## Default Episodes

## Source details

**Canonical URL:** [Default Episodes](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08238.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08238.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08238.pdf.json)

---

### Introduction and objectives
- Purpose: Evaluate empirically mechanisms through which default costs may affect a sovereign government, focusing on four cost types: reputational costs, international trade exclusion costs, costs to the domestic economy through the financial system, and political costs to the authorities.
- Scope: Analyze incidence of those cost channels and assess whether there is empirical basis for—or evidence against—each mechanism rather than precisely quantifying the cost of default.
- Methodological caution: Difficult to isolate causal costs of default from other factors; direction of causality between growth and default is hard to identify.
- Policy relevance: Identification of channels and magnitudes informs the sovereign “default point” concept and affects computation of default probabilities and sovereign bond pricing.

### Conceptual background
- Distinction: Creditor legal remedies weaker for sovereigns; many sovereign assets immune from legal action.
- Definition: “Default” includes any situation where sovereign does not honor original debt contract, including voluntary restructurings where creditors lose value (aligned with credit rating agencies’ concept).
- Theoretical mechanisms emphasized:
  - Reputational costs (Eaton and Gersovitz (1981) framework).
  - Direct sanctions / trade embargoes (Bulow and Rogoff (1989) “sanctions view”).
  - Domestic costs via banking-sector exposures.
  - Political costs from economic decline and banking crises.

### Empirical strategy and limitations
- Objective: Evaluate empirical basis for each mechanism rather than precise cost estimates.
- Identification challenges:
  - Defaults often follow shocks (terms of trade, sudden stops, currency crises) that also depress growth.
  - Decomposition approach: DEFAULT = default_pred_it + v_it, where default_pred_it is predicted probability (logit similar to Manasse et al. (2003)) and v_it is unanticipated residual; anticipated vs unanticipated components interpreted accordingly.
- Approach: Examine relationships between default and GDP growth, borrowing costs (credit ratings and spreads), international trade and trade credit, banking crises, industry-specific effects, and political outcomes.

### Main empirical findings (overview)
- Reputational and borrowing-cost effects: Significant but appear short-lived.
- Trade and trade credit effects: Trade and trade credit negatively affected by default; controlling for trade credit does not modify default effect on trade.
- Growth effects: GDP per capita growth suffers following defaults; negative effect concentrated at impact and short-lived.
- Banking sector: Defaults tend to cause banking crises (default → banking crisis) more than vice versa.
- Industry-level credit crunch: Outside banking-crisis episodes, more credit-dependent industries do not suffer more following sovereign default.
- Political consequences: Severe political costs for incumbent governments and finance ministers, similar to currency crises.

---

### II. Two hundred years of sovereign default — incidence, timing, and duration
- Data sources compared: Standard and Poor’s (first four columns of Table A1), Beim and Calomiris (2000) (columns 5–6), Sturzenegger and Zettelmeyer (2006) (column 7), and Detragiache and Spilimbergo (2001) (last column).
  - Beim and Calomiris tend to merge defaults within five years into single episodes (fewer but longer episodes).
  - Detragiache and Spilimbergo code several episodes not classified as defaults by Standard and Poor’s.
  - Paper uses Standard and Poor’s classifications for remainder due to relative completeness.
- Geographic counts (1824–2004):
  - Latin America: 126 default episodes.
  - Africa: 63 default episodes.
  - Asia: lowest number of defaults among developing regions.
- Default waves and episode lengths by period (selected):
  - 1824–1840: 19 episodes; 14 involved Latin America; average length: more than twenty years.
  - 1841–1860: 6 episodes.
  - 1861–1920: 58 episodes; 41 in Latin America; average length dropped to less than five years by 1881–1920.
  - 1921–1940: 39 episodes; more than half in Latin America and 16 in Europe.
  - 1941–1970: 6 episodes total (Hungary 1941, Japan 1942, Czechoslovakia 1959, Cuba 1960, Costa Rica 1962, Zimbabwe 1965).
  - 1970s: 15 syndicated bank loan default episodes.
  - 1982 debt crisis era: more than 70 default episodes (34 in Africa, 29 in Latin America); average default length ≈ 9 years.
  - 1991–2004: 40 defaults (14 on bonds; 26 on syndicated bank loans).

---

### III. Default and GDP growth — empirical findings and decomposition
- Regression framework:
  - Unbalanced panel up to 83 countries for 1972–2000.
  - Baseline model: GROWTH_it = α + β DEFAULT_it + γ X_it + ε_it, where GROWTH_it is per capita annual real GDP growth; DEFAULT_it is set of dummy variables tracking default episodes.
  - Controls include INV_GDP, POP_GR, GDP_PC70s, SEC_ED, POP, GOV_C1, CIV_RIGHT, DTOT, OPEN, BK_CR, and region dummies (SSA, LAC, TRANS).
- Main growth effects (Table 2 and Table 3 highlights):
  - Baseline (Table 2, column 1): DEFAULT associated with decrease in growth of 1.2 percentage points per year (DEF = -1.239; t-statistic (4.32)***).
  - Dynamic (Table 2, column 2): large impact in first year: drop in growth of 2.6 percentage points (DEF_B on impact = -2.6); lagged default variables not statistically significant.
  - Restricted-sample developing countries (Table 3, column 1): effect slightly smaller but still sizable and highly significant.
- Anticipated vs unanticipated (Table 3, column 2):
  - DEF_PR (anticipated effect): estimate -1.4 percent (negative impact on growth).
  - DEF_U (unanticipated effect): close to -1.0 percent (negative).
  - Both components statistically significant.
- Interpretation:
  - Unanticipated component may capture costs of “unjustified” defaults (weak willingness to pay vs inability), consistent with ability/willingness to pay literature.
  - Effects concentrated at impact and short-lived; often no detectable effects beyond one or two years.

