## _wp1614

## Source details

**Canonical URL:** [_wp1614](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1614.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1614.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1614.pdf.json)

---

### Key findings and statistics
- Dataset coverage:
  - 80 countries—34 advanced economies (AEs) and 46 emerging market economies (EMEs)—over the period 1990–2014.
  - Total contingent liability (CL) episodes identified: 230.
  - Episodes with identified fiscal cost: 174.
- Fiscal cost distribution and magnitudes:
  - Average fiscal cost: about 6 percent of GDP.
  - Median fiscal cost: about 2 percent of GDP.
  - Very large costs (fiscal costs above 20 percent of GDP) are rare but present.
  - Historical examples cited: fiscal costs up to 50 percent of GDP during the Asian and Latin American crises; fiscal costs related to natural disasters historically as high as 10 percent of GDP.
- Frequency and timing:
  - A macro-relevant contingent liability realization occurs on average every 12 years per country.
  - Contingent liability realizations tend to cluster—“when it rains it pours”—often coinciding with major crises.
  - Emerging markets experienced many costly realizations during the Asian Crisis; advanced economies experienced many during the Global Financial Crisis.
- Types and sources:
  - Seven CL categories used: Financial Sector, SOEs, Subnational Government, Natural Disasters, Private non-Financial Sector, Legal, and PPPs.
  - The financial sector accounts for the largest fraction of episodes with the highest costs; subnational government bailouts, support to SOEs, and legal liabilities can also impose substantial costs.

### Definitions, scope, and measurement
- Definition of contingent liabilities:
  - CLs are obligations that do not arise unless particular discrete events occur in the future (Public Sector Debt Statistics Guide, IMF, 2011).
  - Contractual split: explicit CLs (set by law or contract) versus implicit CLs (moral obligations or public expectations).
- Examples of explicit CLs:
  - Guarantees for non-sovereign borrowing, obligations to subnational governments, PPP contract guarantees, guarantees for mortgages, student and small business loans, state insurance schemes (e.g., deposits, minimum returns for private pension funds, agricultural insurance), and export trade guarantees.
- Examples of implicit CLs:
  - Default of subnational governments or public/private entities on nonguaranteed debt, bank failure, investment failure of nonguaranteed pension funds, environmental damage, disaster relief, military financing.
- Accounting and disclosure context:
  - Under accrual accounting, CLs are not recognized unless probability >50 percent and payments can be reliably estimated (then recognized as a provision).
  - Disclosure best practices include description, estimate of financial effect, uncertainties, possible reimbursement, and stocks/flows information.
  - The Fiscal Transparency Code (IMF, 2014) recommends identifying, quantifying, and disclosing guarantees, PPP obligations, explicit government support to financial sector, and direct/indirect support to public corporations at least annually; and disclosing main implicit CL risks in a summary report.

### Dataset construction and methodology
- Three-pillar data collection strategy:
  1. Build a baseline database combining previously available data on CL realizations and cross-check using IMF Staff Reports (SRs).
  2. Use stock-flow analysis and debt forecast error decomposition to guide identification of country-years with potential CL episodes.
  3. Conduct keyword searches of all remaining SRs.
- Primary and complementary sources:
  - Main sources: IMF Staff Reports (Article IV surveillance reports and program reviews).
  - Combined existing datasets: Laeven and Valencia (2008, 2012) and Eurostat (2015) for financial sector fiscal costs; Cordes and others (2014) for subnational bailouts; Flanagan (2008) for large Eastern European legal liabilities; International Disasters Database (University of Leuven) for natural disasters (episodes with damages of at least one percent of GDP); Cebotari and others (2009) for a variety of episodes.
  - Crosschecking: Existing datasets cross-checked against SRs for consistency.
- Stock-flow and forecast-error framework (preserve notation and definitions as in source):
  - Stock-flow adjustment (SF_t) = discrepancy between annual change in gross public debt and the budget deficit; used to flag potential CL realizations.
  - Debt-to-GDP decomposition: variables in percent of GDP, Δd_t denotes annual change in gross public debt to GDP ratio, γ_t denotes nominal GDP growth rate, λ_t = (1+γ_t)/γ_t.
  - Forecast error decomposition: x~_t = x_t - E_{t-1}[x_t]; residual sf_t captures unexpected increases in debt from contingent liabilities and other factors.
- Episode coding:
  - For each CL episode where possible: start year, end year, implicit or explicit CL, type, fiscal response, fiscal cost, trigger, source, and short verbal description. Appendix D provides overview of collected data.
- Verification process:
  - Flag large positive SF and forecast errors (distribution heavily right skewed; truncated below 0 and above 50 percent of GDP).
  - Follow-up verification for right-tail observations using multiple sources.
  - Keyword searches in SRs and complementary sources with keywords such as “recapitalization,” “capital injection,” “restructuring,” “natural disaster,” “contingent,” “SOEs,” “PPP,” etc.

### Stylized facts and econometric insights
- Distributional properties:
  - Distribution of fiscal cost is highly skewed: very large fiscal costs are rare while average cost is pulled up by those large, infrequent events.
- Correlations and drivers:
  - CL realizations are highly correlated with each other and with major crises.
  - Logit regressions indicate CL materializations tend to follow periods of high growth (lagged growth positive) and coincide with low contemporaneous growth and banking crises.
- Institutional and macro links:
  - Countries with stronger institutions and lower growth volatility tend to suffer fewer contingent liability realizations.
  - Evidence points to the role of institutional quality (e.g., bureaucratic quality, corruption indicators referenced in figures) in reducing expected yearly costs.
- Event dynamics:
  - Debt increases by over 15 percent of GDP during a CL realization on average; it rises for roughly three years and then stabilizes at a higher level.
  - Overall fiscal balance falls by about 2 percentage points of GDP on average and remains below the pre-event level for an extended period.
  - GDP growth drops sharply for two years and then reverts to trend.
- Regression evidence (selected exact coefficients and statistics reproduced where provided):
  - Systemic Banking Crisis coefficients (Table 4): 4.439*** (column 1; standard error (0.410)); 5.553*** (column 2; standard error (0.379)). Observations: 2,000 ; 2,000. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
  - Correlations among CL types (Table 5, pairwise coefficients with p-values in parentheses; * indicates significance at 5 percent level) include:
    - Financial vs Private non-financial: 0.0032 (0.8843)
    - Financial vs SOE: 0.0760* (0.0006)
    - Financial vs Subnational: 0.0634* (0.0043)
    - Financial vs PPP: 0.1643* (0)
    - SOE vs Subnational: 0.1117* (0)
    - SOE vs Legal: 0.1796* (0)
    - SOE vs PPP: 0.1220* (0)
    - Subnational vs PPP: 0.2178* (0)
    - (Full table reproduced in source.)

