contracts; whatever was built later under other contracts is not in this number. The report is from 2018. The ECA also records that Greece's PPP route procured 744 km in three tenders where the earlier 82 km of Olympia took up to 20 years and 31 procurements, which is the strongest point in the programme's favour.
๐ฆ๐ฟAzerbaijansince 2010delivered and closedbears on Large project delivery
Azerbaijan's road administration upgraded 94 km of the M4 Baku-Shamakhi highway to a four-lane road and built 48 km of new M3 motorway under the World Bank-financed Third Highway Project, at US$2.47 million per km against US$2.96 million estimated at appraisal, with the loan closing in March 2021 instead of March 2015.
The road administration borrowed from the World Bank for a defined corridor, accepted the lender's procurement and engineering review, held to the original alignment instead of the ministry's costlier viaduct proposal, and reused savings by adding a motorway section through restructuring.
Preconditions
Access to multilateral lending with appraisal estimates that create a published baseline
A road agency able to procure and supervise several works contracts at once
Works contracts in local currency, which turned a devaluation into a dollar saving
Limits: The completion report is the lender's own evaluation of its loan, drafted with the borrower's data. The cost saving is not mainly efficiency: the report says the lower dollar outcome is mainly attributable to the manat losing more than half its value in 2015 to 2017 while works contracts were priced in local currency, so in local terms the project may have overrun. The 72-month slip is computed here from the two closing dates; the scope also changed twice (the M3 section added, part of the M4 dropped in a landslide zone), and the report concedes the final extension may have been avoidable, so not all of it is construction delay. The efficiency rating is only Modest. It is one corridor of 142 km, with a lender supervising procurement, so it is weak evidence about Azerbaijan's road programme outside donor projects.
๐ง๐ฌBulgariasince 2008dismantledbears on Large project delivery
Bulgaria's state utility NEK ordered long-lead reactor equipment for the Belene nuclear plant from Atomstroyexport in 2008. The project was cancelled before any unit was completed, and on 6 December 2016 the Ministry of Energy transferred 601,617,133 euros to NEK so it could pay the damages awarded against it in ICC arbitration case 18086/GZ/MHM.
601,617,133 EUR, state financial assistance transferred to nek to pay the belene arbitration award to atomstroyexport, 2016-12-06. 1,400,000,000 BGN, costs incurred on the belene construction to date, as stated by the minister of energy, 2016. Ministry of Energy of the Republic of Bulgaria, press release of 6 December 2016, retrieved 2026-10-02.
The move, and what it needed
NEK, the state utility, ordered long-lead reactor equipment from a single foreign vendor in 2008 before financing or a final contract price was settled. The government cancelled the project in 2012, the Ministry of Energy advanced NEK repayable assistance in 2016 to pay the vendor's arbitration claim, and the project was revived in 2018.
Preconditions
A state utility allowed to commit to equipment orders before the project's financing was closed
Political control of the decision that changed with each government
A single-vendor contract enforceable in international arbitration
Limits: Belene never produced electricity, so this record documents a delivery that failed before it operated, not a capability that eroded. The headline number is a termination cost, the arbitration damages for equipment ordered in 2008, and not a cost overrun against a budget: the record carries no official baseline budget or schedule for the 2008 phase. The 1.4 billion leva of costs incurred comes from a minister speaking in a television interview. The ministry republished the transcript without a printed date; the 2016 year is inferred from the special assistance law the interview discusses. It is not an audited figure. The project was decided in April 2005, cancelled in March 2012, and revived by a National Assembly decision of 7 June 2018 and a Council of Ministers decision of 29 June 2018. Later decisions on the site and the equipment are not covered by these sources. The 601.6 million euros was advanced to NEK as repayable assistance, not a grant, so the final fiscal loss is not shown here. The later disposal or reuse of the equipment the payment bought is not covered.
๐ง๐พBelarussince 2013still operatingbears on Large project delivery
Belarus started building its first nuclear power plant at Ostrovets in November 2013, a two-unit WWER-1200 plant totalling 2,400 MW. Its regulator's 2019 report scheduled unit 1 for 2019 and unit 2 for 2020. Unit 1 was commissioned in 2021 and unit 2 in October 2023, three years after the 2019 schedule.
