## _wp1018

## Source details

**Canonical URL:** [_wp1018](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp1018.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp1018.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp1018.pdf.json)

---

### Introduction and research question
- Examines whether bank support strategies that commit more fiscal resources improve economic performance during banking crises.
- Uses the Laeven and Valencia (2008) banking-crisis database covering episodes from 1980 to the present to identify crisis response strategies.
- Classifies policy measures into two categories:
  - Measures that involve a commitment of government funds (positive weight).
  - Measures that do not involve a commitment of government funds (negative weight).
- Constructs a policy response index that weights measures according to whether they place public funds at risk, and regresses post-crisis economic performance on this index controlling for other factors.

### Evidence on fiscal costs and government support
- Fiscal cost examples cited:
  - Thai crisis (1997-98): fiscal cost net of recoveries about 35 percent of GDP.
  - Turkish crisis (2000): fiscal cost about 30 percent of GDP.
- Total support packages during the present crisis estimated at:
  - 74 percent of GDP in the U.K.
  - 73 percent of GDP in the U.S.
  - 18 percent of GDP in the Euro area.

### Methodology and data
- Empirical model: Yi = α + βPi + γ'Xi + δi + εi (equation (1)), where i indexes crisis episodes.
- Baseline dependent variable:
  - GDP growth over the period [t, t+2], where t is the year the crisis begins.
- Alternative dependent variables:
  - Crisis duration: number of quarters to return to pre-crisis peak (source: Cecchetti, Kohler, and Upper (2009)).
  - Measures of output loss (CKU and LV) and crisis “intensity” (minimum GDP growth rate during the crisis).
- Key descriptive figures for performance measures:
  - Average duration of a crisis: 11.3 quarters.
  - Correlation coefficients among GDP growth in crisis times, crisis duration, minimum growth, and the CKU loss measure: ranging from 0.70 to 0.86 in absolute value.
  - LV measure: less correlated with the other measures.

### Policy response index — construction, sample, and scores
- Index formula: P_i = Σ_k w_k d_ik, where w_k is the score for policy k and d_ik is a dummy (1 if policy k adopted, 0 otherwise).
- Scoring rules:
  - Policies that shift financial burden to government receive +1.
  - Policies that do not commit public funds receive -1.
- Examples scored +1 (potentially costly to taxpayers): explicit blanket guarantees, nationalization, bank recapitalizations with public funds, setup of an asset management company (AMC).
- Examples scored -1 (do not impose direct fiscal costs): deposit freeze, bank holiday, forbearance.
- Liquidity support: not included in the main index (unclear whether it risks government resources); included in sensitivity checks as an alternative.
- Sample and index statistics:
  - Final sample consists of 40 crisis episodes (excluding ongoing 2007 US and UK crises).
  - Policy index range: low of -2 (Argentina, 1989) to high of 4 (Jamaica, 1996; Sweden, 1991; Turkey, 2000).
  - Mean value: 1.33.
  - Standard deviation: 1.47.
  - Mode value: 1 (1/4 of the episodes).

### Control variables and robustness controls
- Baseline controls:
  - Country long-run growth potential (average growth rate over 1960 to 2007 excluding the three crisis years; alternative: average over three years prior to crisis).
  - World economic growth over the crisis period.
  - Dummy for presence of an IMF-supported program.
- Additional robustness controls:
  - Growth volatility (standard deviation over 1960-2007).
  - Dummy for presence of deposit insurance before crisis.
  - Occurrence of a currency crisis and a sovereign debt crisis around the banking crisis.
  - Pre-crisis GDP per capita, ratio of private credit to GDP, capital account liberalization (0-to-10 index).

### Instrumental variables strategy and diagnostics
- Instrument: political system variable (SYSTEM) from World Bank political institutions database, measured as of crisis year:
  - SYSTEM = 2 if parliamentary system.
  - SYSTEM = 1 if assembly-elected president.
  - SYSTEM = 0 if presidential regime.
- Empirical diagnostics:
  - Simple correlation between SYSTEM and policy index: 0.6.
  - Average index by SYSTEM: parliamentary = 2.7; assembly-elected president = 2; presidential = 0.7.
  - Correlations between SYSTEM and individual policy dummies: absolute value ranging from 0.18 to 0.49.
  - First-stage F-statistic: 17 (above Staiger-Stock threshold of 10).
  - Stock and Yogo assessment: bias of IV estimator relative to OLS approximately 10 percent or less; maximum size distortion of conventional α-level Wald test no more than five percent.
  - Sample-size note: sample size minus number of instruments equals 39.

### Baseline results (OLS and IV) — main empirical findings
- Control-variable behavior:
  - Countries with higher growth potential perform better post-crisis (significant when crisis duration is dependent variable).
  - World growth: a shortfall of 1 percentage point in world growth is associated with a growth shortfall of 2.6 percentage points in the banking crisis country, post-crisis period.
  - IMF programs: associated with worse performance during crises.
- Main findings on policy response index:
  - OLS estimates:
    - Coefficient on policy index is negative and significant for both outcome measures (post-crisis output growth and crisis duration).
    - Economic magnitudes (OLS):
      - Increase in policy index by 1 reduces output growth by almost 0.8 percentage points per year on average for three years.
      - Increase in policy index by 1 increases crisis duration by almost three quarters (0.75 quarters).
  - IV estimates (policy index instrumented with SYSTEM):
    - Coefficient remains statistically significant in both growth and duration regressions.
    - Magnitude increases relative to OLS:
      - Increase in policy index by 1 reduces growth by 0.94 percentage points per year.
      - Increase in policy index by 1 increases length of the crisis by 4.4 quarters.
- Bootstrap small-sample check (10,000 replications):
  - Standard error in output growth regression becomes 0.352.
  - Standard error in duration regression becomes 0.976.
  - Coefficient of policy index remains significant at the 5 percent level (growth regression) and at the 1 percent level (duration regression).

