## wpiea2019069 — References and Appendices (selected findings)

## Source details

**Canonical URL:** [wpiea2019069 — References and Appendices (selected findings)](https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019069.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2019/wpiea2019069.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2019/wpiea2019069.pdf.json)

---

### Major findings on restructuring impacts and transmission channels
- Sample and coverage:
  - Sample of 70 countries with at least one sovereign debt restructuring over 1978–2010.
  - 179 restructurings of privately-held external sovereign debt.
  - 39 banking crises.
  - 48 severe credit declines.
  - 40 severe protracted declines in net capital inflow.
- Overall impacts by restructuring strategy (local projections, AIPW):
  - Post-default restructurings are associated with the most severe and protracted declines in GDP and investment (in percent of GDP), with cumulative contractions of 4 and 1 percentage points over the first 3 years, respectively.
  - Post-default restructurings are associated with severe and prolonged declines in bank credit and sharp increases in lending interest rates.
  - Pre-emptive restructuring events (weakly and strictly) have much milder and short-lived declines in GDP, investment, private credit, and net capital inflows.
  - Weakly preemptive restructuring effects lie between post-default and strictly preemptive effects.
- Role of capital inflow–credit channel and sudden stops:
  - Severe sudden stops (severe protracted declines in net capital inflows) are strongly associated with credit crunches and occur more likely and more severely after post-default restructurings.
  - Severe sudden stops and credit crunches are associated with the worst contractions in investment (in percent of GDP) and GDP, with cumulative declines of 4 and 5 percentage points of GDP over 2 years after a severe sudden stop occurs.
  - The capital inflow–credit channel influences GDP directly and indirectly through private credit and investment and their feedback effects, amplifying the impact of post-default restructurings.
- Banking sector linkage:
  - Logit estimates show countries undergoing post-default restructurings are significantly more likely to experience a banking crisis than those following a preemptive restructuring.
  - The deterioration of the banking sector appears central to the larger output costs observed after post-default restructurings.
- Heterogeneity by negotiation characteristics:
  - Output costs of post-default episodes with short duration or small haircuts are lower than those with long duration or high haircuts, but still higher than for weakly preemptive restructurings—suggesting avoiding missed payments matters more than creditor-friendly terms alone.

### Definitions and classification
- Restructuring strategy definitions (Asonuma and Trebesch (2016)):
  - Strictly preemptive restructurings: implemented without missing any payments (no legal default).
  - Weakly preemptive restructurings: some payments are missed, but only temporarily and after the start of formal or informal negotiations with creditor representatives (no unilateral default).
  - Post-default restructurings: payments are missed unilaterally and without the agreement of creditor representatives (unilateral default prior to negotiations).
- Severe declines definitions (3-year horizon, 25th percentile thresholds):
  - Severe credit declines: cumulative percentage point difference in the credit-to-GDP ratio over three years is less than the 25th percentile of the distribution.
  - Severe protracted net capital inflow declines (sum): sum of cumulative percentage point differences in net capital inflows-to-GDP ratio over three years is less than the 25th percentile of the distribution.
  - Severe protracted net capital inflow declines (difference): cumulative percentage point difference in net capital inflows-to-GDP ratio over three years is less than the 25th percentile of the distribution.
- Banking crisis (Laeven and Valencia (2013)): requires significant signs of financial distress in the banking system and significant intervention measures in response.
- Severe sudden stop (definition vi): percentage point difference in net capital inflows-to-GDP ratio from year 0 to year 1 is less than the 25th or 50th percentile among observations experiencing that type of restructuring.