The Belarusian state chose a single foreign state vendor to build a reference design turnkey, and built up a domestic regulator and operator in parallel under a sequence of IAEA review missions, accepting slippage on the vendor's schedule.
Preconditions
A strategic partner state willing to supply design, construction and credit together
A centralised state able to hold a decade-long build through political and diplomatic opposition
A reference plant of the same design already being built abroad (Leningrad NPP-2)
Limits: This is a schedule figure only. The record carries no official budget or final cost for the plant, so cost performance is unknown. The slip is measured against a schedule that the 2019 report itself calls 'updated'. This record uses the 2019 schedule because it is the one stated in an official source. The schedule is given in years, not months, so the three-year slip for unit 2 is approximate. The 2019 schedule speaks of commissioning while the 2023 figure is the start of commercial operation, so the two milestones may not be identical. Both figures come from the Belarusian regulator reporting on its own country's flagship project. The plant was built turnkey on a Russian design, financed from the republican budget and a Russian state export credit, as the 2019 national report states. The delivery is therefore only partly attributable to Belarusian institutions: they decided, licensed, supervised, financed and now operate it. The source does not show capacity factors or outages after commissioning.
๐ฐ๐ฌKyrgyzstansince 2002delivered and closedbears on Large project delivery
The Kyrgyz Ministry of Transport and Communications completed the third and final phase of the Bishkek-Osh road rehabilitation in July 2007, bringing the rehabilitated length of the only internal surface connection between Bishkek and Osh to 463 km, with a 31.7 percent cost overrun on base cost items and civil works finished 33 months after the October 2004 appraisal schedule.
The Kyrgyz government, through the Ministry of Transport and Communications and a project implementation unit staffed by local consultants experienced on the earlier phases, rehabilitated its only north-south trunk road in phases over a decade, financing each phase with an ADB concessional loan and a counterpart share.
Preconditions
A concessional lender willing to finance the same corridor three times in succession
An implementation unit staffed by local consultants experienced on the earlier phases
Counterpart and maintenance funding from a small budget, which arrived late and only after the works ended
Limits: The record covers phase three only: 120 km of the Bishkek-Osh road and 125 km of secondary roads, not the two earlier phases. The 31.7 percent is on base cost items and is driven by the secondary roads and an added bridge ($7.6 million of the $13.6 million); main-road civil works ran 11.6 percent over per km, and total project cost rose from $50.0 million to $57.1 million, about 14 percent, because contingencies and currency gains absorbed part of the overrun. The 33 months is computed from two dates in the text, and 22 of the delay months came before construction, in loan effectiveness and bidding against a 7-month plan; one civil works contract was terminated and the supervision consultant was sanctioned. The road was 80 percent financed by an ADB concessional loan, and the validation is ADB's evaluation office reviewing ADB's own completion report; it rated the project 'less effective' pending a later performance evaluation that this record does not carry. Traffic grew about 4 percent a year against a forecast of 6 to 7, accidents rose in number and severity after the works, and the road maintenance funding covenants were not met until 2007, so nothing here shows the condition of the road today.
Estonia has run legally binding internet voting since 2005, and at the 2023 Riigikogu election 312,182 votes, 51.1 percent of all votes cast, came in online, the first national election anywhere in which more than half the vote was digital.
The state let any voter cast and re-cast a ballot online during the advance period, with the last vote counting and a paper vote overriding, which turned the coercion problem into a game the voter can always exit.
Preconditions
A universal cryptographic identity already in every pocket
An electoral authority with the engineering staff to own the system rather than rent it
Small scale, which keeps the attack surface and the audit tractable
Where else it has been used: Twenty years on, no other country runs internet voting at national scale: Switzerland trials and withdraws, Norway tried and stopped. The Estonian lesson is that the identity system came first, and copies that start with the voting app fail.