### Summary table by type of contingent liabilities (preserve entries exactly)
- Type of Contingent Liabilities | Number of Episodes | Number of Episodes with Identified Fiscal Costs | Avg. Fiscal Costs (% GDP) | Maximum Fiscal Costs (% of GDP)
- Financial | 91 | 82 | 9.7 | 56.8
- Legal | 9 | 9 | 7.9 | 15.3
- Subnational Government | 13 | 9 | 3.7 | 12.0
- SOEs | 32 | 31 | 3.0 | 15.1
- Natural Disaster(s) | 65 | 22 | 91.6 | 6.0
- Private Non-Financial Sector | 76 | 1 | 7.4 | 4.5
- PPPs | 8 | 5 | 1.2 | 2.0
- Other | 5 | 3 | 1.4 | 2.5
- Total | 230 | 174 | 6.1 | 56.8

### Frequency, probabilities, and country-level experience
- Aggregation rule: sum all episodes that start in exactly the same year in the same country to obtain one country-year observation.
- Probability and cost statistics (Table 2 reproduced exactly):
  - Size (In percent of GDP) | Probability of CL Realization | Average CL Realization in percent of GDP | Number of Years until CL Realization on average = (1/probability)
  - >5 | 2.8% | 15.5 | 36
  - >1 | 5.6% | 9.2 | 18
  - Any | 8.7% | 6.1 | 12
- Interpretation:
  - Average country probability of a macro relevant CL realization in any given year = 8.7 percent → one CL realization every 12 years on average.
  - Average country would be expected to have experienced a CL realization twice in the twenty five-year sample period, with a fiscal cost of 6.1 percent of GDP per episode.
  - Probability of suffering a CL realization of at least 5 percent of GDP = 2.8 percent; conditional fiscal cost = 15.5 percent of GDP.
- Cross-country heterogeneity (Table 3 country examples, exact entries):
  - Argentina: Episodes over 1 percent of GDP = 4; Total number of episodes = 5; Average cost per episode (percent of GDP) = 7.9
  - Brazil: Episodes over 1 percent of GDP = 4; Total number of episodes = 6; Average cost per episode (percent of GDP) = 8.3
  - China: Episodes over 1 percent of GDP = 2; Total number of episodes = 10; Average cost per episode (percent of GDP) = 4.7
  - Hungary: Episodes over 1 percent of GDP = 4; Total number of episodes = 4; Average cost per episode (percent of GDP) = 1.5
  - Indonesia: Episodes over 1 percent of GDP = 3; Total number of episodes = 6; Average cost per episode (percent of GDP) = 15.8
  - Ukraine: Episodes over 1 percent of GDP = 4; Total number of episodes = 4; Average cost per episode (percent of GDP) = 2.9

### Institutions, transparency, and fiscal risk
- Institutional links and quantitative findings:
  - Corruption: Figure 9 shows countries with lower corruption scores have a lower expected cost of CL realizations.
  - Bureaucratic quality: expected fiscal cost is over 30 percent higher in countries with below median bureaucratic quality than in those with an above median score.
  - Volatility channel: standard deviation of growth in countries with below median bureaucratic quality is nearly double that in countries with an above median score; expected cost of CL realizations is nearly 50 percent higher in countries with above median volatility of growth than in those with below median volatility.
  - Results in Figures 9–11 are virtually unchanged when excluding natural disasters.

### Policy implications and recommended practices
- Strengthen fiscal risk disclosure:
  - Identify, quantify, and disclose government guarantees, PPP obligations, explicit financial sector support, and intergovernmental support at least annually.
  - Disclose main implicit contingent liabilities and their estimated magnitudes and likelihoods in fiscal risk summaries.
- Institutional improvements:
  - Improve institutional quality (reduce corruption, strengthen bureaucratic quality) to lower probability and cost of CL realizations.
  - Reduce macroeconomic volatility (growth volatility) to decrease exposure to CL realizations.
- Fiscal and regulatory measures to minimize probability and impact:
  - Limit direct exposure of public sector entities.
  - Require beneficiaries of guarantees to post collateral.
  - Require banks to hold sufficient capital.
  - Adopt regulatory, risk-sharing, and stronger governance measures to mitigate CL risk.
- Use of the dataset:
  - Policymakers and analysts should use comprehensive, cross-checked datasets (including SRs) and stock-flow/forecast error indicators to detect and prepare for CL episodes.

### Representative dataset episodes and modalities (selected exact entries and values)
- Belarus 1999–1999 Financial Sector: Recapitalization of two major state owned banks (Belarusbank and Belagroprombank) in 1999, equivalent to 2.4% of GDP.
- Belgium 2005–2005 SOEs: Eurostat classifies the assumption by government Railway Infrastructure Fund (FIF) in 2005 of EUR 7 400 million (2.5% of GDP) of the debt of SNCB as expenditure (capital transfer).
- Belgium 2008–2011 Financial Sector: 13.5% of GDP SR 2012, SR 2013, World Bank (2012) — Recapitalization of SOBs between 2008 and 2011.
- Bosnia and Herzegovina 2004–2004 Financial Sector: 6.4% of GDP Eurostat (2015) — Bailout of the largest banks in the country.
- Bosnia and Herzegovina 2007–2009 Legal: 15.3% of GDP Flanagan (2008), SR 2008, SR 2007, SR 2006 — Increase in debt by 11.5 percent of GDP in 2007 due to recognition of frozen foreign currency deposits and war claims. Total NPV of settlement 15.3% of GDP.
- China 1998–1998 Financial Sector: 18.0% of GDP Laeven and Valencia (2012) — Laeven and Valencia put total fiscal cost in 1998 at 18 percent of GDP.
- Cyprus 2012–2012 Financial Sector: 9.7% of GDP Eurostat (2015), 2014 SIP — large recapitalizations due to exposure to Greece and local property developers.
- Dominican Republic 2003–2004 Financial Sector: 22.0% of GDP Laeven and Valencia (2012), SR 2003, SR 2005 — public assistance during banking crisis ultimately amounted to 22 percent of GDP.
- Ecuador 1998–2002 Financial Sector: 21.7% of GDP Laeven and Valencia (2012).
- Greece 2009–2013 Financial Sector: 23.1% of GDP Eurostat (2015) — large bank recapitalization in 2012-13.
- Ireland 2008–2011 Financial Sector: 38.9 percent of GDP. (Eurostat (2015))
- Indonesia 1997–2001 Financial Sector: 56.8 percent of GDP. (Laeven and Valencia (2012))
- Korea 1997–1998 Financial Sector: 31.2 percent of GDP. (Laeven and Valencia (2008), Laeven and Valencia (2012))
- Mexico 1994–1996 Financial Sector: 19.3 percent of GDP. (Laeven and Valencia (2008), Laeven and Valencia (2012))
- Selected natural disaster and PPP examples:
  - Thailand 2011–2014 Natural Disaster(s): 3.0 percent of GDP reconstruction program initiated as a response; EM-DAT estimates total damage at 11.7 percent of GDP.
  - Chile 2010–2013 Natural Disaster(s): government contribution 4.0% of GDP for reconstruction.
  - Colombia PPPs cumulative payments 2.0% of GDP by 2004.

### Expected cost metric and headline statistics (preserve exact figures)
- Mean CL realization = 6.1 percent of GDP.
- Median CL realization = 2.3 percent of GDP.
- Total CL realizations recorded = 230.
- Episodes with identified fiscal cost = 174.
- Mean fiscal cost per CL realization = 6.1 percent of GDP.
- Median fiscal cost per CL realization = 2.3 percent of GDP.
- Debt increases by over 15 percent of GDP on average during a CL realization.
- Overall fiscal balance falls by about 2 percentage points of GDP on average during a CL realization.
- Average annual probability of any macro relevant CL realization for a country = 8.7 percent → one event every 12 years on average.
- Probability of CL realization >5 percent of GDP = 2.8 percent; conditional average cost = 15.5 percent of GDP → one event every 36 years on average for >5 percent events.
- Financial sector CL realizations are the most costly on average: Avg. Fiscal Costs = 9.7 percent of GDP; Maximum Fiscal Costs = 56.8 percent of GDP (table entries preserved exactly).
- Over 80 percent of CL realizations in the dataset stem from implicit rather than explicit contingent liabilities.
- Expected cost: 0.53 percent of GDP each year.
- Aggregate example statistic cited: 6.1 percent of GDP.
- Long-run implication: roughly 10 percentage points rise in debt to GDP ratio over a twenty-year period.
- Database coverage: more than 200 episodes across 80 countries over the period 1990–2014.