### Key dataset statistics (Table 1 and Appendix A.4 summaries)
- Panel A: Debt Restructuring Sample
  - # of episodes: Post-default 111; Weakly preemptive 45; Strictly preemptive 23.
  - # of countries: Post-default 60; Weakly preemptive 26; Strictly preemptive 13.
  - Average duration (years): Post-default 5.0; Weakly preemptive 1.1; Strictly preemptive 0.9.
  - Average haircuts (percent): Post-default 48.0; Weakly preemptive 18.2; Strictly preemptive 18.9.
- Panel B: Banking Crisis Sample (3-year and 5-year horizons)
  - # of available episodes (all restructurings, 3-year): 127.
  - # of banking crises within 3 years: 39.
  - Conditional probability of banking crisis after restructurings (all obs., 3-year): 30.7%.
  - Conditional probabilities by strategy (3-year): Post-default 33.0%; Weakly preemptive 17.9%; Strictly preemptive 9.5%.
  - Banking crises after post-default restructurings: 30 cases out of 91 (33 percent) (5-year horizon).
- Panel C: Severe Credit Declines (3-year and 5-year horizons)
  - # of available episodes (3-year, all restructurings): 107.
  - # of severe credit declines within 3 years: 48.
  - Conditional probability (all obs., 3-year): 44.9%.
  - Conditional probabilities by strategy (3-year): Post-default 49.3%; Weakly preemptive 38.5%; Strictly preemptive 25.0%.
  - Severe credit declines in post-default restructurings: 36 cases out of 73 (around 49 percent) (3-year); 39 cases out of 73 (53.4 percent) (5-year).
- Panels D / E: Severe Protracted Net Capital Inflow Declines
  - # of available episodes (3-year, all restructurings): 113.
  - # of severe net capital inflow declines (sum): 40.
  - Conditional probability (sum) (all obs., 3-year): 35.4%.
  - Conditional probabilities by strategy (sum, 3-year): Post-default 37.7%; Weakly preemptive 34.6%; Strictly preemptive 20.0%.
  - # of severe net capital inflow declines (difference): 40; conditional probability (difference) (all obs., 3-year): 35.4%; by strategy (difference, 3-year): Post-default 32.5%; Weakly preemptive 30.8%; Strictly preemptive 10.0%.
- Representative episodes:
  - Post-default: Argentina 2001-05; Russia 1998-2000.
  - Weakly preemptive: Pakistan 1999-2010; Belize 2006-07.
  - Strictly preemptive: Uruguay 2003.
- Sample and period:
  - Sample period: 1970–2013.
  - Benchmark sample: 68 countries that experienced at least one debt restructuring; robustness checks include full country coverage.

### Stylized empirical patterns (summary)
- Output and investment:
  - GDP (real, log-deviation from H-P trend) and investment decline substantially in post-default restructurings, less severely in weakly preemptive restructurings, and are not affected in strictly preemptive cases.
  - Investment declines more than GDP in post-default and weakly preemptive events; investment is the larger margin of adjustment.
- Credit, lending rates, and capital flows:
  - Private credit: severe and prolonged decline after post-default restructurings; recovers to pre-crisis level by the third year since start of restructurings (AIPW patterns).
  - Lending interest rates: large and protracted rise after post-default restructurings; sharp temporary increase accompanied by a sharp drop in private credit, then return to pre-crisis level.
  - Net capital inflows: severe and protracted decline in post-default restructurings (remain subdued for five years); milder decline in weakly preemptive restructurings; no effect for strictly preemptive restructurings.

### Econometric approach and robustness
- Empirical methods:
  - Local projection models (Jordà (2005)) to quantify cumulative effects over horizons h = 1..5 while controlling for dynamics.
  - Baseline estimation: Augmented Inverse Probability Weighted (AIPW) estimator (Jordà et al., 2013) with a probit first stage to estimate propensity scores Pr(R) for R ∈ {PD, Weakly, Strictly}.
  - Auxiliary analyses: logit/probit models for banking crisis likelihood, subsample local projections, two-type and one-type first-stage AIPW variants, OLS local projections.
- Robustness checks include:
  - Alternative coding using S&P default data (S&P sample broadly mirrors post-default results).
  - Exclusion of post-2008 observations.
  - Inclusion/exclusion of IMF-supported programs and Paris Club (official external debt) restructurings.
  - Single external vs dual domestic-external restructurings (single external accounts for 87% of episodes).
  - Enlarged instrument sets (including years since last restructuring, LIEC, EIEC, coerciveness index) — improved ROC classification at the cost of reduced sample size.
  - Controlling for GDP forecasts (IMF WEO 1990–2013) — baseline results robust.