Limits: Turnout by channel measures adoption, not democratic quality: an online share says nothing about coercion outside the polling station, verifiability, or trust in the count, and Estonian i-voting has standing academic critics on all three. Adoption also rides on the national identity infrastructure recorded elsewhere in this dataset, so as evidence for civic participation it shows a lowered cost of voting rather than a raised desire to participate; overall turnout moved far less than the channel share.
๐ฆ๐บAustraliasince 1924still operatingbears on Civic participation
Australia has required citizens to vote since 1924 and turnout at the 2025 federal election was 90.7 percent of enrolled voters, a level no large voluntary-voting democracy reaches.
Parliament made voting a legal duty with a token fine, paired it with automatic-style enrolment, Saturday voting and easy postal and early options, so the state removed every excuse at the same time as it imposed the duty.
Preconditions
Cross-party agreement to keep the duty, since either side could repeal it
An electoral commission trusted to run painless voting
A fine large enough to state the norm and small enough not to embitter
Where else it has been used: Around twenty countries mandate voting and fewer enforce it; Belgium and Luxembourg are the closest working siblings. Where compulsion was dropped, as in Chile in 2012, turnout fell by tens of points within two elections.
Limits: Turnout under compulsion measures compliance, not civic desire, which is exactly the objection to filing it under civic participation; the record is here because the indicator's construct is participation behaviour, and ninety percent of a country voting for a century is behaviour, whatever its cause. The denominator is enrolled voters, and enrolment itself is near-complete only because it is also compulsory. The fine is small and enforcement light, so the mechanism is closer to a maintained norm than to coercion.
๐ท๐ผRwandasince 2007still operatingbears on Volunteering
Rwanda requires a morning of community work on the last Saturday of each month, and the government's own assessment reports average participation of 91.3 percent of adults countrywide in fiscal year 2015-2016, up from 87.8 percent two years earlier.
The state fixed a single national morning each month, closed roads and shops for it, assigned works through the lowest administrative level, and had officials up to the president participate visibly, converting an old rural practice into a synchronised national institution.
Preconditions
An administrative structure that reaches every village
A governing party able to enforce a norm without visible coercion
A rebuilding narrative in which refusing communal work reads as defection
Where else it has been used: The practice descends from communal-labour traditions found across East Africa; Rwanda's addition is the synchronisation and the enforcement. Copies without the administrative reach, or the pressure, produce a cleanup day rather than an institution.
Limits: The number is the state assessing its own compulsory programme in a country where declining to participate carries social and administrative cost, so it cannot be read as volunteering in the survey sense the indicator intends, and no independent measurement exists. Participation is legally required, which is the opposite of the construct's voluntariness. The record is filed anyway because the indicator is a gap and the case is the strongest documented instance of state-organised mass civic labour, provided the compulsion is read alongside the number.
๐ง๐ทBrazilsince 1990still operatingbears on Civic participation
Since Law 8.142 established them in 1990, Brazil's SUS has required permanent, deliberative health councils at the national, state and municipal levels, with users holding 50 percent of seats and health workers and government or providers holding 25 percent each.
Brazil made participation part of the health system's operating structure: users, workers, government and providers share a permanent council whose decisions feed into policy formulation and control of execution.
Preconditions
A national health system with responsibilities distributed across government levels
A legal requirement that makes councils permanent and deliberative
Representative organisations able to select users and workers for the councils
Limits: A mandated seat structure creates a route for participation but does not show whether councils meet regularly, deliberate independently, change health priorities or attract people beyond organised representatives. The 50/25/25 composition is a legal and administrative design, not a measure of actual attendance or influence, and the ministry describes the system without an independent evaluation of outcomes.
๐ฉ๐ชGermanysince 1950still operatingbears on Volunteering
Germany's federal civil-protection agency trains and equips a volunteer force that reached about 88,000 people across 669 local units in May 2025, supported by about 2,200 paid staff.
The federal government supplies equipment, doctrine and a small professional backbone while local volunteer units recruit, train and retain the people who operate it, turning civic participation into standing disaster-response capacity.