### Data limitations and avenues for further research
- Coverage limitations:
  - Some episodes may not have been captured; dataset could be extended to pre-1990 periods and low-income countries.
  - Coverage of reported fiscal costs varies by country and period: in the 1990s coverage tends to be central government; in the 2000s coverage expanded to general government mainly in advanced economies.
- Measurement notes:
  - Gross fiscal costs are used (immediate outlays) due to limited data on net costs (e.g., asset recoveries).
  - For natural disasters, EM-DAT damage estimates are reported and noted as "not the fiscal costs."
- Research opportunities:
  - More analytical work on causes and consequences of CL realizations.
  - Comparisons across countries and time to inform prevention and management of CL shocks.
  - Use of stock-flow and forecast-error indicators combined with SR verification to improve early detection of CL episodes.

*Source: _wp1614 - References (IMF working paper content provided).*

### References .............................................................................................................

### References

### Key findings and statistics
- Dataset coverage:
  - 80 countries—34 advanced economies (AEs) and 46 emerging market economies (EMEs)—over the period 1990–2014.
  - Total contingent liability (CL) episodes identified: 230.
  - Episodes with identified fiscal cost: 174.
- Fiscal cost distribution and magnitudes:
  - Average fiscal cost: about 6 percent of GDP.
  - Median fiscal cost: about 2 percent of GDP.
  - Very large costs (fiscal costs above 20 percent of GDP) are rare but present.
  - Historical examples cited: fiscal costs up to 50 percent of GDP during the Asian and Latin American crises; fiscal costs related to natural disasters historically as high as 10 percent of GDP.
- Frequency and timing:
  - A macro-relevant contingent liability realization occurs on average every 12 years per country.
  - Contingent liability realizations tend to cluster—“when it rains it pours”—often coinciding with major crises.
  - Emerging markets experienced many costly realizations during the Asian Crisis; advanced economies experienced many during the Global Financial Crisis.
- Types and sources:
  - Seven CL categories used: Financial Sector, SOEs, Subnational Government, Natural Disasters, Private non-Financial Sector, Legal, and PPPs.
  - The financial sector accounts for the largest fraction of episodes with the highest costs; subnational government bailouts, support to SOEs, and legal liabilities can also impose substantial costs.

### Definitions, scope, and measurement
- Definition of contingent liabilities:
  - CLs are obligations that do not arise unless particular discrete events occur in the future (Public Sector Debt Statistics Guide, IMF, 2011).
  - Contractual split: explicit CLs (set by law or contract) versus implicit CLs (moral obligations or public expectations).
- Examples of explicit CLs:
  - Guarantees for non-sovereign borrowing, obligations to subnational governments, PPP contract guarantees, guarantees for mortgages, student and small business loans, state insurance schemes (e.g., deposits, minimum returns for private pension funds, agricultural insurance), and export trade guarantees.
- Examples of implicit CLs:
  - Default of subnational governments or public/private entities on nonguaranteed debt, bank failure, investment failure of nonguaranteed pension funds, environmental damage, disaster relief, military financing.
- Accounting and disclosure context:
  - Under accrual accounting, CLs are not recognized unless probability >50 percent and payments can be reliably estimated (then recognized as a provision).
  - Disclosure best practices include description, estimate of financial effect, uncertainties, possible reimbursement, and stocks/flows information.
  - The Fiscal Transparency Code (IMF, 2014) recommends identifying, quantifying, and disclosing guarantees, PPP obligations, explicit government support to financial sector, and direct/indirect support to public corporations at least annually; and disclosing main implicit CL risks in a summary report.

### Dataset construction and methodology
- Data collection strategy (three pillars):
  1. Build a baseline database combining previously available data on CL realizations and cross-check using IMF Staff Reports (SRs).
  2. Use stock-flow analysis and debt forecast error decomposition to guide identification of country-years with potential CL episodes.
  3. Conduct keyword searches of all remaining SRs.
- Primary and complementary sources:
  - Main sources: IMF Staff Reports (Article IV surveillance reports and program reviews).
  - Combined existing datasets: Laeven and Valencia (2008, 2012) and Eurostat (2015) for financial sector fiscal costs; Cordes and others (2014) for subnational bailouts; Flanagan (2008) for large Eastern European legal liabilities; International Disasters Database (University of Leuven) for natural disasters (episodes with damages of at least one percent of GDP); Cebotari and others (2009) for a variety of episodes.
  - Crosschecking: Existing datasets cross-checked against SRs for consistency.
- Stock-flow adjustments and forecast error decomposition:
  - Stock-flow adjustment defined as the discrepancy between the annual change in gross public debt and the budget deficit (indicator to guide CL episode search).
  - Acknowledged caveat: changes in debt unexplained by deficit could reflect CL realizations or other factors (e.g., exchange rate changes).

### Stylized facts and econometric insights
- Skewness and rarity:
  - Distribution of fiscal cost is highly skewed: very large fiscal costs are rare, while average cost is pulled up by those large, infrequent events.
- Correlations and drivers:
  - CL realizations are highly correlated with each other and with major crises.
  - Logit regressions indicate CL materializations tend to follow periods of high growth and coincide with low growth periods and banking crises.
- Institutional and macro links:
  - Countries with stronger institutions and lower growth volatility tend to suffer fewer contingent liability realizations.
  - Evidence points to the role of institutional quality (e.g., bureaucratic quality, corruption indicators referenced in figures) in reducing expected yearly costs.
- Coverage notes:
  - Coverage of reported fiscal costs varies by country and period: in the 1990s coverage tends to be central government; in the 2000s coverage expanded to general government mainly in advanced economies.
  - Gross fiscal costs are used (immediate outlays) due to limited data on net costs (e.g., asset recoveries).

### Tables, figures, and appendices (listed in source)
- Tables included (selection of titles and page references as listed):
  - Table 1. Average Fiscal Cost of Contingent Liability Realizations (p.13)
  - Table 2. Probability of Contingent Liability Realizations (p.16)
  - Table 3. Contingent Liability Realizations: Country Cases (p.17)
  - Table 4. Correlation between Contingent Liability Realizations and Crises (p.20)
  - Table 5. Correlation between Different Types of Contingent Liability Realizations (p.20)
  - Table 6. Triggers of Contingent Liability Realizations (p.22)
- Figures included (selection of titles and page references as listed):
  - Figure 1. Distribution of Positive Stock Flow Adjustment and Forecast Error (p.11)
  - Figure 2. Distribution of Contingent Liability Realizations 1990–2014 (p.12)
  - Figure 3. Contingent Liability Realizations by Year and Type (p.14)
  - Figure 5. Contingent Liability Realizations by Year and Country Group (p.15)
  - Figure 9. Expected Yearly Cost of Contingent Liability Realizations and Corruption (p.23)
  - Figure 11. Bureaucratic Quality, Volatility of Growth and Cost of Contingent Liability Realizations (p.24)
- Appendices:
  - Appendix A. List of Countries (p.26)
  - Appendix B. Data Sources (p.26)
  - Appendix C. Robustness Tests (p.27)
  - Appendix D. Dataset (p.28)

### Policy implications and recommended practices (as emphasized in text)
- Strengthen fiscal risk disclosure:
  - Identify, quantify, and disclose government guarantees, PPP obligations, explicit financial sector support, and intergovernmental support at least annually.
  - Disclose main implicit contingent liabilities and their estimated magnitudes and likelihoods in fiscal risk summaries.
- Institutional improvements:
  - Improve institutional quality (reduce corruption, strengthen bureaucratic quality) to lower probability and cost of CL realizations.
  - Reduce macroeconomic volatility (growth volatility) to decrease exposure to CL realizations.
- Use of the dataset:
  - Policymakers and analysts should use comprehensive, cross-checked datasets (including SRs) and stock-flow/forecast error indicators to detect and prepare for CL episodes.