---

### IV. Default and reputation — access, costs, and market discrimination
- Market access:
  - Default does not lead to permanent exclusion from international capital markets.
  - Countries lose access during default but regain access after restructuring; external factors/investor mood often more important determinants.
  - Gelos et al. (2004): countries that defaulted in the 1980s regained access in about 4 years.
- Credit ratings regressions (Table 4):
  - Rating model R^2 = 0.91 in column 1 (explains more than 90 percent of cross-country variance in credit ratings).
  - DEFAULT coefficient examples: -1.669; -1.486; -1.855 (t-statistics (3.10)***, (2.86)***, (3.57)***).
  - Estimated rating effect summarized: "1.7 notches, slightly lower than the estimate of Cantor and Packer (2.5 notches)."
  - Robustness: adding DEBT_GDP and OR_SIN changes sample by 16 observations but leaves results essentially unchanged; SDTOT not statistically significant; only defaults in the 1995–2002 period significantly correlated with credit ratings over 1999–2002.
- Bond spreads (Table 5 highlights):
  - DEF1YEAR coefficients across columns: 412.863; 307.746; 433.912; 305.783; 389.342; 249.764; 267.770; 249.175 (many significant).
  - DEF2YRS coefficients across columns: 246.746; 188.244; 267.262; 162.114; 238.877; 145.339; 134.276; 144.640.
  - Interpretation: default raises bond spreads substantially on impact; effect attenuates over time.

---

### Trade credit, defaults, and bilateral trade
- Trade-credit regression (Table 6 and Table 7 summaries):
  - Column 6 (Arellano and Bond estimator): effect of default on trade credit is negative and large only in the first and second year of default; negative effect short lived.
  - Gravity model (equation (5)) results (Table 7):
    - DEF_NS (dummy if either country in pair in default) coefficients across columns: -0.206; -0.319; -0.054; -0.054; -0.047; -0.104 (various significance levels).
    - NSTC (log stock of trade credit) positive and significant; estimated elasticity of trade to trade credit ≈ 7 percent.
    - Controlling for trade credit does not affect relationship between default and trade: coefficients of DEF_NS and DEF_AVT unchanged between columns 3 and 4.
  - Interpretation: trade and trade credit suffer around defaults; effects tend to be short lived; trade-credit channel does not fully explain trade decline.

---

### Default and the domestic banking system
- Data and unconditional counts (1975–2000, 149 countries, 3,874 observations):
  - 111 banking crises (unconditional probability = 2.9 percent).
  - 85 default episodes (unconditional probability = 2.2 percent).
- Conditional probabilities (Table 8 exact figures):
  - Probability of a banking crisis in year t conditional on debt default in year t or t−1 = 14.1 percent.
  - This is an 11 percentage point increase relative to unconditional probability (2.9 percent); difference statistically significant (P value = 0.0 for test P(BC/DEF)>P(BC)).
  - Probability of default conditional on having a banking crisis = 4.5 percent (difference from unconditional default probability 2.2 percent; P value = 0.1).
- Interpretation: default episodes appear to increase probability of banking crisis more than banking crises increase probability of default; caveats due to few twin crisis cases and annual timing.
- Industry-level credit-crunch test (Table 9 results):
  - Specification follows Rajan and Zingales (1998) with DEF × EXT interaction.
  - Coefficients on DEF × EXT not statistically significant in various specifications; column 1 DEF positive (wrong sign) and not significant; columns splitting first/second/third year show negative coefficients but not statistically significant.
  - BK_CR*EXT coefficient examples: -2.277; -2.164; -2.282 (t-statistics (2.29)**, (2.16)**, (2.29)**) indicating banking crises interact with external dependence to hurt sectors, but default itself does not show consistent credit-crunch pattern.
- Conclusion: evidence does not support a systematic default-driven domestic credit crunch outside banking-crisis episodes.

---

### Political implications of default
- Theoretical framework and incentives:
  - Social cost immediate default = D_0; delayed default cost = D_1 (with D_1 > D_0); avoidance success probability = Π.
  - Avoidance socially optimal iff (1−Π)D_1 < D_0, i.e., Π > (D_1−D_0)/D_1.
  - Politician loses job upon default with probability θ; politician utility U = Φ W + (1−Φ) R (R = rents).
  - Politician chooses to attempt avoidance based on inequality (8) in text; implications: self-interested politicians (R>0 and Φ<1) may delay default even when socially suboptimal; extreme case Φ=0 implies postponement even if Π=0.
- Empirical evidence on political costs (Table 10, Table 11, Table 12 summaries):
  - Elections (Table 10): among 19 democracies with electoral data before/after defaults (1980–2003), ruling coalitions lost votes in 18 cases (exception: Ukraine); average decrease in electoral support = 16 percentage points.
  - Change in chief executive: in 50 percent of cases (11 out of 22 episodes) there was a change in the chief of the executive in the year of default or the following year (more than twice the normal-time probability reported by Frankel (2005)).
  - Minister turnover (Table 11):
    - One year later: tranquil years = 19.40; after a default = 25.70; difference = 6.40; P value = 0.04.
    - 18 months later: tranquil years = 47.30; after a default = 57.70; difference = 10.40; P value = 0.01.
    - Bond defaults: probability of turnover more than doubles to over 40 percent in one-year window; using 18-month window bond defaults show more than 90 percent of finance ministers losing their job (tranquil-time turnover = 47 percent).
  - Regime split (Table 12): political cost of defaulting on bank loans higher in dictatorships; cost of defaulting on sovereign bonds higher in democracies; pooled results show higher turnover of economic policymakers in dictatorships.
- Interpretation: defaults are associated with large political costs for incumbents and top economic officials; bond defaults particularly perilous for finance ministers.