### Additional results and measures of crisis performance
- Policy index and crisis intensity (minimum growth rate):
  - OLS: an increase of one in the policy index results in a decline in the minimum growth rate of over one percentage point.
  - IV: standard error increases substantially and coefficient is not significant in some specifications.
- Alternative crisis-performance measures (selected estimates preserved exactly):
  - 4-year window [t, t+3]: POLICY RESPONSE OLS -0.722*** (0.243); IV -0.847** (0.400).
  - Pre-crisis trend specification: POLICY RESPONSE OLS -0.795*** (0.268); IV -0.899** (0.430).
  - Minimum growth: POLICY RESPONSE OLS -1.080** (0.461); IV -0.945 (0.755).
  - CKU output loss: POLICY RESPONSE OLS 5.775* (2.984); IV 7.958 (4.917).
  - LV output loss: POLICY RESPONSE OLS 6.618** (2.754); IV 8.799** (3.545).

### Sensitivity analyses on policy index construction
- Three alternative policy-index variants tested:
  - Drop AMC (zero weight to AMC).
  - Add liquidity support and treat it as putting fiscal resources at risk.
  - 0-1 score: give score of zero to deposit freezes, bank holidays, and regulatory forbearance (instead of -1).
- Key variant results (Table 8 highlights, numbers preserved):
  - Drop AMC: POLICY RESPONSE OLS -0.748* (0.371); IV -1.170* (0.632); First-stage F statistic 16.00; Prob > F 0.000.
  - Add liquidity support: POLICY RESPONSE OLS -0.746*** (0.264); IV -0.952** (0.485); First-stage F statistic 12.54; Prob > F 0.001.
  - 0-1 score: POLICY RESPONSE OLS -1.032*** (0.293); IV -1.298** (0.632); First-stage F statistic 8.34; Prob > F 0.007.
- Nonlinearity tests:
  - Interactions of the policy index with world growth and with the IMF program dummy were tested; no evidence found of non-linear effects.

### Fiscal policy during crises
- Fiscal expansion definition: difference between the average fiscal deficit (as a share of GDP) during the years of the crisis and the fiscal deficit in the year before the crisis (positive = increase in deficit during crisis).
- Endogeneity concern: fiscal stance is jointly endogenous with output growth; instrument for output growth in fiscal regression is world GDP growth.
- Findings:
  - Political regime has no independent effect on fiscal expansion once IMF program dummy and GDP growth (instrumented) are controlled for (Table 9).
  - Same conclusion when controlling for occurrence of a debt crisis or for the policy response index.
  - If fiscal expansion is added directly to the baseline regression, fiscal expansion has a negative and significant coefficient and the negative association between the policy index and growth during the crisis remains in both OLS and IV regressions.

### Comparison with prior literature and interpretation
- Novelty: first study to summarize crisis containment and resolution policies by their potential burden on taxpayers and to carefully address endogeneity of policy response.
- Relation to prior findings:
  - Honohan and Klingebiel (2003): found liquidity support, extensive forbearance, and guarantees associated with higher fiscal costs.
  - Claessens, Klingebiel, and Laeven (2005): found support policies do not reduce output cost; good institutions improve outcomes.
  - Cecchetti, Kohler, and Upper (2009): found AMC associated with longer crises and forbearance with larger output losses; did not address endogeneity.
- Interpreted channels for why risky taxpayer bailouts are costly:
  - Hinder private-sector-led financial sector restructuring, delaying resolution.
  - Politically influenced credit allocation, undermining efficient restructuring.
  - Effective early interventions that avoided large taxpayer bailouts may have prevented panic and asset liquidation.

### Limitations and caveats
- Sample size of 40 episodes is small.
- Insufficient information to evaluate the size of taxpayer exposures associated with each policy measure.
- Index simplification reduces diverse policy responses to a single fiscal-risk dimension; does not capture intensity, timing, sequencing, or implementation details.
- Liquidity support treatment is ambiguous.
- Country-specific conditions may make some response strategies feasible in some countries but not in others.
- Endogeneity concerns are difficult to fully lay to rest; evidence is suggestive but not definitive.

### Appendix — key summary statistics (exact figures)
- Output growth during crisis (%): Obs. 40; Mean 0.65; Std. Dev. 3.12; Min. -10.45; Max. 6.03
- Crisis duration (# quarters): Obs. 40; Mean 11.35; Std. Dev. 8.86; Min. 0; Max. 33
- CKU output loss (%): Obs. 40; Mean 18.43; Std. Dev. 28.61; Min. 0; Max. 129.3
- LV output loss (%): Obs. 37; Mean 19.91; Std. Dev. 25.91; Min. 0; Max. 97.66
- Minimum growth rate (%): Obs. 40; Mean -4.16; Std. Dev. 5.40; Min. -21.6; Max. 4.8
- Policy response index: Obs. 40; Mean 1.33; Std. Dev. 1.47; Min. -2; Max. 4
- System: Obs. 40; Mean 0.6; Std. Dev. 0.87; Min. 0; Max. 2
- Trend growth (long-term) (%): Obs. 40; Mean 3.65; Std. Dev. 1.99; Min. -3.27; Max. 8.02
- Trend growth (pre-crisis) (%): Obs. 40; Mean 1.38; Std. Dev. 4.43; Min. -13.15; Max. 8.66
- World growth during crisis (%): Obs. 40; Mean 2.84; Std. Dev. 0.57; Min. 1.35; Max. 3.78
- IMF program: Obs. 40; Mean 0.55; Std. Dev. 0.50; Min. 0; Max. 1
- Growth volatility: Obs. 40; Mean 4.68; Std. Dev. 2.10; Min. 1.57; Max. 9.89
- Deposit insurance: Obs. 40; Mean 0.5; Std. Dev. 0.51; Min. 0; Max. 1
- Currency crisis: Obs. 40; Mean 0.58; Std. Dev. 0.50; Min. 0; Max. 1
- Debt crisis: Obs. 40; Mean 0.13; Std. Dev. 0.33; Min. 0; Max. 1
- Private credit to GDP (%): Obs. 40; Mean 49.93; Std. Dev. 44.39; Min. 1.77; Max. 205.15
- Per capita GDP ($ thousand): Obs. 40; Mean 8.84; Std. Dev. 7.14; Min. 0.89; Max. 32.12
- Capital controls: Obs. 39; Mean 4.53; Std. Dev. 3.26; Min. 0; Max. 10