*Source: _wp1614 - References (IMF working paper content provided).*

### Appendix B provides a detailed list of sources for all data used in this study.

### _wp1614 - Appendix B provides a detailed list of sources for all data used in this study.

### Methodology: debt decomposition and identification of CL realizations
- Stock-flow identity in nominal terms (equation (1)):
  - SFOBDD
    1
    (where D_t denotes gross public debt in nominal terms at time t, OB_t denotes the overall balance and SF_t denotes the residual term referred to as the stock flow adjustment).
- Debt-to-GDP decomposition (equation (2)):
  - tt tttt ttt t t t t t t sfobdd sfobddd sfobdd         1 11 1 1 1 1    
    (where γ_t denotes the nominal GDP growth rate at time t, small letters denote variables in percent of GDP, Δd_t denotes the annual change in gross public debt to GDP ratio, and λ_t = (1+γ_t)/γ_t captures GDP growth effects).
- Forecast error decomposition (equation (3)):
  - ttttt bodd  ~ ~ ~ 1
    (where x~_t = x_t - E_{t-1}[x_t] is the difference between the IMF WEO forecast for year t made in year t-1 and the outturn for year t based on WEO data submitted in year t+1; ε_t is the forecast error residual; residual sf_t captures unexpected increases in debt from contingent liabilities and other factors).
- Data sources used for calculations: IMF’s WEO and Fiscal Monitor databases (fiscal balances, interest payments, public debt, and GDP).
- Three pillars for identifying CL realizations:
  - Stock-flow adjustment and forecast error decompositions to flag large positive SF and forecast errors (distribution heavily right skewed; distribution truncated below 0 and above 50 percent of GDP).
  - Follow-up verification for right-tail observations using as many sources as possible to confirm CL realization.
  - Key word searches in Surveillance Reports (SRs) and complementary sources (Debt Sustainability Analyses, Selected Issues Papers, IMF Fiscal Transparency Evaluations, academic papers, Ministries of Finance, Central Banks). Keywords: “recapitalization,” “capital injection,” “restructuring,” “natural disaster,” “contingent,” “SOEs,” “PPP,” etc.
- For each CL episode recorded where possible: start year, end year, implicit or explicit CL, type (financial sector, SOE, etc.), fiscal response, fiscal cost, trigger, source, and short verbal description. Appendix D provides overview of collected data.

### Data and descriptive statistics
- Total CL realizations captured: 230.
- CL realizations with identified fiscal cost: 174.
- Distribution of fiscal cost (174 realizations):
  - Mean CL realization = 6.1 percent of GDP.
  - Median CL realization = 2.3 percent of GDP.
  - Several episodes exceed 20 percent of GDP; a fairly large number exceed 10 percent of GDP.
- Summary table by type of contingent liabilities (preserve entries exactly):

  - Type of Contingent Liabilities | Number of Episodes | Number of Episodes with Identified Fiscal Costs | Avg. Fiscal Costs (% GDP) | Maximum Fiscal Costs (% of GDP)
  - Financial | 91 | 82 | 9.7 | 56.8
  - Legal | 9 | 9 | 7.9 | 15.3
  - Subnational Government | 13 | 9 | 3.7 | 12.0
  - SOEs | 32 | 31 | 3.0 | 15.1
  - Natural Disaster(s) | 65 | 22 | 91.6 | 6.0
  - Private Non-Financial Sector | 76 | 1 | 7.4 | 4.5
  - PPPs | 8 | 5 | 1.2 | 2.0
  - Other | 5 | 3 | 1.4 | 2.5
  - Total | 230 | 174 | 6.1 | 56.8

  (Source: Authors’ calculations — table reproduced exactly as in source.)

- Prevalence of implicit versus explicit CLs:
  - Over 80 percent of CL realizations in the dataset stem from implicit rather than explicit contingent liabilities.

### Frequency, probabilities, and country-level experience
- Aggregation rule: sum all episodes that start in exactly the same year in the same country to obtain one country-year observation.
- Probability and cost statistics (Table 2 reproduced exactly):

  - Size (In percent of GDP) | Probability of CL Realization | Average CL Realization in percent of GDP | Number of Years until CL Realization on average = (1/probability)
  - >5 | 2.8% | 15.5 | 36
  - >1 | 5.6% | 9.2 | 18
  - Any | 8.7% | 6.1 | 12

  (Source: Authors’ calculations.)

- Interpretation:
  - Average country probability of a macro relevant CL realization in any given year = 8.7 percent → one CL realization every 12 years on average.
  - Average country would be expected to have experienced a CL realization twice in the twenty five-year sample period, with a fiscal cost of 6.1 percent of GDP per episode.
  - Probability of suffering a CL realization of at least 5 percent of GDP = 2.8 percent; conditional fiscal cost = 15.5 percent of GDP.
- Cross-country heterogeneity:
  - Over the twenty five-year period average number of CL realizations per country ≈ 2 (median = 2); several countries up to 4–5 realizations.
  - Table 3 country examples (exact entries):
    - Argentina: Episodes over 1 percent of GDP = 4; Total number of episodes = 5; Average cost per episode (percent of GDP) = 7.9
    - Brazil: Episodes over 1 percent of GDP = 4; Total number of episodes = 6; Average cost per episode (percent of GDP) = 8.3
    - China: Episodes over 1 percent of GDP = 2; Total number of episodes = 10; Average cost per episode (percent of GDP) = 4.7
    - Hungary: Episodes over 1 percent of GDP = 4; Total number of episodes = 4; Average cost per episode (percent of GDP) = 1.5
    - Indonesia: Episodes over 1 percent of GDP = 3; Total number of episodes = 6; Average cost per episode (percent of GDP) = 15.8
    - Ukraine: Episodes over 1 percent of GDP = 4; Total number of episodes = 4; Average cost per episode (percent of GDP) = 2.9