---

### Policy implications and interpretation
- Default point and debt safety:
  - Accurately identifying channels and magnitudes of default costs necessary to determine a sovereign’s “default point” and how “safe” a given debt level is.
- Market functioning and borrowing costs:
  - If default costs operate largely through international trade, more open economies would have a higher default point, be less risky for lenders, and face lower borrowing costs (other things equal).
- Domestic policy:
  - Because defaults can trigger banking crises, pay attention to bank exposures to government debt and crisis-resolution mechanisms to mitigate domestic costs of sovereign default.
- Political economy:
  - High political costs of default influence policymakers’ incentives and timing of restructurings; may induce delay or “gambles for redemption,” sometimes at social cost.

---

### Conclusions and open questions
- Robust findings:
  - Default costs observable but generally short lived — effects rarely detectable beyond one or two years.
  - Sharp increase in government turnovers and ministerial replacements following debt crises.
- Key patterns:
  - Reputation measures (credit ratings and spreads) worsen on impact but tend to normalize over time.
  - Trade and trade credit decline following defaults, but trade-credit channel does not fully explain trade losses.
  - Defaults tend to precede banking crises; however, defaults do not systematically produce industry-level credit crunches outside banking crises.
- Caveats:
  - Identifying causal effects remains challenging due to confounding shocks and endogeneity.
  - Some results (e.g., absence of bank-lending effects, role of trade credit) warrant caution and further investigation.
- Suggested avenues for research:
  - Deeper study of policymakers’ decision-making and timing of defaults.
  - Further quantification of default-cost channels and investigation of mechanisms that mitigate domestic consequences (bank exposure limits, crisis-resolution frameworks).

*Source: _wp08238 - 1.   Default Episodes (IMF working paper excerpt).*

### 1.   Default Episodes ..................................................................................................

### Default Episodes

### Introduction and objectives
- Purpose: Evaluate empirically the suspected mechanisms through which default costs may affect a sovereign government, focusing on four types of cost: reputational costs, international trade exclusion costs, costs to the domestic economy through the financial system, and political costs to the authorities.
- Scope: Analyze incidence of those cost channels and assess whether there is empirical basis for—or evidence against—each mechanism rather than precisely quantifying the cost of default.
- Key methodological caution: It is difficult to isolate the causal costs of default from other factors that cause both debt default and economic downturns; direction of causality between growth and default is hard to identify.
- Policy relevance: Identifying channels and magnitudes of default costs informs the sovereign “default point” concept (the point at which the cost of servicing debt in full contractual terms exceeds the comprehensive costs of restructuring) and affects the computation of default probabilities and sovereign bond pricing.

### Conceptual framework and background
- Rationale for sovereign debt existence: Presence of costly sovereign defaults is the mechanism that makes sovereign debt possible (Dooley, 2000).
- Distinction from private debt: Creditor legal remedies are weaker for sovereigns; many sovereign assets are immune from legal action and enforcement of court judgments is often impossible.
- Definition used: “Default” encompasses any situation in which the sovereign does not honor the original terms of the debt contract, including voluntary restructurings where creditors lose value; aligned with credit rating agencies’ concept.
- Traditional mechanisms emphasized in literature:
  - Reputational costs (canonical Eaton and Gersovitz (1981) model).
  - Direct sanctions such as trade embargoes (Bulow and Rogoff (1989) and related “sanctions view” literature).
- Recent emphases:
  - Domestic costs via the banking sector because banks in many emerging economies hold significant government bonds; sovereign default can weaken bank balance sheets and precipitate bank runs.
  - Political costs: domestic economic decline and banking crises can impose political costs on incumbent parties and policymaking authorities.

### Empirical strategy and limitations
- Objective: Evaluate empirical basis for each mechanism rather than provide precise cost estimates.
- Identification challenges highlighted:
  - Negative correlation between default and growth may reflect common causes rather than a causal effect of default.
  - Difficulty in testing causality and isolating the effect of default episodes from contemporaneous shocks.
- Approach: Examine relationships between default and GDP growth, borrowing costs (credit ratings and interest rate spreads), international trade and trade credit, banking crises, industry-specific effects, and political outcomes.

### Main empirical findings
- Reputational and borrowing-cost effects:
  - Reputational costs, as reflected in credit ratings and interest rate spreads, are significant but appear to be short-lived.
- Trade and trade credit effects:
  - Evidence that trade and trade credit are negatively affected by default.
  - Controlling for trade credit does not seem to modify the effect of default on trade.
- Growth and macroeconomic effects:
  - Growth in the domestic economy suffers following defaults, and more so in cases where the causes for default seem less compelling.
  - The negative growth effect also seems to be short-lived.
- Banking sector and financial system:
  - Default episodes seem to cause banking crises and not vice versa.
  - Outside of banking crisis episodes, more credit-dependent industries do not suffer more than other industries following a sovereign default.
- Political consequences:
  - Political consequences of a debt crisis are dire for incumbent governments and finance ministers, broadly in line with what happens in currency crises.

### Policy implications and interpretation
- Default point and debt safety: Accurately identifying the channels and magnitudes of default costs is necessary to determine a sovereign’s “default point” and to assess how “safe” a given level of debt is.
- Market functioning and borrowing costs:
  - If costs of default operate largely through international trade, more open economies would have a higher default point, be less risky for lenders, and face lower borrowing costs (other things equal).
- Domestic policy implications:
  - Because defaults can trigger banking crises, attention to bank exposures to government debt and mechanisms for crisis resolution are central to mitigating domestic costs of sovereign default.
- Political economy:
  - Awareness that default episodes carry severe political costs for incumbents may affect policy choices and the timing of restructurings.