### Heterogeneity: duration, haircuts, and expected forecasts
- Average durations and haircuts:
  - Average duration: post-default 5.0 years; weakly preemptive 1.1 years; strictly preemptive 0.9 years.
  - Average haircuts (percent): post-default 48.0; weakly preemptive 18.2; strictly preemptive 18.9.
- Key heterogeneity findings:
  - Post-default episodes with short expected duration or small expected haircuts have smaller output costs than long-duration/high-haircut post-defaults, but still worse than weakly preemptive episodes.
  - First-stage Poisson/Negative Binomial and probit predictions show:
    - GDP growth forecasts, long-run: coefficients like -0.484*** (0.056) and -0.547*** (0.064) in some columns (duration regressions).
    - # of past post-default episodes: positive and significant predictors of longer duration and higher haircuts (e.g., 0.219*** (0.057)).
    - # of past preemptive episodes: negative and significant predictors of duration/haircuts (e.g., -0.307*** (0.090)).
  - AIPW local projection estimates: GDP declines are more severe and prolonged for post-default restructurings with low GDP growth forecasts, long expected duration, and high expected haircuts (statistically significant negative sums reported in Appendix D.2).

### Capital inflow–credit channel and severe sudden stops (quantitative summary)
- Co-occurrence counts:
  - 48 severe credit declines (1978–2010) coincide with restructurings.
  - 40 (34) severe protracted net capital inflow declines coincide with restructurings (variant counts reported).
  - Under the 5-year criterion: 49 and 35 (32) episodes identified, respectively.
- Amplification:
  - Private credit, investment, and GDP decline more severely when restructurings coincide with a severe sudden stop than when they do not.
  - AIPW and OLS local projections report substantially more negative cumulative sums for post-default with sudden stops versus without (examples: OLS Panel C Sum = -23.63*** with sudden stops vs -15.76*** without; AIPW one-type Post-default with sudden stops Sum = -24.89*** vs without Sum = -10.17***).
- Annualized regressions (Table D4):
  - Post-default, d.NCI < 25th: annualized GDP growth coefficient = -3.302*** (example OLS column).
  - Post-default, d.NCI ≥ 25th: coefficient = -1.183** (less negative).
  - Post-default, d.NCI < 25th: annualized investment change = -3.894***.
  - Differences between strategies statistically significant (e.g., difference ≈ -2.120* in one specification).

### Robustness and extensions (selected)
- Large sample including countries without restructurings (122 countries, 1970–2013, OLS):
  - Post-default Sum on GDP: -13.22***.
  - Post-default Sum on Investment: -10.30***.
  - Post-default Sum on Private credit: -11.97***.
  - Post-default Sum on Lending rates: 48.77***.
  - Post-default Sum on Net capital inflows: -14.16***.
- S&P default episodes (AIPW):
  - S&P AIPW GDP Sum (x1): -9.216***.
  - S&P AIPW GDP Sum (x2): -16.68***.
  - S&P results broadly mirror post-default restructuring patterns.
- IMF-supported programs and Paris Club involvement:
  - Declines in GDP and investment following post-default or weakly preemptive restructurings are significantly moderated by official financing or official debt treatment (AIPW/OLS panels; Figure E4).
- Single external restructurings (no domestic restructuring during external restructure):
  - Single external accounts for 156 of 179 episodes (87% share).
  - AIPW results for single external episodes similar to overall external results (post-default GDP Sum = -14.70*** in one specification).