### Principal conclusion
- No evidence of a trade-off where larger fiscal commitments to bailouts improve post-crisis economic performance; policies that put public money at risk are associated with lower output growth and delayed recovery. Evidence is robust across OLS and IV specifications, alternative index constructions, and multiple performance measures, but limited by sample size and index simplification.

*Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp1018.pdf*

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

### _wp1018 - References .............................................................................................................

### Introduction
- The paper examines whether bank support strategies that commit more fiscal resources improve economic performance during banking crises.
- Uses the Laeven and Valencia (2008) banking-crisis database covering episodes from 1980 to the present to identify crisis response strategies.
- Policy measures are classified into two categories:
  - Measures that involve a commitment of government funds (positive weight).
  - Measures that do not involve a commitment of government funds (negative weight).
- Constructs a policy response index that weights measures according to whether they place public funds at risk, and regresses post-crisis economic performance on this index controlling for other factors.

### Evidence on Fiscal Costs and Government Support
- Fiscal cost examples cited:
  - Thai crisis (1997-98): fiscal cost net of recoveries about 35 percent of GDP.
  - Turkish crisis (2000): fiscal cost about 30 percent of GDP.
- Total support packages during the present crisis estimated at:
  - 74 percent of GDP in the U.K.
  - 73 percent of GDP in the U.S.
  - 18 percent of GDP in the Euro area.

### Main Empirical Finding
- OLS regressions indicate that post-crisis economic performance tends to be worse when policymakers adopt policies that are risky for the government budget.
- The effect is described as robust and economically sizable.
- This finding is broadly consistent with prior literature: Honohan and Klingebiel (2003) and Claessens, Kliengebiel, and Laeven (2005) also fail to find evidence of a trade-off between limiting economic costs and protecting fiscal resources.

### Endogeneity Concerns and Instrumental Variables Strategy
- A key concern: endogeneity of the policy response (milder crises may elicit less fiscal-risky policies).
- Instrumental variables approach:
  - Uses political institutions as instruments, hypothesizing that political systems conducive to larger governments are more likely to choose bank-support policies that put fiscal resources at risk.
  - Persson and Tabellini (1999) motivate the choice of political-system indicators (parliamentary vs. presidential).
  - In the data, the political-system indicator explains about 40 percent of the sample variation in the banking crisis policy response index.
- IV results:
  - Results continue to hold after instrumenting the policy index with the political-system indicator.
  - No evidence found that endogeneity biases the OLS coefficient downward.
  - Interpreted as evidence against a trade-off between faster recovery and limiting taxpayer risk.

### Interpretation, Limitations, and Caveats
- The authors advise caution in interpreting results due to several limitations:
  - Small sample size.
  - Insufficient information to evaluate the size of taxpayer exposures associated with each policy measure.
  - Specific country conditions may make some response strategies feasible in some countries but not in others.
  - Endogeneity concerns are often difficult to fully lay to rest.
- Conclusion: evidence is suggestive but does not imply that bank support measures requiring large fiscal outlays should be avoided in any circumstance.

### Organization of the Paper
- Section II explains methodology and data, including construction of the policy index and the instrumentation strategy.
- The paper discusses related literature and prior empirical contributions throughout, referencing Honohan and Klingebiel (2003), Claessens, Laeven, and Klingebiel (2005), Cecchetti, Kohler, and Upper (2009), Laeven and Valencia (2008), Brown and Dinç (2009), Dell’Ariccia, Detragiache, and Rajan (2008), Demirgüç-Kunt and Detragiache (1998), Kaminsky and Reinhart (1999), Domaç and Martinez Peria (2003), Reinhart and Rogoff (2008), Dziobeck and Pazarbasioglu (1997), Hoelscher and Quintyn (2003), and Calomiris, Klingebiel, and Laeven (2003).

*Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp1018.pdf*

### Section III reports the baseline results and sensitivity analysis and relates our finding to the

### _wp1018 - Section III reports the baseline results and sensitivity analysis and relates our finding to the

### Methodology and data — empirical model and performance measures
- Estimated model: Yi = α + βPi + γ'Xi + δi + εi (equation (1)), where i indexes crisis episodes.
- Baseline dependent variable:
  - GDP growth over the period [t, t+2], where t is the year the crisis begins.
- Alternative dependent variables:
  - Duration of the crisis: number of quarters to return to pre-crisis peak (peak defined as maximum quarterly GDP over a window of four quarters before and after the crisis). Source: Cecchetti, Kohler, and Upper (2009) (CKU).
  - Measures of output loss (CKU and LV) and crisis “intensity” (minimum GDP growth rate during the crisis).
- Key descriptive figures for performance measures:
  - Average duration of a crisis: 11.3 quarters.
  - Correlation coefficients among GDP growth in crisis times, crisis duration, minimum growth, and the CKU loss measure: ranging from 0.70 to 0.86 in absolute value.
  - LV measure: less correlated with the other measures.