*Source: _wp08238 - 1.   Default Episodes (IMF working paper excerpt).*

### Section VIII we conclude.

### _wp08238 - Section VIII we conclude.

### II. Two hundred years of sovereign default — incidence, timing, and duration
- Data sources and classification
  - Table A1 compares four sources: Standard and Poor’s (first four columns), Beim and Calomiris (2000) (columns 5–6), Sturzenegger and Zettelmeyer (2006) (column 7), and Detragiache and Spilimbergo (2001) (last column).
  - Beim and Calomiris (2000) tend to merge defaults that occur within five years into single episodes, producing fewer but longer episodes.
  - Detragiache and Spilimbergo (2001) code as defaults several episodes not classified as defaults by Standard and Poor’s.
  - The paper uses Standard and Poor’s classifications (first four columns of Table A1) for the remainder of the analysis due to relative completeness.

- Geographic distribution and counts (1824–2004)
  - Latin America: 126 default episodes.
  - Africa: 63 default episodes.
  - Asia: lowest number of defaults among developing regions.
  - Historical context: Latin America’s lead partly reflects earlier independence and earlier access to international financial markets compared with Africa.

- Default waves and episode lengths by period
  - 1824–1840: 19 default episodes; 14 involved Latin America; other 5 involved Greece, Portugal, and Spain (three episodes). Average length: more than twenty years.
  - 1841–1860: 6 default episodes (relatively tranquil), followed by a lending boom.
  - 1861–1920: 58 default episodes; 41 in Latin America and 8 in Africa. Average length dropped to less than five years by 1881–1920.
  - 1921–1940: 39 default episodes; more than half in Latin America and 16 in Europe. Last period with Western European defaults.
  - 1941–1970: very few defaults (six episodes in total). Of these six: Hungary in 1941, Japan in 1942, Czechoslovakia in 1959, Cuba in 1960, Costa Rica (1962), Zimbabwe (1965).
  - 1970s lending boom: 15 episodes of defaults on syndicated bank loans occurred in the 1970s.
  - 1982 debt crisis (post-Mexico August 1982): more than 70 default episodes (34 in Africa, 29 in Latin America). Average default lasted approximately 9 years during this crisis era.
  - Restructuring via Brady Bonds followed syndicated bank loan restructurings and helped create a bond market for emerging-market debt.
  - 1991–2004: 40 defaults (14 on bonds and 26 on syndicated bank loans). Syndicated bank loan defaults mainly in Africa; bond defaults mainly in Latin America.

### III. Default and GDP growth — empirical findings and decomposition
- Regression framework
  - Unbalanced panel up to 83 countries for 1972–2000.
  - Model (1): GROWTH_it = α + β DEFAULT_it + γ X_it + ε_it, where GROWTH_it is per capita annual real GDP growth; X is a matrix of controls (see list below); DEFAULT is a set of dummy variables tracking default episodes.
  - Controls include: INV_GDP, POP_GR, GDP_PC70s, SEC_ED, POP, GOV_C1, CIV_RIGHT, DTOT, OPEN, BK_CR, and regional dummies for SSA, LAC, TRANS. Substituting country fixed effects for regional dummies does not change results.

- Main growth effects
  - Column 1 (baseline): default associated with a decrease in growth of 1.2 percentage points per year.
  - Column 2 (dynamic): DEF_B (beginning of default episode) and three lags (DEF_B1, DEF_B2, DEF_B3) show a large impact in the first year: drop in growth of 2.6 percentage points; no statistically significant effects of the lagged default variables.
  - Columns 3–4: augment with END_DEF and two lags (exit from default); END_DEF dummies not statistically significant and do not affect estimated effect of default.

- Causality and decomposition strategy
  - Concern: defaults often follow shocks (terms of trade, sudden stops, currency crises) that also depress growth, so correlation may not be causal.
  - Decomposition of DEFAULT into anticipated and unanticipated components:
    - default_pred_it + v_it = default_it
    - default_pred_it: predicted probability of default from a logit (model similar to Manasse et al. (2003); full results in Table A2).
    - v_it: residual (unanticipated) component.
  - Interpretation: default_pred_it captures effects associated with increased probability of default (anticipated), while v_it captures additional effects from the act of default itself (unanticipated).

- Results from Table 3 (restricted sample of developing countries)
  - Sample: Table 3 sample has 843 observations versus 2,048 observations in Table 2.
  - Column 1 (restricted-sample baseline): effect of default is slightly smaller than before but at 1 percent still sizable and highly statistically significant.
  - Column 2 (anticipated vs unanticipated):
    - DEF_PR (anticipated effect): estimate 1.4 percent (negative impact on growth).
    - DEF_U (unanticipated effect): close to 1.0 percent (negative).
    - Both anticipated and unanticipated components are statistically significant.
  - Column 3 (dynamic split):
    - DEF_PRB (anticipated on impact): negative, quite large, and statistically significant (investigated for outliers without finding evidence).
    - DEF_U shows a large and statistically significant overall effect but no significant negative effect in the first year.
  - Column 4: augmenting column 2 with END_DEF and two lags does not change results.

- Interpretation and alternative readings
  - The unanticipated component may capture costs of “unjustified” defaults (weak willingness to pay versus inability), consistent with literature distinguishing “ability” and “willingness” to pay (Grossman and van Huyck, 1988).
  - Identifying “avoidable” defaults directly is difficult; many unilateral repudiations stem from political regime changes where downturns may be endogenous to political shifts.

### IV. Default and reputation — access, costs, and market discrimination
- Access to international capital markets
  - Evidence indicates default does not lead to permanent exclusion from international capital markets.
  - Countries lose access during default, but once restructuring is concluded, markets do not systematically discriminate in terms of access between defaulters and non-defaulters.
  - External factors and investor mood are often more important determinants of market access than default history.
  - Example: 1930s–1960s exclusion from world capital markets affected both defaulters and non-defaulters among Latin American countries.
  - Gelos et al. (2004): countries that defaulted in the 1980s regained access to international credit in about 4 years.