### Policy implications and recommendations (faithful extraction)
- Quick creditor agreement matters:
  - Even if a sovereign must miss payments (default), adverse effects on GDP, investment, and banking sector can be mitigated by reaching an agreement with creditors quickly (short duration), and/or inflicting on creditors the smallest possible losses (small haircuts) needed to restore solvency.
- Preemptive restructurings:
  - Preemptive restructurings (without missing payments) are largely able to avoid or mitigate output and banking sector costs associated with restructuring.
- Constraints and caveats:
  - Sudden large shocks or multiple-equilibria "bad equilibrium" contexts may force sovereigns to stop servicing debt; nonetheless, quick suspension agreed with creditors is associated with better outcomes than unilateral suspension.
  - Haircut design caveat: the haircut should be large enough to restore solvency; offering a small haircut that requires another restructuring is counter-productive and can lead to even higher costs.
- Official financing implications:
  - Where feasible, official financing and creditor cooperation that allow countries to restructure without missing payments can attenuate long-run costs; delaying adjustment and requests for official support until default becomes inevitable imposes costs.

### Selected exact numeric highlights (preserve reporting)
- Counts and rates:
  - 48 severe credit declines (1978–2010) coincide with restructurings.
  - 40 (34) severe protracted net capital inflow declines coincide with restructurings (variant counts reported).
  - Banking crises after post-default restructurings: 30 cases out of 91 (33 percent).
  - Severe credit declines in post-default restructurings: 36 cases out of 73 (around 49 percent).
  - Net capital inflow declines in post-default restructurings: 38 (33) percent on average; in weakly preemptive and strictly preemptive: 35 and 20 (31 and 10) percent, respectively.
- Durations:
  - Average duration of restructuring strategies: post-default 5.0 years, weakly preemptive 1.1 years, strictly preemptive 0.9 years.
- Sample and period:
  - Sample period: 1970–2013.
  - Benchmark sample: 68 countries experiencing at least one restructuring.

*Source: wpiea2019069 — References; Appendix A.3; Appendix C; Appendix D.1; Appendix E (selected tables and figures).*

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

### wpiea2019069 - References

### Major findings on restructuring impacts and transmission channels
- Sample and coverage:
  - Sample of 70 countries with at least one sovereign debt restructuring over 1978–2010.
  - 179 restructurings of privately-held external sovereign debt.
  - 39 banking crises.
  - 48 severe credit declines.
  - 40 severe protracted declines in net capital inflow.
- Overall impacts by restructuring strategy (local projections, AIPW):
  - Post-default restructurings are associated with the most severe and protracted declines in GDP and investment (in percent of GDP), with cumulative contractions of 4 and 1 percentage points over the first 3 years, respectively.
  - Post-default restructurings are associated with severe and prolonged declines in bank credit and sharp increases in lending interest rates.
  - Pre-emptive restructuring events (weakly and strictly) have much milder and short-lived declines in GDP, investment, private credit, and net capital inflows.
  - Weakly preemptive restructuring effects lie between post-default and strictly preemptive effects.
- Role of capital inflow–credit channel and sudden stops:
  - Severe sudden stops (severe protracted declines in net capital inflows) are strongly associated with credit crunches and occur more likely and more severely after post-default restructurings.
  - Severe sudden stops and credit crunches are associated with the worst contractions in investment (in percent of GDP) and GDP, with cumulative declines of 4 and 5 percentage points of GDP over 2 years after a severe sudden stop occurs.
  - The capital inflow–credit channel influences GDP directly and indirectly through private credit and investment and their feedback effects, amplifying the impact of post-default restructurings.
- Banking sector linkage:
  - Logit estimates show countries undergoing post-default restructurings are significantly more likely to experience a banking crisis than those following a preemptive restructuring.
  - The deterioration of the banking sector appears central to the larger output costs observed after post-default restructurings.
- Heterogeneity by negotiation characteristics:
  - Output costs of post-default episodes with short duration or small haircuts are lower than those with long duration or high haircuts, but still higher than for weakly preemptive restructurings—suggesting avoiding missed payments matters more than creditor-friendly terms alone.