### Policy response index — construction and sample
- Index construction:
  - Policy index for crisis episode i: P_i = Σ_k w_k d_ik, where w_k is the score for policy k and d_ik is a dummy (1 if policy k adopted, 0 otherwise).
  - Scoring: policies that shift financial burden to government receive +1; policies that do not commit public funds receive -1.
  - Examples scored +1 (potentially costly to taxpayers): explicit blanket guarantees, nationalization, bank recapitalizations with public funds, setup of an asset management company (AMC).
  - Examples scored -1 (do not impose direct fiscal costs): deposit freeze, bank holiday, forbearance.
  - Liquidity support is not included in the main index (unclear whether it risks government resources); included in sensitivity checks as an alternative.
- Sample and index statistics:
  - Data source: LV database; final sample consists of 40 crisis episodes (excluding ongoing 2007 US and UK crises).
  - Policy index range: low of -2 (Argentina, 1989) to high of 4 (Jamaica, 1996; Sweden, 1991; Turkey, 2000).
  - Mean value: 1.33.
  - Standard deviation: 1.47.
  - Mode value: 1 (1/4 of the episodes).

### Control variables
- Baseline control variables:
  - Country long-run growth potential (calculated as average growth rate over 1960 to 2007 excluding the three crisis years; alternative: average over three years prior to crisis in robustness tests).
  - World economic growth over the crisis period.
  - Dummy for presence of an IMF-supported program.
- Additional robustness controls considered:
  - Volatility of GDP growth (standard deviation over 1960-2007).
  - Dummy for presence of deposit insurance before crisis.
  - Occurrence of a currency crisis and a sovereign debt crisis around the banking crisis.
  - Pre-crisis GDP per capita, ratio of private credit to GDP, degree of capital account liberalization (zero-to-ten index from Fraser Institute’s Economic Freedom of the World).

### Instrumental variable for policy index and instrument validity
- Instrument used: political system variable (SYSTEM) from World Bank political institutions database, measured as of crisis year:
  - SYSTEM = 2 if parliamentary system.
  - SYSTEM = 1 if assembly-elected president.
  - SYSTEM = 0 if presidential regime.
- Empirical relationships and diagnostics:
  - Simple correlation between SYSTEM and policy index: 0.6.
  - Average index by SYSTEM: parliamentary = 2.7; assembly-elected president = 2; presidential = 0.7.
  - Correlations between SYSTEM and individual policy dummies: absolute value ranging from 0.18 to 0.49.
  - First-stage F-statistic: 17 (above Staiger-Stock threshold of 10).
  - Stock and Yogo assessment: bias of IV estimator relative to OLS approximately 10 percent or less; maximum size distortion of conventional α-level Wald test no more than five percent.
  - Sample size consideration: sample size minus number of instruments equals 39 (cited McFadden (1999) guidance).

### Limitations noted in methodology
- Index simplification:
  - Reduces diverse policy responses to a single dimension (whether policies put fiscal resources at risk).
  - Does not capture intensity of interventions or detailed implementation (timing, sequencing).
  - Sparse data on intervention intensity prevents finer disaggregation.
- Liquidity support treatment is ambiguous and treated in sensitivity analysis.
- Possible heterogeneity in index components addressed by reporting correlations with SYSTEM.

### Results — baseline specification (OLS and IV)
- Control-variable behavior:
  - Countries with higher growth potential perform better post-crisis (significant when crisis duration is dependent variable).
  - World growth: a shortfall of 1 percentage point in world growth is associated with a growth shortfall of 2.6 percentage points in the banking crisis country, post-crisis period.
  - IMF programs: associated with worse performance during crises.
- Main findings on the policy response index:
  - OLS results:
    - Coefficient on policy index is negative and significant for both outcome measures (post-crisis output growth and crisis duration).
    - Economic magnitudes (OLS):
      - Increase in policy index by 1 reduces output growth by almost 0.8 percentage points per year on average for three years.
      - Increase in policy index by 1 increases crisis duration by almost three quarters (0.75 quarters).
  - IV results (policy index instrumented with SYSTEM):
    - Coefficient remains statistically significant in both growth and duration regressions.
    - Magnitude increases relative to OLS (consistent with correction for attenuation measurement error):
      - Increase in policy index by 1 reduces growth by 0.94 percentage points per year.
      - Increase in policy index by 1 increases length of the crisis by 4.4 quarters.
- Bootstrap small-sample check:
  - With 10,000 replications:
    - Standard error in output growth regression becomes 0.352.
    - Standard error in duration regression becomes 0.976.
  - Coefficient of policy index remains significant at the 5 percent level (growth regression) and at the 1 percent level (duration regression).

### Comparison with existing literature
- Novelty:
  - First study to summarize crisis containment and resolution policies by their potential burden on taxpayers and to carefully address endogeneity of policy response.
- Relation to prior findings:
  - Honohan and Klingebiel (2003): found liquidity support, extensive forbearance, and guarantees associated with higher fiscal costs; present study differs in sample, focus on economic performance rather than fiscal cost, policy classification by fiscal risk, and IV treatment.
  - Claessens, Klingebiel, and Laeven (2005): found support policies do not reduce output cost; good institutions improve outcomes.
  - Cecchetti, Kohler, and Upper (2009): found AMC associated with longer crises and forbearance with larger output losses; did not address endogeneity.