- Reputation and borrowing costs
  - Empirical work examines indirect and direct measures of market discrimination via credit ratings and spreads.
  - Cantor and Packer (1996): small set of explanatory variables explains more than 90 percent of variance in credit ratings; dummy for countries that defaulted after 1970 associated with a drop of two notches in credit rating.
  - Reinhart et al. (2003): default history associated with lower Institutional Investor ratings.

- Cross-country rating model (equation (3))
  - RATING_i = α + β DEFAULT_i + γ X_i + ε_i, where RATING measures average credit ratings over 1999–2002; X measured over 1999–2002.
  - Ratings: Standard and Poor’s foreign-currency long-term ratings converted to numerical values (20 = AAA, 19 = AA+, 18 = AA, ... SD = 0).
  - Explanatory variables include: LGDP_PC, GDPGR, LINF, CG_BAL, CA_BAL, EXDEXP, and IND (industrial country dummy).
  - Column 1 of Table 4 (1970–2002 default dummy): regression R^2 = 0.91 (explains more than 90 percent of cross-country variance in credit ratings). Default history is negatively correlated with credit ratings; point estimates indicate default history leads to a drop in credit rating (exact magnitude reported in Table 4).

*Source: _wp08238 - Section VIII we conclude.*

### 1.7 notches, slightly lower than the estimate of Cantor and Packer (2.5 notches).

### _wp08238 - 1.7 notches, slightly lower than the estimate of Cantor and Packer (2.5 notches).

### Ratings regressions and robustness checks
- Estimated rating effect cited: "1.7 notches, slightly lower than the estimate of Cantor and Packer (2.5 notches)."
- Additional control variables added (column 2):
  - DEBT_GDP (public debt over GDP)
  - OR_SIN (index of original sin, Eichengreen et al. (2005))
  - Both controls have the right sign and are statistically significant; 16 observations lost; results essentially unchanged.
- Column 3 augmentation:
  - Adds SDTOT (standard deviation of the terms of trade) over the period 1991–2002.
  - SDTOT has the right sign but is not statistically significant; other results unchanged.
- Column 4 specification:
  - Replaces a single default dummy with seven dummy variables tracking default history (DEF1800, DEF1900_50, DEF1950_70, etc.).
- Additional robustness notes:
  - Model estimated using average ratings for the 2000-2004 period and explanatory variables averaged over the 1990-2000 period; results did not change.
  - Using external debt over GDP (data from World Bank’s GDF) yields identical results. For industrial countries EXDEXP set to zero; EXDEXP can be thought of as EE*(1-IND).
  - Countries in default over the entire 1999–2004 period were dropped in estimations; results robust to keeping them.

- Key interpretive finding:
  - "The results indicate that defaults episodes do not have a long-term impact on credit ratings. In fact, only defaults in the 1995–2002 period are significantly correlated with credit ratings over the 1999–2002 period."

*Italicized source attribution line.*

### 0.8 to -0.13. Column 6 reproduces the model of column 4 adding the lagged dependent variable

### _wp08238 - 0.8 to -0.13. Column 6 reproduces the model of column 4 adding the lagged dependent variable

### Trade credit, defaults, and bilateral trade
- Column 6 (Arellano and Bond (1991) estimator, with lagged dependent variable) finds the effect of default on trade credit is negative and large only in the first and second year of the default; the negative effect is short lived.
- Gravity model estimated (equation (5)):
  - Dependent variable: LTR_{i j t}, log of bilateral trade between country i and country j at time t.
  - Key regressors:
    - NSDEF_{i j t}, dummy = 1 if in year t either country i or country j is in default (measured using Standard and Poor’s data) and the i–j pair consists of a developing and industrial country.
    - NSTC_{i j t}, set equal to the log of the stock of official trade credit received by the developing country in the pair in year t; takes value 0 if the pair is two industrial or two developing countries.
  - Controls: country-pair fixed effect μ_{i j} and X_{i j t} (same set used by Rose (2005) in fixed effect regressions — log of total GDP, log of GDP per capita, regional trade agreement dummy, colony dummy, currency union dummy — augmented with default interacted with average trade).
- Empirical results (Table 7 summary across columns):
  - Column 1 reproduces Rose (2005): defaults associated with a large and statistically significant decline in bilateral trade flows between advanced and emerging/developing economies.
  - Column 2 adds DEF_AVT (default dummy interacted with log of average trade): DEF_AVT positive and statistically significant; including it increases point estimates of DEF_NS.
  - Column 3: same as column 2 but sample restricted to observations with trade credit data; DEF_NS remains negative and statistically significant but impact quantitatively smaller in restricted sample.
  - Column 4 augments with NSTC_{i j t} (log total stock of trade credit to the developing country in year t):
    - NSTC positive and statistically significant.
    - Estimated elasticity of trade to trade credit approximately 7 percent.
    - Controlling for trade credit does not affect the relationship between default and trade: coefficients of DEF_NS and DEF_AVT in column 4 are identical to column 3.
  - Columns 5 and 6 repeat the experiment focusing on total non-bank trade credit and total bank trade credit, respectively; results essentially unchanged.
- Interpretation: although trade credit correlates positively with trade, controlling for trade credit does not mediate the negative correlation between default and bilateral trade; effect of default on trade and trade credit tends to be short lived.