### When CL realizations occur: macro connections and crisis timing
- Event-study evidence (average dynamics around CL realizations):
  - Debt increases by over 15 percent of GDP during a CL realization on average; it rises for roughly three years and then stabilizes at a higher level.
  - Overall fiscal balance falls by about 2 percentage points as a share of GDP on average and remains below the pre-event level for an extended period.
  - GDP growth drops sharply for two years and then reverts to trend.
- Correlation between CL realizations and crises (Table 4 coefficients reproduced exactly for columns (1) and (2)):
  - Dependent variables: All CL Realization (column 1); Financial Sector CL Realization (column 2).
  - Systemic Banking Crisis: 4.439*** (column 1); 5.553*** (column 2)
    - (standard errors: (0.410) and (0.379) respectively)
  - Currency Crisis: 0.795 (0.507) ; 0.806 (0.680)
  - Sovereign Debt Crisis: 0.729 (1.398) ; -0.722 (1.111)
  - Constant: -2.465*** (0.0853) ; -4.137*** (0.183)
  - Observations: 2,000 ; 2,000
  - Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
  - (Source: Authors’ calculations.)
- Correlations among CL types (Table 5 reproduced exactly, showing pairwise correlation coefficients with p-values in parentheses; * indicates significance at 5 percent level):
  - Financial vs Private non-financial: 0.0032 (0.8843)
  - Financial vs SOE: 0.0760* (0.0006)
  - Financial vs Subnational: 0.0634* (0.0043)
  - Financial vs Legal: 0.0005 (0.9808)
  - Financial vs PPP: 0.1643* (0)
  - Private non-financial vs SOE: -0.012 (0.5903)
  - Private non-financial vs Subnational: -0.0086 (0.6989)
  - Private non-financial vs Legal: -0.0137 (0.537)
  - Private non-financial vs PPP: -0.0068 (0.7584)
  - SOE vs Subnational: 0.1117* (0)
  - SOE vs Legal: 0.1796* (0)
  - SOE vs PPP: 0.1220* (0)
  - Subnational vs Legal: 0.0527* (0.0177)
  - Subnational vs PPP: 0.2178* (0)
  - Legal vs PPP: -0.021 (0.3442)
  - (P-values in parentheses. * indicates significance at 5 percent level. Source: Authors’ calculations.)
- Regression evidence on triggers of CL realizations:
  - Logit specification (equation (4)):
    - logit(E(Y_i | X_i)) = x_i' β
      (where Y_i is an indicator for a CL realization, x_i is a vector of covariates, β is a vector of regression coefficients).
  - Key regression findings (summary):
    - Probability of a CL realization is positively correlated with lagged GDP growth and negatively correlated with contemporaneous GDP growth (column 1 of Table 6).
    - Results robust to inclusion of contemporaneous and lagged monetary variables (inflation and exchange rate), dummy for country-specific systemic banking crises, and volatility of growth.
    - Past inflation and depreciations do not seem to affect the probability; volatility of growth does not have a consistent effect.
    - Interpretation: boom periods (high lagged growth) may coincide with excessive risk-taking; growth reversals (weak contemporaneous growth) are associated with higher probability of CL realizations.
  - Robustness checks (Appendix C references):
    - Using output gap instead of GDP growth yields very similar results.
    - Results qualitatively unchanged when including natural disasters.
    - Table C.1 shows robustness to past fiscal variables (debt and deficit), oil prices, year fixed effects, and using linear probability model or cost rather than occurrence as dependent variable.

### Institutions, transparency, and CL realizations
- Summary insight:
  - Countries with stronger institutions and lower volatility of growth are less exposed to CLs.
  - Fiscal transparency and accountability help identify, monitor, and prevent fiscal risks.
  - Strengthening institutions is highlighted as a key step in preventing costly CL realizations.

### Key statistics and stylized facts (compact bullets)
- Total CL realizations recorded = 230.
- Episodes with identified fiscal cost = 174.
- Mean fiscal cost per CL realization = 6.1 percent of GDP.
- Median fiscal cost per CL realization = 2.3 percent of GDP.
- Debt increases by over 15 percent of GDP on average during a CL realization.
- Overall fiscal balance falls by about 2 percentage points of GDP on average during a CL realization.
- Average annual probability of any macro relevant CL realization for a country = 8.7 percent → one event every 12 years on average.
- Probability of CL realization >5 percent of GDP = 2.8 percent; conditional average cost = 15.5 percent of GDP → one event every 36 years on average for >5 percent events.
- Financial sector CL realizations are the most costly on average: Avg. Fiscal Costs = 9.7 percent of GDP; Maximum Fiscal Costs = 56.8 percent of GDP (table entries preserved exactly).
- Over 80 percent of CL realizations in dataset are implicit rather than explicit.

_Authors’ calculations (content reproduced from source)._

### 6.1 percent of GDP. This gives an expected cost of 0.53 percent of GDP each year.

### _wp1614 - 6.1 percent of GDP. This gives an expected cost of 0.53 percent of GDP each year.

### Expected cost metric and headline statistics
- Expected cost: 0.53 percent of GDP each year.
- Aggregate example statistic cited: 6.1 percent of GDP.
- Interpretation: countries should expect on average ½ percent of GDP of debt annually due to CL materializations.
- Long-run implication: roughly 10 percentage points rise in debt to GDP ratio over a twenty-year period.
- Database coverage: more than 200 episodes across 80 countries over the period 1990–2014.

### Empirical findings on institutional quality and fiscal cost
- Corruption:
  - Figure 9 shows countries with lower corruption scores have a lower expected cost of contingent liability (CL) realizations.
- Bureaucratic quality:
  - Figure 10 shows a significant difference between good and bad performers.
  - Quantitative result: the expected fiscal cost is over 30 percent higher in countries with below median bureaucratic quality than in those with an above median score.
- Volatility channel:
  - Panel 11(a): the standard deviation of growth in countries with below median bureaucratic quality is nearly double that in countries with an above median score.
  - Panel 11(b): the expected cost of CL realizations is nearly 50 percent higher in countries with above median volatility of growth than in those with below median volatility.
- Robustness:
  - Results in Figures 9–11 are virtually unchanged when excluding natural disasters.

### Stylized facts from the database and regression analysis
- Costliest CL shocks are related to the financial sector.
- CL realizations tend to occur at times of crisis and many materializations occur simultaneously ("when it rains it pours").
- Boom-bust cycles can act as triggers for CL materializations.
- Countries with stronger institutions and lower volatility of growth tend to suffer less from CL realizations.
- Regression highlights (selected coefficients and significance reported in the source tables):
  - GDP Growth coefficients in CL Realization regressions: negative and statistically significant in multiple specifications (e.g., -0.111***, -0.0726***, -0.0882**).
  - Systemic Banking Crisis dummy: positive and significant (e.g., 4.077***, 3.941***, 4.476***).
  - Lagged GDP Growth often positive and significant (e.g., 0.0627***, 0.0502**, 0.0714*).

### Policy recommendations and institutional measures
- Strengthen fiscal frameworks, analysis, and reporting of fiscal risks, including CLs.
- Minimize probability of CL realizations through:
  - Limiting direct exposure of public sector entities.
  - Requiring beneficiaries of guarantees to post collateral.
  - Requiring banks to hold sufficient capital.
- Improve governance to reduce volatility and moral hazard:
  - Enhance bureaucratic quality and capacity to govern without drastic policy interruptions.
  - Adopt regulatory, risk-sharing, and stronger governance measures to mitigate CL risk.

### Data limitations and avenues for further research
- Some episodes may not have been captured; dataset could be extended to pre-1990 periods and low-income countries.
- The dataset enables:
  - More analytical work on causes and consequences of CL realizations.
  - Comparisons across countries and time to inform prevention and management of CL shocks.

*Source: Authors’ calculations.*

### 2.4 SR 1999, SR 2000 Recapitalization of two major state owned banks (Belarusbank and

### _wp1614 - 2.4 SR 1999, SR 2000 Recapitalization of two major state owned banks (Belarusbank and Belagroprombank) in 1999, equivalent to 2.4% of GDP.