### Sensitivity analyses
- Additional controls (Table 6):
  - Including growth volatility: volatility not significant; coefficients of other variables unchanged.
  - Including deposit insurance dummy: not significant; no alteration of main results.
  - Controlling for currency crisis and sovereign debt crisis occurrence: not significant; IMF dummy may capture these effects better.
  - Controlling for private credit/GDP, GDP per capita, and capital controls: private credit and capital control coefficients significant and as expected; policy response coefficient remains large and statistically significant at the five percent level in OLS and IV.
- Alternative measures of crisis performance (Table 7):
  - Extending crisis window to four years [t, t+3]: estimated coefficients largely unchanged; regression fit worse compared with baseline.
  - Using three-year pre-crisis average growth as trend: policy index coefficient remains very close to baseline; pre-crisis growth positively and significantly associated with growth during the crisis.
  - Measuring crisis intensity as minimum real GDP growth rate during crisis: dataset notes mean of this variable = -4.2 percent (further details truncated in provided content).

*Source: _wp1018 - Section III reports the baseline results and sensitivity analysis and relates our finding to the (PDF chapter/section).*

### 21.6 percent in Estonia in 1992 to 4.8 percent in Vietnam in 1997. The policy response index

### _wp1018 - 21.6 percent in Estonia in 1992 to 4.8 percent in Vietnam in 1997. The policy response index

### Main empirical finding on policy response and crisis performance
- The policy response index is significantly and negatively correlated with crisis intensity in the OLS regression.
- In the IV regression the standard error increases substantially and the coefficient is not significant in some specifications.
- Based on the OLS estimates, an increase of one in the policy index results in a decline in the minimum growth rate of over one percentage point.
- Policies that are riskier for the government (blanket guarantees, nationalization, recapitalization with public funds, asset management companies) are positively associated with larger output losses measured by CKU and LV (IV coefficient of the CKU measure is significant only at the 10.6 percent level in one specification).
- Instrumenting policy response with political regime (SYSTEM) yields a first-stage coefficient of 0.945*** (standard error 0.229) and a First-stage F statistic of 17.02 (Prob > F = 0.000), supporting the instrument’s strength in the baseline.

### Sensitivity analysis: alternative policy response indexes
- Three alternative policy index variants were tested:
  - Drop AMC (zero weight to AMC on grounds that AMCs acquiring assets at recovery value do not result in government losses).
  - Add liquidity support and treat it as a policy that puts fiscal resources at risk.
  - 0-1 score: give score of zero to deposit freezes, bank holidays, and regulatory forbearance (as opposed to minus one in baseline).
- Key results across variants:
  - Dropping the AMC: no effect on the coefficient of the policy index; regression fit worsens slightly; IV results continue to hold.
  - Adding liquidity support: results very close to baseline for OLS and IV; first-stage F statistic is a bit lower.
  - 0-1 score variant: policy index negatively correlated with economic performance in OLS and IV; coefficient larger than baseline, but instrument is weaker (First-stage F statistic reported as 8.34 and 12.54 in some columns).
- Table 8 specific estimates (Output growth during crisis):
  - Drop AMC: POLICY RESPONSE OLS -0.748* (0.371); IV -1.170* (0.632); First-stage F statistic 16.00; Prob > F 0.000.
  - Add liquidity support: POLICY RESPONSE OLS -0.746*** (0.264); IV -0.952** (0.485); First-stage F statistic 12.54; Prob > F 0.001.
  - 0-1 score: POLICY RESPONSE OLS -1.032*** (0.293); IV -1.298** (0.632); First-stage F statistic 8.34; Prob > F 0.007.

### Nonlinearity tests
- Interactions of the policy index with world growth and with the IMF program dummy (in separate regressions) were tested to detect non-linear effects.
- No evidence was found that policy orientation effects on crisis performance are non-linear (results not reported).

### Fiscal policy during the crisis
- Fiscal expansion is defined as the difference between the average fiscal deficit (as a share of GDP) during the years of the crisis and the fiscal deficit in the year before the crisis (a positive value corresponds to an increase in the deficit during the crisis).
- Concern: fiscal stance is jointly endogenous with output growth; instrument used for output growth in the fiscal regression is world GDP growth.
- Regressing fiscal expansion on political regime and controls (with output growth instrumented by world growth) yields:
  - Political regime has no independent effect on fiscal expansion once IMF program dummy and GDP growth (instrumented) are controlled for (Table 9).
  - Similar conclusion when controlling for occurrence of a debt crisis or for the policy response index.
- Additional note: If fiscal expansion is added directly to the baseline regression of Table 4, fiscal expansion has a negative and significant coefficient and the negative association between the policy index and growth during the crisis remains in both OLS and IV regressions.

### Baseline regression summary and robustness
- Baseline (Table 5) results for dependent variables Output growth during crisis and Crisis duration:
  - POLICY RESPONSE: OLS -0.783** (0.292); IV -0.936* (0.479) for Output growth during crisis.
  - POLICY RESPONSE: OLS 2.804*** (0.827); IV 4.381*** (1.421) for Crisis duration.
  - TREND GROWTH, WORLD GROWTH, and IMF PROGRAM coefficients presented; WORLD GROWTH positive and significant in growth regressions and negative for crisis duration, IMF PROGRAM associated with worse growth and longer crisis duration.
  - Observations 40 in baseline regressions; R-squared values reported (e.g., 0.430 for OLS growth).
- Additional controls (Table 6) show policy response coefficient remains negative and significant across alternative specifications including growth volatility, deposit insurance, currency crisis, debt crisis, private credit, per capita GDP, and capital controls. Example estimates:
  - Policy response across columns ranges from -0.784** to -0.856*** in OLS and from -0.889* to -1.016** in IV, with standard errors preserved exactly as reported.