### Default and the domestic banking system
- Data and sample:
  - Banking-crisis index constructed using Glick and Hutchinson (2001), Caprio and Klingebiel (2003), Dell’Ariccia et al. (2005).
  - Sample: 149 countries, 1975–2000 period, total of 3,874 observations.
  - Observations:
    - 111 banking crises (unconditional probability = 2.9 percent).
    - 85 default episodes (unconditional probability = 2.2 percent).
- Conditional probabilities:
  - Probability of a banking crisis in year t conditional on debt default in year t or year t-1 = 14 percent.
  - This represents an 11 percentage point increase relative to unconditional probability (2.9 percent).
  - Difference between conditional and unconditional probability is statistically significant.
  - Probability of default conditional on having a banking crisis is only two percentage points higher than unconditional probability; difference not statistically significant at conventional confidence levels.
- Interpretation: default episodes may increase the probability of a banking crisis more than banking crises increase the probability of default; caution due to few “twin” crisis cases and annual data timing limitations.
- Credit-crunch test (industry-level, Rajan and Zingales (1998) methodology, specification (6)):
  - Dependent variable: real value added growth for industry j in country i at time t.
  - Controls: country-industry fixed effects a_{i j}, country-year fixed effects b_{i t}, industry-year fixed effects c_{j t}, lagged sector share SHVA to control for convergence/mean reversion.
  - Variable of interest: DEF × EXT (interaction between default dummy and index of external financial dependence).
  - Interpretation of β: negative β would indicate sectors requiring more external finance are disproportionately affected by defaults (evidence of credit crunch).
- Results (Table 9 summary):
  - Column 1 (all years in default): coefficient on DEF positive (wrong sign) and not statistically significant.
  - Column 2 (three dummies for first, second, third year of default): coefficients tend to be negative but never statistically significant (neither individually nor jointly).
  - Columns 3 and 4 augment with interaction between banking crisis and external dependence (as in Dell’Ariccia et al., 2005): results unchanged.
- Conclusion: evidence does not support the credit-crunch hypothesis — defaults do not appear to have a special negative effect on industries that depend more on external finance.

### Political implications of default
- Political costs and policymaker incentives:
  - High political costs of default can increase willingness to pay and sustainable debt, but can also induce “gambles for redemption” and prolong costly delays.
  - Delaying default can be costly for at least three reasons:
    - (i) Non-credible restrictive fiscal policies are ineffective and lead to output contractions.
    - (ii) Delayed defaults may prolong uncertainty and high interest rates, negatively affecting investment and banks’ balance sheets.
    - (iii) Delayed default may have direct harmful effects on the financial sector.
- Simple formal framework (notation from text):
  - Social cost of immediate default = D_0.
  - If avoidance measures succeed with probability Π, no future default; if fail, delayed default cost = D_1 (with D_1 > D_0).
  - Trying to avoid default is socially optimal iff (1−Π)D_1 < D_0, equivalently Π > (D_1−D_0)/D_1.
  - Politician loses job upon default with probability θ. Politician’s utility U = Φ W + (1−Φ) R, with R rents, W social welfare, Φ in [0,1].
  - Politician chooses to attempt avoidance if inequality (8) in text holds (policy condition depends on R, Φ, θ, Π, D_0, D_1).
  - Implication: politicians with R>0 and Φ<1 (self-interested) will try to delay default even when socially suboptimal; extreme case Φ=0 implies postponement even if Π=0.
- Empirical evidence on political costs:
  - Table 10 (democracies that defaulted, 1980–2003): out of 19 countries with electoral data before/after defaults, ruling coalitions lost votes in 18 cases (exception: Ukraine).
  - Average decrease in electoral support for ruling governments after default = 16 percentage points.
  - In 50 percent of cases (11 out of 22 episodes) there was a change in the chief of the executive in the year of default or the following year (more than twice probability in normal times reported by Frankel (2005)).
  - Table 11 (IMF governor/finance minister turnover):
    - In tranquil years probability of change of IMF governor = 19.4 percent; after a default probability = 26 percent (difference p-value = 0.04).
    - Bond defaults: probability of turnover more than doubles to over 40 percent.
    - Using an 18-month window: bond defaults show more than 90 percent of finance ministers losing their job in 18 months following default (tranquil-time turnover = 47 percent).
  - Table 12 (regime split: dictatorships vs democracies):
    - Political cost of defaulting on bank loans higher in dictatorships.
    - Cost of defaulting on sovereign bonds higher in democracies.
    - For all defaults pooled, higher turnover of economic policymakers observed in dictatorships.
    - 18-month window results broadly similar.
- Interpretation: defaults are associated with large political costs for incumbents and top economic officials; bond defaults particularly perilous for finance ministers.

### Conclusions and broader takeaways
- Overall findings:
  - Default costs are significant but short lived; effects rarely detectable beyond one or two years.
  - Reputation measures (credit ratings and spreads) are tainted by default but only for a short time.
  - Some evidence that international trade and trade credit are negatively affected by default episodes, but unable to trace this to the volume of trade credit.
  - Debt defaults appear to cause banking crises (default → banking crisis), but weak evidence for default-driven credit crunches in domestic markets.
  - Defaults significantly shorten the tenure of governments and economic officials.
- Robust/striking findings:
  - The short-lived nature of effects — “we almost never can detect effects beyond one or two years.”
  - Sharp increase in government turnovers following debt crises.
- Caveats and open questions:
  - Default costs remain vaguely defined and difficult to quantify in some dimensions.
  - Results on how international trade credit affects the link between trade and default, and the finding that default episodes do not seem to affect bank lending, appear less plausible and warrant caution.
  - Decision-making process and timing of defaults by policymakers is relatively unexplored; possibilities include strategic delay to avoid reputational costs (Grossman and Van Huyck (1988) model) or self-interested postponement to protect political careers.
- Suggested avenues:
  - Further investigation of policymaker decision-making regarding timing of default and the political economy determinants of postponement.