### Notable recapitalizations and fiscal impacts (selected entries from the dataset)
- Belarus 1999–1999 Financial Sector: Recapitalization of two major state owned banks (Belarusbank and Belagroprombank) in 1999, equivalent to 2.4% of GDP.  
- Belgium 2005–2005 SOEs: Eurostat classifies the assumption by government Railway Infrastructure Fund (FIF) in 2005 of EUR 7 400 million (2.5% of GDP) of the debt of SNCB as expenditure (capital transfer).  
- Belgium 2008–2011 Financial Sector: 13.5% of GDP SR 2012, SR 2013, World Bank (2012) — Recapitalization of SOBs between 2008 and 2011.  
- Bosnia and Herzegovina 2004–2004 Financial Sector: 6.4% of GDP Eurostat (2015) — Bailout of the largest banks in the country.  
- Bosnia and Herzegovina 2007–2009 Legal: 15.3% of GDP Flanagan (2008), SR 2008, SR 2007, SR 2006 — Increase in debt by 11.5 percent of GDP in 2007 due to recognition of frozen foreign currency deposits and war claims. Total NPV of settlement 15.3% of GDP.  
- Brazil 1990–1994 Financial Sector: restructuring of public banks and resolution of private ones (Laeven and Valencia (2012)) — various fiscal impacts listed in dataset.  
- Brazil 1993–1993 Subnational Government: 7.0% of GDP Cordes and others (2014), Cebotari and others (2009) — Subnational debt R$39.4bn refinanced with federal public banks.  
- Brazil 1994–1998 Financial Sector: 13.2% of GDP Laeven and Valencia (2012) — restructuring and resolution including “good bank/bad bank” approaches.  
- Brazil 1996–2002 Legal: 8.4% of GDP SR 2000, Cebotari et al. (2009), SR 2005 — Recognition of old debt as part of fiscal transparency moves; source provides year-by-year SR 2000 and other breakdowns.  
- Bulgaria 1996–1997 Financial Sector: 14.0% of GDP Laeven and Valencia (2012) — Bailout and recapitalization mainly of state-owned banks.  
- Canada 2009–2009 Private Non-Financial Sector: 0.9% of GDP 2010 SIP — 9.7 CAD billion from Federal Government and 4.9 CAD billion from Ontario to Chrysler and General Motors in the form of transfers of debt obligations and equity shares.  
- Chile 2006–2010 Other: 2.5% of GDP SR 2008 — 0.5% of GDP per year to recapitalize the Central Bank over 5 years.  
- China 1998–1998 Financial Sector: 18.0% of GDP Laeven and Valencia (2012) — Laeven and Valencia put total fiscal cost in 1998 at 18 percent of GDP; other sources estimate capital injections/restructuring of largest state-owned banks accounting for about 70 percent of banking system assets with a fiscal cost estimated at 3.4% of GDP in one note.  
- China 2003–2005 Financial Sector: 9.0% of GDP SR 2006, SR 2005, SIP 2005 — Two of the four large SOBs recapitalized with USD45bn of foreign exchange reserves in 2003; USD15bn capital injection for ICBC in 2005; total financial support to ICBC estimated at USD$80bn (4.3% of GDP).  
- Colombia 1998–2000 Financial Sector: 6.3% of GDP Laeven and Valencia (2012) — widespread bank weakness and restructuring.  
- Croatia 1998–1999 Financial Sector: 6.9% of GDP Laeven and Valencia (2012) — mainly SOB bailout; four state-owned banks entered rehabilitation with overall cost 6.1% of GDP; subsequent failures raised costs.  
- Cyprus 2012–2012 Financial Sector: 9.7% of GDP Eurostat (2015), 2014 SIP — large recapitalizations due to exposure to Greece and local property developers.  
- Czech Republic 1996–2000 Financial Sector: 6.8% of GDP Laeven and Valencia (2012); BIS (2006) — recapitalization/restructuring of small private banks and large state owned banks.  
- Denmark 2008–2009 Financial Sector: 4.4% of GDP Eurostat (2015) — two subsequent bank support packages in 2008-09.  
- Dominican Republic 2003–2004 Financial Sector: 22.0% of GDP Laeven and Valencia (2012), SR 2003, SR 2005 — public assistance during banking crisis ultimately amounted to 22 percent of GDP, including an unprecedented 15 percent of GDP to resolve a single institution.  
- Ecuador 1998–2002 Financial Sector: 21.7% of GDP Laeven and Valencia (2012) — failures and runs led to closure or takeover of 16 out of 40 banks accounting for 65% of assets.  
- Egypt 1991–1991 Financial Sector: 4.5% of GDP World Bank (2003), SR 1992 — four public banks given capital assistance; US$2,090 million in U.S. dollar-denominated bonds in 1991.  
- Estonia 1992–1994 Financial Sector: 1.9% of GDP Laeven and Valencia (2012) — recapitalizations and mergers of major banks; liquidation/merger of small credit institutions.  
- Finland 1991–1995 Financial Sector: 12.8% of GDP Laeven and Valencia (2012) — government took control of three banks accounting for 31% of system deposits.  
- France 1994–1995 Financial Sector: 0.7% of GDP World Bank (2003) — Credit Lyonnais experienced serious solvency problems with estimated losses about $10 billion.  
- Germany 2008–2010 Financial Sector: 11.9% of GDP Eurostat (2015) — substantial bank bailout due to global financial crisis.  
- Greece 2009–2013 Financial Sector: 23.1% of GDP Eurostat (2015) — large bank recapitalization in 2012-13.  
- Hungary 1991–1995 Financial Sector: 10.0% of GDP Laeven and Valencia (2012) — eight banks insolvent; SOBs portfolio cleaned and recapitalized.  
- Hungary 2011–2012 SOEs: 1.3% of GDP SR 2011 — Assumption of debt of MAV and BKV in 2011.  
- Iceland 1993–1993 Financial Sector: 1.0% of GDP World Bank (2003), SR 1993 — capital injection into Landsbanki totaling ISK4.25 billion.  
- (Additional country entries follow similarly in the dataset for many episodes involving SOEs, financial sector support, natural disasters, subnational government bailouts, legal settlements, and PPP-related calls on guarantees.)

### Patterns and key statistics observed in the dataset
- Recapitalization and bailout episodes span multiple subtypes of contingent liabilities: Financial Sector, SOEs, Subnational Government, PPPs, Natural Disaster(s), and Legal.  
- Fiscal impact amounts are reported precisely as percent of GDP; examples include single-event impacts (e.g., 2.4% of GDP for Belarus 1999) and multi-year cumulative impacts (e.g., Brazil 1996–2002 Legal 8.4% of GDP).  
- Several large banking crises produced very large fiscal costs:
  - China 1998 Financial Sector: 18.0% of GDP (Laeven and Valencia (2012)).  
  - Dominican Republic 2003–2004 Financial Sector: 22.0% of GDP.  
  - Ecuador 1998–2002 Financial Sector: 21.7% of GDP.  
  - Greece 2009–2013 Financial Sector: 23.1% of GDP.  
- Natural disasters and PPP-related contingent liabilities are also material in some episodes (examples in the dataset include Chile 2010–2013 Natural Disaster(s) government contribution 4.0% of GDP for reconstruction; Colombia PPPs cumulative payments 2.0% of GDP by 2004).  
- The dataset records specific modalities of support: capital injections, assumption of enterprise debt, transfers of debt obligations and equity, recognition of previously unrecognized liabilities, and debt takeovers by central government or specialized funds.