### Definitions and classification
- Restructuring strategy definitions (Asonuma and Trebesch (2016)):
  - Strictly preemptive restructurings: implemented without missing any payments (no legal default).
  - Weakly preemptive restructurings: some payments are missed, but only temporarily and after the start of formal or informal negotiations with creditor representatives (no unilateral default).
  - Post-default restructurings: payments are missed unilaterally and without the agreement of creditor representatives (unilateral default prior to negotiations).
- Severe declines definitions (3-year horizon, 25th percentile thresholds):
  - Severe credit declines: cumulative percentage point difference in the credit-to-GDP ratio over three years is less than the 25th percentile of the distribution.
  - Severe protracted net capital inflow declines (sum): sum of cumulative percentage point differences in net capital inflows-to-GDP ratio over three years is less than the 25th percentile of the distribution.
  - Severe protracted net capital inflow declines (difference): cumulative percentage point difference in net capital inflows-to-GDP ratio over three years is less than the 25th percentile of the distribution.
- Banking crisis (Laeven and Valencia (2013)): requires significant signs of financial distress in the banking system and significant intervention measures in response.

### Key dataset statistics (Table 1 summaries)
- Panel A: Debt Restructuring Sample
  - # of episodes: Post-default 111; Weakly preemptive 45; Strictly preemptive 23.
  - # of countries: Post-default 60; Weakly preemptive 26; Strictly preemptive 13.
  - Average duration (years): Post-default 5.0; Weakly preemptive 1.1; Strictly preemptive 0.9.
  - Average haircuts (percent): Post-default 48.0; Weakly preemptive 18.2; Strictly preemptive 18.9.
- Panel B: Banking Crisis Sample (3 years since start of debt restructurings, t + 1 to t + 3)
  - # of available episodes (all restructurings): 127.
  - # of banking crises within 3 years: 39.
  - Conditional probability of banking crisis after restructurings (all obs.): 30.7%.
  - Conditional probabilities by strategy: Post-default 33.0%; Weakly preemptive 17.9%; Strictly preemptive 9.5%.
- Panel C: Severe Credit Declines (3-year horizon)
  - # of available episodes (all restructurings): 107.
  - # of severe credit declines within 3 years: 48.
  - Conditional probability (all obs.): 44.9%.
  - Conditional probabilities by strategy: Post-default 49.3%; Weakly preemptive 38.5%; Strictly preemptive 25.0%.
- Panel D / E: Severe Protracted Net Capital Inflow Declines
  - # of available episodes (all restructurings): 113.
  - # of severe net capital inflow declines (sum): 40.
  - Conditional probability (sum) (all obs.): 35.4%.
  - Conditional probabilities by strategy (sum): Post-default 37.7%; Weakly preemptive 34.6%; Strictly preemptive 20.0%.
  - # of severe net capital inflow declines (difference): 40 (same count reported); conditional probability (difference) (all obs.): 35.4%; by strategy (difference): Post-default 32.5%; Weakly preemptive 30.8%; Strictly preemptive 10.0%.
- Representative episode examples listed in Table 1 (selected):
  - Post-default representative episodes: Argentina 2001-05; Russia 1998-2000.
  - Weakly preemptive representative episodes: Pakistan in 1999-2010; Belize 2006-07.
  - Strictly preemptive representative episodes: Uruguay 2003.