*Italic source attribution: Content reproduced from the supplied IMF PDF chapter/section.*

### REFERENCES

### _wp08238 - REFERENCES

### Major literature themes cited
- Sovereign default theory and empirical analyses
  - Ades et al., 2000; Arellano and Bond, 1991; Bulow and Rogoff, 1989; Eaton and Gersovitz, 1981; Grossman and Van Huyck, 1988; Krugman, 1988; Merton, 1974; Tomz, 2007.
- Costs and consequences of defaults and crises
  - De Paoli, Hoggarth, and Saporta, 2006; Dell'Ariccia, Detragiache, and Rajan, 2005; Díaz-Alejandro, 1983; Frankel, 2005; Reinhart, Rogoff, and Savastano, 2003; Sturzenegger and Zettelmeyer, 2006.
- Banking crises, twin crises, and crisis propagation
  - Caprio and Klingebiel, 2003; Glick and Hutchinson, 2001; Kaminsky and Reinhart, 1999; Detragiache and Spilimbergo, 2001.
- Sovereign risk measurement and bond markets
  - Cantor and Packer, 1996; Gapen et al., 2005; Kealhofer, 2003; Gelos, Sahay, and Sandleris, 2004; Love, Preve, and Sarria-Allende, 2005.
- Historical studies and datasets on sovereign lending/defaults
  - Beim and Calomiris, various; Standard & Poor’s historical default dataset; Sturzenegger & Zettelmeyer, 2006; World Bank, Global Development Finance 2004.

### Key empirical tables and exact results (selected)
- Table 2. Default and Growth, Panel 1972–2000 (columns (1)-(4)):
  - INV_GDP: 1.211; 1.152; 1.205; 1.146 (t-statistics: (8.63)***, (8.08)***, (8.58)***, (8.04)***)
  - POP_GR: -0.120; -0.119; -0.121; -0.118 (t-statistics: (1.22), (1.22), (1.24), (1.20))
  - GDP_PC70s: -0.121; -0.124; -0.121; -0.125 (t-statistics: (7.25)***, (7.34)***, (7.24)***, (7.37)***)
  - GOV_C1: 2.965; 2.974; 2.970; 3.000 (t-statistics: (2.91)***, (2.89)***, (2.89)***, (2.89)***)
  - BK_CR: -1.087; -1.068; -1.092; -1.080 (t-statistics: (4.64)***, (4.53)***, (4.65)***, (4.57)***)
  - DEF: -1.239; -1.184; -1.282; -1.370 (t-statistics: (4.32)***, (3.82)***, (4.38)***, (4.06)***)
  - Observations: 2048; 1985; 2048; 1985
  - R-squared: 0.22; 0.22; 0.22; 0.22

- Table 3. Default and Growth, Panel 1972–2000 (alternative specification):
  - INV_GDP: 1.607; 1.584; 1.635; 1.584 (t-statistics: (5.11)***, (5.00)***, (4.58)***, (5.03)***)
  - POP: 0.006; 0.006; 0.005; 0.006 (t-statistics: (5.36)***, (5.29)***, (4.12)***, (5.25)***)
  - GOV_C1: 3.402; 3.281; 3.084; 3.299 (t-statistics: (2.95)***, (2.76)***, (2.45)**, (2.75)***)
  - BK_CR: -1.364; -1.328; -1.188; -1.324 (t-statistics: (3.81)***, (3.73)***, (3.36)***, (3.71)***)
  - Observations: 843; 843; 726; 843
  - R-squared: 0.26; 0.26; 0.28; 0.26

- Table 4. Default and Credit Ratings, Cross Section Regression, 1999–2002:
  - LGDP_PC: 1.627; 1.418; 1.215; 1.366 (t-statistics: (4.69)***, (3.83)***, (3.20)***, (3.47)***)
  - LINF: -0.707; -0.817; -0.727; -0.932 (t-statistics: (3.48)***, (3.88)***, (3.04)***, (3.65)***)
  - EXDEXPGDF: -0.834; -0.776; -0.750; -0.761 (t-statistics: (2.67)***, (3.05)***, (2.03)**, (2.13)**)
  - DEFAULT: -1.669; -1.486; -1.855 (t-statistics: (3.10)***, (2.86)***, (3.57)***)
  - DEBT_GDP: -0.022; -0.020; -0.020 (t-statistics: (2.99)***, (2.16)**, (2.73)***)
  - Observations: 68; 595; 568
  - R-squared: 0.91; 0.94; 0.95; 0.92

- Table 5. Defaults and Bond Spreads, Panel Regression, 1997–2004 (selected coefficients):
  - LGDP_PC: -200.578; -1424.802; -218.969; -1237.708; -216.274; -1663.319; -47.260; -1172.255 (various columns; many significant with ( )***)
  - LINF: 46.061; 25.281; 55.359; 31.052; 54.589; 33.042; 36.787; 26.325
  - EXDEXPGDF: 166.770; 207.386; 169.660; 213.708; 192.966; 246.435; 96.262; 189.341 (all columns significant at *** or **)
  - DEF1YEAR: 412.863; 307.746; 433.912; 305.783; 389.342; 249.764; 267.770; 249.175 (many significant at *** or **)
  - DEF2YRS: 246.746; 188.244; 267.262; 162.114; 238.877; 145.339; 134.276; 144.640
  - Observations by column: 150; 150; 162; 162; 144; 144; 144; 144
  - R-squared examples: 0.56; 0.53; 0.58; 0.58

- Table 6. Default and Trade Credit:
  - DEFAULT: -0.800; -0.800; -0.800; -0.900; -0.134; 0.011 (t-statistics: (5.85)***, (5.74)***, (5.95)***, (5.85)***, (4.88)***, (0.39))
  - DRER: -0.100; -0.100; -0.300; -0.300; -0.266; -0.259 (t-statistics: (1.74)*, (1.73)*, (2.50)**, (2.50)**, (4.78)***, (4.69)***)
  - Observations: 1060; 1060; 1059; 1059; 872; 872
  - R-squared: 0.07; 0.07; 0.07; 0.07