### Alternative crisis performance measures
- Table 7 reports robustness of the negative association across measures:
  - 4-year window: POLICY RESPONSE OLS -0.722*** (0.243); IV -0.847** (0.400).
  - Pre-crisis trend specification: POLICY RESPONSE OLS -0.795*** (0.268); IV -0.899** (0.430).
  - Minimum growth: POLICY RESPONSE OLS -1.080** (0.461); IV -0.945 (0.755).
  - CKU output loss: POLICY RESPONSE OLS 5.775* (2.984); IV 7.958 (4.917).
  - LV output loss: POLICY RESPONSE OLS 6.618** (2.754); IV 8.799** (3.545).
- First-stage statistics remain strong in many of these alternative specifications (e.g., First-stage F statistic 17.02, 18.15, 30.93 depending on panel).

### Political system, policy choices, and summary statistics
- Policy index construction (Table 2): More risk to taxpayers = Blanket guarantee (1), Nationalization (1), Recapitalization (1), Asset management companies (1); Less risk to taxpayers = Deposit freeze (-1), Bank holiday (-1), Forbearance (-1).
- Distribution and episodic data (Table 3): 40 crisis episodes listed with POLICY INDEX mean 1.33 and S.D. 1.47; SYSTEM mean 0.6 and S.D. 0.87. SYSTEM coding: 2 for parliamentary regime; 1 for assembly-elected president; 0 for presidential regime.
- Cross-correlations (Table 4) show SYSTEM positively correlated with blanket guarantee (0.495), AMC (0.332), and other costly interventions; negative correlations with deposit freeze (-0.264) and bank holiday (-0.233).

### Conclusions and interpretation
- No evidence of a trade-off where larger fiscal commitments to bailouts improve post-crisis economic performance; instead, policies that put public money at risk are associated with lower output growth and delayed recovery.
- Parliamentary regimes are more likely to adopt high-fiscal-risk rescue policies; this provides a valid instrument for policy response and strengthens the IV results (policy effect becomes larger when instrumented).
- Possible channels for why risky taxpayer bailouts are costly:
  - Hinder private-sector-led financial sector restructuring, delaying resolution.
  - Politically influenced credit allocation, undermining efficient restructuring.
  - Effective early interventions that avoided large taxpayer bailouts may have prevented panic and asset liquidation.
- Limitations noted:
  - Sample size of 40 episodes is small.
  - Policy characterization is simplistic (single dimension: whether policy puts public money at risk).
  - More research is needed to explore mechanisms.

*Source: _wp1018 - 21.6 percent in Estonia in 1992 to 4.8 percent in Vietnam in 1997. The policy response index*