### Methodology and robustness
- Empirical approach:
  - Use local projection models (Jordà (2005)) to quantify cumulative effects over long horizons while controlling for dynamic feedback.
  - Baseline estimation uses Augmented Inverse Probability Weighted (AIPW) estimator (Jordà et al., 2013) to address endogeneity and selection into restructuring strategies.
  - Auxiliary analyses include logit models for banking crisis likelihood and subsample local projections to identify state-dependent responses (Auerbach and Gorodnichenko (2016); Jordà and Taylor (2016)).
- Robustness and auxiliary checks:
  - Robustness checks include alternative coding (S&P sovereign defaults), exclusion of post-2008 observations, consideration of episodes with and without official external debt restructurings or IMF-supported programs, and analyses of duration and haircut heterogeneity.
  - The restructuring strategy classification is made after completion of the episode.

_Source: wpiea2019069 - References._

### Appendix A.3—as follows:

### Appendix A.3

### Definitions of event types
- Definition iv. A ‘severe credit decline’ is an event where the cumulative percentage point difference in the credit-to-GDP ratio over three or five years,
  푑
  푖,푡+푠
  퐶
  =
  100×(퐶푟푒푑푖푡
  푖,푡+푠
  /퐺퐷푃
  푖,푡+푠
  −퐶푟푒푑푖푡
  푖,푡
  /퐺퐷푃
  푖,푡
  ), for 푠=3,5, is less than the 25th percentile of the distribution.

- Definition v. A ‘severe protracted net capital inflow decline’ is an event where:
  - (a) the sum of cumulative percentage point differences in the net capital inflows-to-GDP ratio over three or five years,
    푑
    푖,푡+푟
    푁퐶퐼,푆푈푀
    =100×
    ∑
    (푁퐶퐼
    푖,푡+푠
    /퐺퐷푃
    푖,푡+푠
    −푁퐶퐼
    푖,푡
    /퐺퐷푃
    푖,푡
    )
    푟
    푠=1
    for 푟=3,5, is less than the 25th percentile of the distribution, or
  - (b) the cumulative percentage point difference in the net capital inflows-to-GDP ratio over three or five years,
    푑
    푖,푡+푟
    푁퐶퐼,퐷퐼퐹
    =100×(푁퐶퐼
    푖,푡+푟
    /퐺퐷푃
    푖,푡+푟
    −푁퐶퐼
    푖,푡
    /퐺퐷푃
    푖,푡
    ) for 푟=3,5, is less than the 25th percentile of the distribution.

- Severe-decline dummies: the severe credit decline dummy (severe net capital inflow decline dummy) is set to 1 if a severe credit decline (protracted net capital inflow decline) occurs or continues in that year and 0 otherwise.

### Co-occurrence counts and sample evidence
- Over 1978–2010:
  - 48 severe credit declines occur simultaneously with (or follow) sequentially debt restructurings.
  - 40 (34) severe protracted net capital inflow declines occur simultaneously with (or follow) sequentially debt restructurings (reported in column 1 in Panels C, D, and E in Table 1).
- Under the 5-year criterion (Panels C, D, and E in Table A4 in Appendix A.4):
  - 49 severe credit decline episodes and 35 (32) severe protracted net capital inflow decline episodes are identified, respectively.
- Stylized fact 2: Banking crises, and severe credit and net capital inflow declines occur more frequently following post-default restructurings.

### Stylized empirical patterns (figures and panels summarized)
- Figure 1 and Figure 2 show that severe declines in credit and net capital inflows occur simultaneously with or follow sequentially debt restructurings; debt restructurings often precede or coincide with banking crises (notable cases: Argentina 2001–05 and Russia 1998–99).
- Panel summaries (from Table 1 and figures):
  - In 30 cases out of 91 post-default restructurings, banking crises occur in subsequent years (33 percent).
  - Likelihood of banking crises after:
    - post-default restructurings: 33 percent (30/91).
    - weakly preemptive restructurings: 27 percent.
    - strictly preemptive restructurings: 20 percent.
  - Banking crises precede debt restructurings in:
    - 16 cases out of 39 weakly preemptive episodes (41 percent).
    - 9 cases out of 21 strictly preemptive episodes (42 percent).
  - Panel C (Table 1): Severe credit declines occur in around 49 percent of post-default restructurings (36 cases out of 73 cases where credit data are available).
    - Severe credit declines occur in 39 percent of weakly preemptive restructurings and 25 percent of strictly preemptive restructurings.
  - Panels D and E (Table 1): Severe protracted net capital inflow declines occur in:
    - 38 (33) percent of post-default restructurings on average.
    - 35 and 20 (31 and 10) percent of weakly preemptive and strictly preemptive restructurings, respectively (two variant counts reported).