*Source: Excerpt from the dataset listing contingent liability episodes and fiscal impact amounts as percent of GDP in the IMF working paper chapter.*

### 44.2 Laeven and Valencia

### 44.2 Laeven and Valencia

### Financial sector crises — major episodes and fiscal impacts
- Indonesia 1997–2001, Financial Sector — 56.8 percent of GDP. Widespread systemic banking crisis associated with the Asian crisis; closure of more than 60 banks and gross outlays of over 50 percent of GDP. (Laeven and Valencia (2012))
- Korea 1997–1998, Financial Sector — 31.2 percent of GDP. Large, systemic banking crisis related to the Asian crisis. (Laeven and Valencia (2008), Laeven and Valencia (2012))
- Mexico 1994–1996, Financial Sector — 19.3 percent of GDP. Currency and banking crisis with blanket guarantee, recapitalization, intervention of 15 banks (1994-2001). (Laeven and Valencia (2008), Laeven and Valencia (2012))
- Malaysia 1997–1999, Financial Sector — 16.4 percent of GDP. Large systemic banking crisis with non-performing loans peaking at 25-35 percent. (Laeven and Valencia (2008), Laeven and Valencia (2012))
- Japan 1997–2001, Financial Sector — 14.0 percent of GDP. Significant public funds for loan losses, bank recapitalizations, and depositor protection after stock market crash and real estate collapse. (Laeven and Valencia (2008), Laeven and Valencia (2012))
- Ireland 2008–2011, Financial Sector — 38.9 percent of GDP. One of the largest bank bailouts ever; recapitalization of the three largest banks. (Eurostat (2015))
- Netherlands 2008, Financial Sector — 12.8 percent of GDP. Bailout of large banks (Fortis, ABN Amro and ING Group). (Eurostat (2015))
- Portugal 2007–2014, Financial Sector — 11.0 percent of GDP. Recapitalization of all major banks. (Eurostat (2015))
- Kazakhstan 2008–2012, Financial Sector — 3.7 percent of GDP. Recapitalization of various banks after the global financial crisis. (Laeven and Valencia (2012))
- Latvia 2008–2010, Financial Sector — 9.5 percent of GDP. Nationalization of the second largest bank. (Eurostat (2015))
- Lithuania 2010–2011, Financial Sector — 2.8 percent of GDP. Financial sector support in the wake of the global financial crisis. (Eurostat (2015))
- Mexico 1995–1999, Financial Sector — 4.2 percent of GDP. Direct assistance to impaired debtor groups. (SR 2000)

### Natural disasters — estimated damages and fiscal costs
- Indonesia 1997 (fires) — Total damages estimated at 3.1% of GDP in 1997 and 1.3% of GDP in 1998. (EM-DAT International Disasters Database)
- New Zealand 2010–2015, Natural Disaster(s) — 5.0 percent of GDP. Total cost of the earthquakes (Sept-10/Feb-11) was 15 percent of GDP; costs to central government estimated at 2.8% of GDP. (SR 2013, SR 2014)
- Japan 2011–2014, Natural Disaster(s) — 4.0 percent of GDP. Fiscal cost over a few years following the Great East Japan Earthquake of 2011. (Article IV SR 2012, SR 2011)
- Iran 1990–1991, Natural Disaster(s) — 5.7 percent of GDP. Total damage due to large earthquake estimated at 9%; World Bank estimated total reconstruction costs at US$5.25 billion (US$990 million in foreign exchange). (EM-DAT International Disasters Database, SR 1991)
- Pakistan 2010, Natural Disaster(s) — Total damage from flood estimated at 5.4% of GDP. (EM-DAT International Disasters Database)
- Philippines 2014, Natural Disaster(s) — 1.0 percent of GDP. Government earmarked about ₱ 120 billion (1 percent of GDP) for reconstruction spending in 2014 after Typhoon Yolanda (EMDAT estimates total damage at 4.6 percent of GDP). (EM-DAT International Disasters Database, SR 2014)
- India 2008, Natural Disaster(s) — Estimated total damages estimated at 2.8% of GDP. (EM-DAT International Disasters Database)
- India 2014, Natural Disaster(s) — Estimated total damages estimated at 1.1% of GDP. (EM-DAT International Disasters Database)
- Oman 2007, Natural Disaster(s) — 0.4 percent of GDP. Total damage estimated at 9.57 percent of GDP; 55 million rials (0.36 percent of GDP) of cyclone-relief spending in 2007. (EM-DAT International Disasters Database, SR 2007)
- Morocco 2000, Natural Disaster(s) — 0.5 percent of GDP. Severe drought; EM-DAT estimates 2.3 percent of GDP in 1999. (EM-DAT International Disasters Database, SR 2000)

### SOEs, PPPs, legal and subnational fiscal events
- Portugal 2001–2013, SOEs — 12.1 percent of GDP. Reclassification of SOE debt into the general government. (Portugal Fiscal Transparency Evaluation (2014))
- Jordan 2011–2014, SOEs — 15.1 percent of GDP. Debt assumption and transfers to heavily loss making electricity company due to interruptions in gas supply from Egypt. (SR December 2014, SR December 2010, Article IV SR 2012)
- Macedonia FYR 1999–2010, Legal — 10.1 percent of GDP. Cash servicing of frozen foreign currency deposits; NPV of settlement estimated at 10.1% of GDP to be settled over a 12 year period. (Flanagan (2008))
- Lithuania 1996–2010, Legal — 5.4 percent of GDP. Compensation related to frozen domestic currency deposits after collapse of Soviet Union; NPV of settlement estimated at 5.4% of GDP over a 10 year period. (Flanagan (2008))
- Mexico 1997, PPPs — 1.6 percent of GDP. Toll road debt assumption including a final USD7.7bn debt assumption in 1997. (FAD "Public-Private Partnerships and Fiscal Risks" (2007))
- Indonesia 1998, PPPs — Substantial obligations on PPP contracts in power plants and roads became due during the Asian crisis. (Cebotari et al. (2009))
- Italy 2006, SOEs — 0.9 percent of GDP. Government assumptions of railroad related debt. (SR 2006, SR 2005, SIP 2005)
- Malta 2011–2015, SOEs — 1.8 percent of GDP. Restructuring of Airmalta involving a total government injection of €130 million (about 1¾ percent of GDP) to be completed by 2016. (SR 2014)
- Pakistan 2013, SOEs — 1.5 percent of GDP. Government cleared circular debt of energy companies. (SR 2013)
- New Zealand 1987–1990, Financial Sector (state bank recapitalization) — 1.0 percent of GDP. Capital injection for BNZ equal to 1 percent of GDP. (World Bank (2003))

### Representative smaller fiscal impacts and notable entries
- India 1993, Financial Sector — Non-performing assets 11% in 1993-04. (Laeven and Valencia (2012))
- India 2008, Financial Sector — 0.4 percent of GDP. Post-crisis stabilization measures included 0.4% of GDP of recapitalization of public banks. (SR 2008)
- Iran 2003–2005, Private Non-Financial Sector — 4.5 percent of GDP. Contingent liabilities related to letters of credit; realization/coverage 2.2% of GDP in 2002/03, and 1.3% of GDP in 2003/04. (SR 2004)
- Jordan 2004–2005, Financial Sector — 0.3 percent of GDP. Bank restructuring after fraud case; capital injections of 0.3-0.8% of GDP. (SR 2004)
- Norway 1991–1993, Financial Sector — 2.7 percent of GDP. Banking crisis with capital injections into three of the four largest banks. (Laeven and Valencia (2008), Laeven and Valencia (2012))
- Luxembourg 2008–2009, Financial Sector — 6.9 percent of GDP. Participation in bailouts of Fortis and Dexia. (Eurostat (2015))
- Poland 1992–1994, Financial Sector — 3.5 percent of GDP. State-owned banks experienced solvency problems and received liquidity support. (Laeven and Valencia (2008), Laeven and Valencia (2012))
- Philippines 1997–2001, Financial Sector — 13.2 percent of GDP. NPL increases necessitated recapitalization to restore asset quality. (Laeven and Valencia (2008), Laeven and Valencia (2012), Alburo (1999))

*Source: Excerpt from _wp1614 - 44.2 Laeven and Valencia (IMF PDF).*

### 0.8 SR July 2012, SR

### _wp1614 - 0.8 SR July 2012, SR

### Summary of dataset and scope
- Compilation of country episodes of realized contingent liabilities drawn from IMF Surveillance Reports (SR) and external sources, including Laeven and Valencia (2008, 2012), EM-DAT International Disasters Database, Eurostat (2015), World Bank (2003), and country-specific SRs.
- Episodes are classified by subtype of contingent liability (Financial Sector, SOEs, Natural Disaster(s), Subnational Government, PPPs, Legal, Private Non-Financial Sector).
- For natural disasters, estimated damages from EM-DAT are reported and noted as "not the fiscal costs."  
- The dataset reports start and end dates as reflected in IMF SRs or external sources and notes that these may not exactly coincide with the actual start/end dates of realizations.