### References

### _wp1018 - References

### References
- Alesina, Alberto, Felipe Campante, and Guido Tabellini, 2008, “Why is Fiscal Policy Often Pro-Cyclical?,” Journal of the European Economic Association, 6, 1006-38.
- Akitoby, Bernardin, and Thomas Stratman, 2008, “Fiscal Policy and Financial Markets,” Economic Journal, 118, 1971-1985.
- Brown, Craig, O., and I. Serdar Dinç, 2009, “Too Many to Fail? Evidence of Regulatory Forbearance When the Banking Sector is Weak,” Review of Financial Studies, forthcoming.
- Calomiris, Charles, Daniela Klingebiel, and Luc Laeven, 2003, “A Taxonomy of Financial Crisis Restructuring Mechanisms,” in Patrick Honohan and Luc Laeven (Eds., Systemic Financial Crises: Containment and Resolution, (Cambridge: Cambridge University Press).
- Cecchetti, Stephen, G., Marion Kohler, and Christian Upper, 2009, “Financial Crises and Economic Activity,” Paper prepared for the 2009 Federal Reserve Bank of Kansas City’s symposium at Jackson Hole.
- Cerra, Valerie, and Sweta Saxena, 2008, “Growth Dynamics: The Myth of Economic Recovery,” American Economic Review, 98, 439–457.
- Claessens, Stijn; Klingebiel, Daniela and Luc Laeven, 2005, “Crisis Resolution, Policies, and Institutions: Empirical Evidence”, in Patrick Honohan and Luc Laeven (Eds., Systemic Financial Crises: Containment and Resolution, (Cambridge: Cambridge University Press).
- Dell’Ariccia, Giovanni, Detragiache, Enrica, and Raghuram Rajan, 2008, “The Real Effect of Banking Crises,” Journal of Financial Intermediation, 17, 89–112.
- Demirgüç-Kunt, Asli and Enrica Detragiache, 1998, “The Determinants of Banking Crises in Developing and Developed Countries,” IMF Staff Papers, 45, 81–109.
- Demirgüç-Kunt, Asli and Enrica Detragiache, and Poonam Gupta, 2006, “Inside the Crises: An Empirical Analysis of Banking Systems in Distress,” Journal of International Money and Finance, 25, 702–718.
- Domaç, Ilker and Maria Soledad Martinez Peria, 2003, “Banking Crises and Exchange Rate Regimes: Is There a Link?,” Journal of International Economics 61 pp. 41–72.
- Djankov, Simeon; McLiesh, Caralee and Andrei Shleifer, 2007, “Private Credit in 129 Countries,” Journal of Financial Economics 84, pp. 299–329.
- Dziobek, Claudia, and Celia Pazarbasioglu, 1997, “Lessons from Systemic Bank Restructuring: a Survey of 24 Countries,” IMF Working Paper 97/161.
- Frydl, Edward, 1999, “The Length and Cost of Banking Crises”, IMF Working paper 99/30.
- Hoelscher, David, and Marc Quintyn, 2003, “Managing Systemic Banking Crises,” IMF Occasional Paper, No. 224.
- Hoggarth, Glenn, Reis, R., and Victoria Saporta, 2002, “Output Costs of Banking System Instability: Some Empirical Evidence,” Journal of Banking and Finance, 26, 825–55.
- Hoggarth, Glenn, and Jack Reidhill, and Peter Sinclair, 2003, “Resolution of Banking Crises: A Review,” Bank of England Financial Stability Review, December, 109–121.
- Honohan, Patrick, and Daniela Klingebiel, 2003, “The Fiscal Cost Implications of an Accommodating Approach to Banking Crises,” Journal of Banking and Finance, 27, 1539–1560.
- Hutchinson, Michael M., and Ilan Noy, 2005, “How Bad Are Twins? Output Costs of Currency and Banking Crises,” Journal of Money, Credit, and Banking, 37, 725–752.
- Kaminsky, Graciela L. and Carmen M. Reinhart, 1999, “The Twin Crises: The Causes of Banking and Balance of Payments Problems,” American Economic Review 89, pp. 473–500.
- Klingebiel, Daniela, 2000, “The Use of Asset Management Companies in the Resolution of Banking Crises: Cross-Country Experience,” World Bank Policy Research Working Paper No. 2248.
- Kroszner, Randall S., Luc Laeven, and Daniela Klingebiel, 2007, "Banking Crises, Financial Dependence, and Growth," Journal of Financial Economics, 84, 187–228.
- Laeven, Luc, and Fabian Valencia, 2008, “Systemic Banking Crises: A New Database,” IMF Working Paper No. 08/224.
- Laeven, Luc, and Fabian Valencia, 2008, “The Use of Blanket Guarantees in Banking Crises“, IMF Working Paper 08/250.
- McFadden, Daniel L., 1999, Econometrics/Statistics Reader, available on the internet at http://www.econ.berkeley.edu/~mcfadden/e240b_sp03/e240b.html.
- Perotti, Roberto, 2004, “Estimating the Effects of Fiscal Policy in OECD Countries.” Working Paper, IGIER-Bocconi.
- Persson, Torsten and Guido Tabellini, 1999, “The Size and Scope of Government: Comparative Politics with Rational Politicians,” European Economic Review, 43, 699–735.
- Persson, Torsten; Roland, Gerard and Guido Tabellini, 2000, “Comparative Politics and Public Finance,” Journal of Political Economy, 108, 1121–1161.
- Reinhart, Carmen M., and Kenneth S. Rogoff, 2008, “Banking Crises: An Equal Opportunity Menace,” National Bureau of Economic Research Working Paper 14587.
- Staiger, Douglas and James H. Stock, 1997, “Instrumental Variables Regression with Weak Instruments,” Econometrica, 65, 557–586.
- Stock, James and Motohiro Yogo, 2002, “Testing for Weak Instruments in Linear IV Regression”, NBER Technical Working Paper No. 284.

### Appendix — Variable Definitions and Data Sources
- I. Cost of crisis measures
  - Output growth during crisis: Average growth rate of real GDP over the period [t,t+2], where t is the start year of crisis — Source: WDI
  - Crisis duration: Number of quarters until GDP reverts to pre-crisis peak — Source: Cecchetti et al. (2009)
  - CKU output loss: Cumulative GDP decline during crisis duration — Source: Cecchetti et al. (2009)
  - LV output loss: Sum of output gaps over [t,t+3], where potential output is calculated by extrapolating trend GDP growth rate during 3 pre-crisis years — Source: LV database
  - Minimum growth rate: Minimum observed growth rate during crisis — Source: LV database
- II. Policy and instrument
  - Policy response index: An index constructed by scoring policies based on the degree of risk imposed on taxpayers — Source: Authors’ calculation
  - System: Indicator of a country’s political system: 2=Parliamentary; 1=Assembly-elected president; 0=Presidential — Source: WB database of political institutions
- III. Control variables
  - Trend growth (long-term): Average trend growth rate of real GDP over 1960-2007, excluding three crisis years — Source: WDI
  - Trend growth (pre-crisis): Average growth rate over 3 pre-crisis years — Source: WDI
  - World growth during crisis: Average growth rate of world real GDP over the period [t,t+2] — Source: WDI
  - IMF program: Indicator variable (0,1), a value of 1 indicates that an IMF program was put in place in response to the crisis — Source: LV database
  - Growth volatility: Standard deviation of GDP growth, 1960-2007 — Source: WDI
  - Deposit insurance: Indicator variable (0,1), a value of 1 indicates that a deposit insurance scheme was in place at the start of crisis — Source: LV database
  - Currency crisis: Indicator variable (0,1), a value of 1 indicates that a currency crisis occurred during [t-1,t+1] — Source: LV database
  - Debt crisis: Indicator variable (0,1), a value of 1 indicates that a sovereign debt crisis occurred during [t-1,t+1] — Source: LV database
  - Private credit to GDP: Ratio of bank credit to private sector to GDP, averaged over 3 pre-crisis years — Source: WDI
  - Per capita GDP: GDP per capita as of t-1 — Source: WDI
  - Capital controls: 0-to-ten rating based on degree of international capital control and foreign ownership/investment restrictions, measured at most 5 years before crisis year — Source: “Economic Freedom of the World” database, Fraser Institute