- Table 7. Default and Trade: Does Trade Credit Matter?
  - DEF_NS: -0.206; -0.319; -0.054; -0.054; -0.047; -0.104 (t-statistics: (16.46)***, (25.21)***, (1.68)*, (1.66)*, (1.47), (3.00)***)
  - LGDP: 0.315; 0.353; 0.393; 0.393; 0.393; 0.392 (t-statistics: (40.18)***, (45.03)***, (38.75)***, (38.76)***, (38.73)***, (38.39)***)
  - LTC_TOTNS, LTC_NBNKNS, LTC_BNKNS individual coefficients reported in columns; when included they are significant at *** levels.
  - Observations across columns: 234457; 234457; 151371; 151371; 151243; 147057
  - R-squared range: 0.08 to 0.12

- Table 8. Probabilities of Default and Banking Crisis (exact probabilities):
  - Unconditional probability of a banking crisis (111 episodes): 2.9
  - Probability of a banking crisis conditional on a default: 14.1
  - P value on a test P(BC/DEF)>P(BC): 0.0
  - Unconditional probability of a sovereign default (85 episodes): 2.2
  - Probability of a default conditional on a banking crisis: 4.5
  - P value on a test P(DEF/BC)>P(DEF): 0.1

- Table 9. Default and Industry Value-Added Growth:
  - SHVA: -1.251; -1.25; -1.253; -1.252; -1.253 (t-statistics: (15.18)***, (15.17)***, (15.21)***, (15.19)***, (15.21)***)
  - BK_CR*EXT: -2.277; -2.164; -2.282 (t-statistics: (2.29)**, (2.16)**, (2.29)**)
  - Observations: 15872
  - R-squared: 0.46

- Table 10. Defaults and Elections (selected country entries, exact values preserved):
  - Argentina: Year of default 2001; Year Votes 1999 37.50; 2003 16.90; Change in votes -20.60; Change in the chief of the executive: YES
  - Bolivia: Year of default 1989; Year Votes 1985 26.42; 1989 19.64; Change -6.78; YES
  - Costa Rica: 1981 default; votes 1978 39.66; 1982 25.79; Change -13.87; YES
  - Table lists multiple further country-year vote changes and indicates where chief executive changed.

- Table 11. Default and the Probability of Replacing the Minister of Finance (exact percentages and p values):
  - One year later, Tranquil years: 19.40; After a default: 25.70; Difference: 6.40; P value: 0.04
  - 18 months later, Tranquil years: 47.30; After a default: 57.70; Difference: 10.40; P value: 0.01
  - Subsamples by default type (International Bank Loans, Sovereign Bonds) and period (1977-1989, 1990-2004) reported with exact values and p values.

- Table 12. Default and the Probability of Replacing the Minister of Finance by Type of Default and Government (exact percentages and p values):
  - Example: Democracies, One year later, Tranquil years: 21.90; After a default: 23.30; Difference: 1.40; P value: 0.79
  - Democracies, 18 months later, Tranquil years: 51.10; After a default: 53.50; Difference: 2.40; P value: 0.69
  - Dictatorships show different magnitudes; exact cell values and p values are reported in the table.

- Appendix — Table A1. Private Lending to Sovereign. Default and Rescheduling:
  - Extensive country-by-country listings across Standard & Poor’s (1824-2004), Beim & Calomiris (1800-1992), Sturzenegger & Zettelmeyer (1874-2003), and Detragiache & Spilimbergo (1973-1991). Examples of exact year ranges are reported for many countries (e.g., Egypt 1816–1880 under Beim & Calomiris; Liberia multiple episodes including 1875–1898; a wide set of LAC, Africa, Asia, and Europe entries).
  - Notes preserve dataset definitions, e.g., S&P default definition, Beim & Calomiris inclusion rules, Sturzenegger and Zettelmeyer federal/central government focus, Detragiache & Spilimbergo debt-crisis classification criteria (including arrears > 5 percent of total commercial debt outstanding).

- Table A2. Logit Model for the Probability of Default (selected coefficients and fit):
  - US real treasury bill rate_1: 0.184 (t-statistic (2.13)) in DEF; 0.127 (t-statistic (1.29)) in DEF_B
  - Real GDP growth_1: -0.063 (t-statistic (3.91)) in DEF; -0.066 (t-statistic (3.36)) in DEF_B
  - Volatility on inflation_1: 0.093 (t-statistic (3.16)) in DEF; 0.017 (t-statistic (0.76)) in DEF_B
  - Concessional debt to total debt_1 x Dummy 90s: 6.343 (t-statistic (2.09)) in DEF; 4.413 (t-statistic (1.39)) in DEF_B
  - Observations: 1416; R-square: 0.313 (DEF) and 0.147 (DEF_B)

### Data and methodological notes captured in the appendix
- Standard & Poor’s definition of sovereign default: failure to meet principal or interest on due date (or within specified grace period); rescheduling or exchange offers with less favorable terms count as default.
- Beim & Calomiris dataset: includes private lending through bonds, supplier’s credits, or bank loans; defines extended period of reduced/rescheduled payments (six months or more); combines periods within five years.
- Sturzenegger & Zettelmeyer: focuses on federal/central government defaults; excludes some defaults tied solely to wars/revolutions except when coincident with clusters; lists year of initial default/rescheduling in sequences.
- Detragiache & Spilimbergo: classify a debt crisis if (i) arrears > 5 percent of total commercial debt outstanding, or (ii) rescheduling/restructuring agreement with commercial creditors as listed in the GDF.

*Source: _wp08238 - REFERENCES*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08238.pdf_