### Major types of contingent liabilities and illustrative episodes
- Financial Sector
  - Russia 1998–1998: Impact amount 6.0 (percent of GDP). "720 banks (accounting for 4% of sector assets and 32% of retail deposits) were deemed insolvent."
  - Turkey 2000–2001: Impact amount 32.0 (percent of GDP). "Banks had a high exposure to the government through large holdings of public securities... Two banks were closed and 19 taken over by the Savings Deposit Insurance Fund."
  - Uruguay 2002–2005: Impact amount 20.0 (percent of GDP). "Runs on both private and public banks required significant state intervention."
  - Spain 2008–2013: Impact amount 5.0 (percent of GDP). "High NPLs as consequence of burst property bubble."
  - United Kingdom 2008–2010: Impact amount 11.4 (percent of GDP). "Substantial purchases of shares of troubled banks at the height of the financial crisis."
  - United States 2008–2012: Impact amount 4.0 (percent of GDP). "Financial crisis bailouts and support excluding support to automotive industry."
  - Sweden 1991–1995: Impact amount 3.6 (percent of GDP). "Three large banks, one of which was largely state owned, were nationalized due to sizeable loan losses."
- State-Owned Enterprises (SOEs)
  - United Arab Emirates 2008–2010: Impact amount 9.0 (percent of GDP). "Support to Government related enterprises, mainly by the government of Abu Dhabi. The number represents a rough approximation (upper bound)."
  - South Africa 2009–2010: Impact amount 0.8 SR 2010 (percent of GDP). "0.75 percent of GDP for one-off lending to the state-owned electricity company Eskom."
  - Ukraine 2008–2008: Impact amount 1.2 (percent of GDP). "1.2 percent of GDP debt payment for Naftogaz."
  - Serbia 2009–2012: Impact amount 0.8 SR 2013 (percent of GDP). "Called SOE guarantees."
- Natural Disaster(s)
  - Thailand (2011–2014): Impact amount 3.0 (percent of GDP). "EMDAT estimates the total damage at 11.7 percent of GDP. 3% of GDP reconstruction investment program initiated as a response."
  - Romania 2000: Impact amount 1.6% of GDP (total damage due to drought/flood). (Reported as "Total damage due to drought/flood estimated at 1.6% of GDP.")
  - Sri Lanka 2004–2007: Impact amount 1.5 (percent of GDP). "EMDAT estimates the total damage at 6.4 percent of GDP. Government spending on reconstruction and relief amounted to 1½ percentage points of GDP."
  - United States 2005–2007: Impact amount 0.5 (percent of GDP). "Total damage from hurricanes (Katrina) estimated at 1.2% of GDP. Fiscal cost range of US$62.3bn-100bn. We use the lower bound."
- Subnational Government
  - Spain 2012–2012: Impact amount SR 2012 (reported as central government loans of EUR18bn (1.4% pf GDP) and loans of EUR18bn (1.4% of GDP) for other regions). "Two financing facilities... deficit target was relaxed from 1.3 percent in 2011 (and 2012) to 1.5 percent of GDP under the 2012 SGP."
  - United Arab Emirates 2008–2009: Impact amount 3.2 (percent of GDP). "Emergency bailout loan from Abu Dhabi to Dubai in the wake of Dubai World debt distress."
- Legal / Frozen Deposits / Compensation
  - Serbia 1998–2015: Impact amount 11.5 (percent of GDP). "Frozen foreign currency deposits (over 30 percent of GDP at end-2001) are repaid over the next decade via mandatory conversion of citizens FX Deposits into Public Bonds... NPV of 11.5% of GDP at least."
  - Ukraine 1996–2008: Impact amount 3.5 (percent of GDP). "Compensation related to frozen domestic currency deposits... The total fiscal cost was estimated at 3.5% of GDP by Flanagan (2008)."

### Common drivers, mechanisms, and government responses (as reported)
- High non-performing loans (NPLs), often linked to lending to state-owned enterprises or exposure to property sectors, triggered bank failures and recapitalizations (examples: Romania, Spain, Thailand).
- Large exposures to government securities and exchange rate mismatches amplified banking sector vulnerability (example: Turkey 2000–2001).
- Bank nationalizations, closures, recapitalizations, and blanket deposit guarantees were frequently used to restore stability (examples: Sweden, Russia, Turkey, United Kingdom).
- Runs and cross-border spillovers (deposit dollarization, non-resident deposits) magnified crises (example: Uruguay 2002–2005 linked to Argentina crisis).
- SOE distress led to direct lending, equity injections, and guarantee calls (examples: South Africa - Eskom; Ukraine - Naftogaz; United Arab Emirates - Abu Dhabi support).
- Natural disasters produced sizable estimated damages (EM-DAT totals) and led to reconstruction packages and direct fiscal outlays (examples: Thailand 2011–2014 reconstruction program of 3% of GDP; Sri Lanka reconstruction spending of 1½ percentage points of GDP).

### Selected key statistics and fiscal impact values (preserved exactly as in source)
- Russia 1998: 6.0 (percent of GDP).
- Turkey 2000–2001: 32.0 (percent of GDP).
- Thailand 1997–2000: 43.8 (percent of GDP).
- Uruguay 2002–2005: 20.0 (percent of GDP).
- Slovenia 1992–1994: 14.6 (percent of GDP).
- Slovenia 2009–2014: 18.2 (percent of GDP).
- United Kingdom 2008–2010: 11.4 (percent of GDP).
- Serbia 1998–2015: 11.5 (percent of GDP) NPV at least (frozen FX deposits converted to public bonds).
- United Arab Emirates 2008–2010 (SOEs): 9.0 (percent of GDP) (rough approximation, upper bound).
- United States 2008–2012: 4.0 (percent of GDP).
- Sweden 1991–1995: 3.6 (percent of GDP).
- Thailand 2011–2014 (natural disaster reconstruction program): 3.0 (percent of GDP).
- Ukraine 2008–2012 (financial sector recapitalization): 4.5 (percent of GDP).
- Spain 2012 (regional rescue fund): EUR18bn (1.4% pf GDP) loans for nine regions; EUR18bn (1.4% of GDP) loans for 14 regions; deficit target relaxed to 1.5 percent of GDP under the 2012 SGP.
- United States 2005–2007 (Katrina): 0.5 (percent of GDP) used in dataset (total damage estimated at 1.2% of GDP; fiscal cost range US$62.3bn-100bn; dataset uses lower bound).

### Remarks on data construction (from source notes)
- Start and end dates reflect reporting in IMF SRs or external sources and may not coincide with actual start/end dates of realizations.
- For episodes ongoing while data was collected, 2014 was chosen as the cut-off year.
- EM-DAT damage estimates reported for natural disasters are not fiscal costs.

*Source: IMF working paper dataset compilation (excerpts from _wp1614 - 0.8 SR July 2012, SR).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1614.pdf_