Preconditions
Local associations able to recruit and hold volunteers over many years
Federal finance for specialised equipment, training and paid coordination
Employers willing to release volunteers when an emergency triggers deployment
Where else it has been used: Volunteer fire and civil-protection systems exist across Europe, but the German combination of a federal technical agency and locally embedded unpaid units is unusually institutionalised. A copy has to fund readiness between disasters, not just recruit after one.
Limits: Membership in one emergency-service organisation is not the share of adults who volunteered in the last month, which is what the indicator asks for, and it cannot be compared directly with countries that organise civil protection through paid services. The agency reports its own membership, the count includes more than 17,000 youth members, and it does not show how often each member trains or deploys.
๐ฏ๐ตJapansince 1965operating below its peakbears on Volunteering
Japan's Fire and Disaster Management Agency reported 732,223 volunteer fire-corps members on 1 April 2025, a fall of 14,458 in one year in the local disaster-response corps that it describes as a core of community resilience.
Japan embeds trained local volunteers in municipal fire organisations, gives them equipment and a formal role in prevention and disaster response, and is now adding function-specific and student units as the general membership base shrinks. The erosion shows that a long-standing civic institution needs an active recruitment and role-design strategy to remain deployable.
Preconditions
Municipal fire organisations able to train, equip and insure residents who serve alongside professional crews
A social norm and compensation structure that make recurring emergency service compatible with work and family life
Flexible roles, youth and women recruitment, and a data system that reveals membership and participation gaps
Where else it has been used: Volunteer emergency corps travel when local legitimacy is matched with practical support and flexible participation; simply asking communities to volunteer does not replace equipment, training or a reason to stay.
Limits: Membership is a capacity proxy, not the number of volunteers who turn out, hours trained, response quality or lives and property protected. The agency reports a one-year decline but this record does not identify its causes or establish a long-run causal trend; local corps differ in duties, age structure and recruitment, and the national total can hide uneven resilience.
๐บ๐ธUnited Statessince 1977operating below its peakbears on Political polarisation
The Congressional Research Service records 20 federal funding-gap shutdowns with at least one full day between FY1977 and FY2019, including the 35-day partial shutdown from 21 December 2018 to 25 January 2019, the longest in the period it surveys.
The US appropriations process allows a lapse in budget authority to stop or furlough parts of the federal government, turning unresolved political conflict into an interruption of administration. Continuing resolutions and contingency plans soften the shock but do not remove the incentive or the coordination cost of brinkmanship.
Preconditions
Separated institutions in which the executive and legislature can block or condition annual appropriations
A legal shutdown rule that makes funding lapses administratively real rather than automatically extending the prior budget
Political parties willing to use public-service disruption as leverage, plus contingency systems that limit but cannot eliminate the damage
Where else it has been used: Funding-gap reversals travel as a warning: constitutional checks need a credible continuity mechanism, or polarised bargaining can turn a disagreement into a measurable loss of state capacity.
Limits: Shutdown duration and frequency are institutional consequences that can reflect constitutional rules, divided government and bargaining strategy, not a direct measure of public attitudes or latent polarisation. The CRS series ends in FY2019 for this historical resource, and a shutdown's operational cost varies with which agencies lose budget authority.
๐ต๐นPortugalsince 2017delivered and closedbears on Civic participation
Portugal's first national participatory-budget cycle drew almost 80,000 votes and committed โฌ3.18 million to winning projects, which were written into the 2018 state budget.
The central government opened a national proposal and voting platform, reserved a defined slice of the state budget, and promised to place winning projects in departmental budgets so participation could produce a visible delivery rather than only consultation.
Preconditions
A public budget line that can be set aside before proposals are voted on
A technical process for screening proposals without removing the final choice from citizens
Departments and regions willing to implement projects they did not originate
Where else it has been used: Participatory budgeting travels when a vote is linked to a protected implementation route; a platform without a budget, screening and delivery owner produces participation theatre.
Limits: A single cycle's votes do not establish sustained civic participation, representativeness, deliberative quality or whether every winning project was delivered on time. The government announcement reports the programme's own participation and budget totals, and the national cycle was later redesigned rather than becoming a stable annual measure of citizen voice.