### Table A2 — Summary Statistics (naming and exact figures preserved)
- Output growth during crisis (%): Obs. 40; Mean 0.65; Std. Dev. 3.12; Min. -10.45; Max. 6.03
- Crisis duration (# quarters): Obs. 40; Mean 11.35; Std. Dev. 8.86; Min. 0; Max. 33
- CKU output loss (%): Obs. 40; Mean 18.43; Std. Dev. 28.61; Min. 0; Max. 129.3
- LV output loss (%): Obs. 37; Mean 19.91; Std. Dev. 25.91; Min. 0; Max. 97.66
- Minimum growth rate (%): Obs. 40; Mean -4.16; Std. Dev. 5.40; Min. -21.6; Max. 4.8
- Policy response index: Obs. 40; Mean 1.33; Std. Dev. 1.47; Min. -2; Max. 4
- System: Obs. 40; Mean 0.6; Std. Dev. 0.87; Min. 0; Max. 2
- Trend growth (long-term) (%): Obs. 40; Mean 3.65; Std. Dev. 1.99; Min. -3.27; Max. 8.02
- Trend growth (pre-crisis) (%): Obs. 40; Mean 1.38; Std. Dev. 4.43; Min. -13.15; Max. 8.66
- World growth during crisis (%): Obs. 40; Mean 2.84; Std. Dev. 0.57; Min. 1.35; Max. 3.78
- IMF program: Obs. 40; Mean 0.55; Std. Dev. 0.50; Min. 0; Max. 1
- Growth volatility: Obs. 40; Mean 4.68; Std. Dev. 2.10; Min. 1.57; Max. 9.89
- Deposit insurance: Obs. 40; Mean 0.5; Std. Dev. 0.51; Min. 0; Max. 1
- Currency crisis: Obs. 40; Mean 0.58; Std. Dev. 0.50; Min. 0; Max. 1
- Debt crisis: Obs. 40; Mean 0.13; Std. Dev. 0.33; Min. 0; Max. 1
- Private credit to GDP (%): Obs. 40; Mean 49.93; Std. Dev. 44.39; Min. 1.77; Max. 205.15
- Per capita GDP ($ thousand): Obs. 40; Mean 8.84; Std. Dev. 7.14; Min. 0.89; Max. 32.12
- Capital controls: Obs. 39; Mean 4.53; Std. Dev. 3.26; Min. 0; Max. 10

### Table A3 — Cross Correlations among Variables (row-wise entries)
- Trend growth / Pre-crisis growth: 0.6083
- Trend growth / World growth during crisis: 0.0208
- Trend growth / IMF program: -0.0702
- Trend growth / Growth volatility: -0.6122
- Trend growth / Deposit insurance: -0.1374
- Trend growth / Currency crisis: 0.0688
- Trend growth / Debt crisis: -0.0182
- Trend growth / Private credit: 0.465
- Trend growth / GDP per capita: 0.0856
- Trend growth / Capital controls: -0.0185

- Pre-crisis growth / World growth during crisis: 0.0542
- Pre-crisis growth / IMF program: -0.1629
- Pre-crisis growth / Growth volatility: -0.5024
- Pre-crisis growth / Deposit insurance: 0.0213
- Pre-crisis growth / Currency crisis: 0.0543
- Pre-crisis growth / Debt crisis: -0.0811
- Pre-crisis growth / Private credit: 0.2301
- Pre-crisis growth / GDP per capita: 0.0134
- Pre-crisis growth / Capital controls: 0.1033

- World growth during crisis / IMF program: 0.0564
- World growth during crisis / Growth volatility: 0.2499
- World growth during crisis / Deposit insurance: -0.07
- World growth during crisis / Currency crisis: -0.2448
- World growth during crisis / Debt crisis: 0.0581
- World growth during crisis / Private credit: -0.0971
- World growth during crisis / GDP per capita: -0.3857
- World growth during crisis / Capital controls: 0.1939

- IMF program / Growth volatility: 0.2941
- IMF program / Deposit insurance: 0.1777
- IMF program / Currency crisis: 0.5283
- IMF program / Debt crisis: 0.1824
- IMF program / Private credit: -0.2262
- IMF program / GDP per capita: -0.2544
- IMF program / Capital controls: -0.1046

- Growth volatility / Deposit insurance: -0.1052
- Growth volatility / Currency crisis: -0.0859
- Growth volatility / Debt crisis: 0.0149
- Growth volatility / Private credit: -0.3893
- Growth volatility / GDP per capita: -0.3044
- Growth volatility / Capital controls: -0.011

- Deposit insurance / Currency crisis: 0.1257
- Deposit insurance / Debt crisis: 0.0669
- Deposit insurance / Private credit: -0.1013
- Deposit insurance / GDP per capita: 0.3552
- Deposit insurance / Capital controls: 0.1001

- Currency crisis / Debt crisis: 0.1639
- Currency crisis / Private credit: 0.069
- Currency crisis / GDP per capita: -0.0383
- Currency crisis / Capital controls: -0.1687

- Debt crisis / Private credit: -0.218
- Debt crisis / GDP per capita: -0.1552
- Debt crisis / Capital controls: -0.0425

- Private credit / GDP per capita: 0.5481
- Private credit / Capital controls: 0.1153

- GDP per capita / Capital controls: 0.3027

*Source: _wp1018 - References (Appendix tables and bibliography).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp1018.pdf_