- Figure 3: All banking crises, severe credit and net capital inflow declines occur more likely following post-default restructurings than weakly or strictly preemptive restructurings (3-year horizon; confirmed with 5-year criterion in Figure A1).

### Stylized facts on output, investment, credit, and capital flows
- Stylized fact 3:
  - GDP (real, log-deviation from H-P trend) and investment (real, normalized at pre-crisis t = 0) decline substantially in post-default restructurings, less severely in weakly preemptive restructurings, and are not affected in strictly preemptive cases.
  - Investment declines more than GDP in post-default and weakly preemptive events; investment is the larger margin of adjustment.
- Stylized fact 4:
  - Private credit and capital inflows remain below the pre-crisis level and lending interest rates experience a sharp increase in post-default restructurings.
  - Private credit: severe and prolonged decline after post-default restructurings; recovers to pre-crisis level by the third year since start of restructurings.
  - Lending interest rates: sharp and temporary increase in post-default restructurings, accompanied by a sharp drop in private credit, followed by return to pre-crisis level.
  - Net capital inflows: severe and protracted decline in post-default restructurings (remain subdued for five years); milder decline in weakly preemptive restructurings; no effect for strictly preemptive restructurings.

### Data, sample coverage, and event coding
- Data sources:
  - Real US-dollar GDP, real US-dollar GDP per capita, and investment (US dollar) from Penn World Table 8.0 (Feenstra et al., 2015).
  - Private credit and lending interest rates from World Bank WDI.
  - Net capital inflows from IMF WEO.
- Sample period: 1970–2013. (Sample after 2008 dropped in certain analysis because of Global Financial Crisis influence.)
- Benchmark sample: 68 countries that experienced at least one debt restructuring over the specified horizon; robustness checks include full country coverage (Appendix E.1).
- Each restructuring considered an independent event when debt instruments subject to exchange and dates differ; overlapping observations included and treated per robustness checks (Appendix E.4).
- Debt restructuring episode coding: when more than one strategy occurs within [−2, +3] years around start year t=0, coded with the worst strategy only.

### Econometric approach and addressing endogeneity
- Endogeneity concern: restructuring strategy is an endogenous sovereign choice; conventional OLS may be biased.
- First-stage prediction (probit) uses instruments/explanatory variables:
  - external debt-to-GDP ratio (lag), total interest payments-to-GDP ratio (lag), change in credit ratings (lag), cyclical component of log of GDP per capita (lag), GDP growth rate (lag), freedom index of civil liberty (lag), counts of past restructurings and banking crises in last six years, and country fixed effects.
- Table 2 highlights pre-treatment differences:
  - Significant differences found in credit ratings, changes in credit ratings, total interest payments-to-GDP ratio at start of restructurings vs normal times.
  - Preemptive restructurings are not uniformly associated with more favorable ex ante outcomes (e.g., preemptive restructurings had worse credit ratings on average than post-default restructurings).
- Probit results (Table 3) and classification:
  - ROC areas computed for three types of restructurings based on different regressor sets (x1 and x2); x1 chosen as first-stage instrument set.
  - Kernel density estimates show substantial separation between treatment and control predicted likelihoods.

### Augmented Inverse Probability Weighted (AIPW) local projections
- First stage: estimate propensity scores Pr(R) using probit for each restructuring strategy R ∈ {PD, Weakly, Strictly}.
- Second stage: apply inverse of estimated propensity score 1/Pr(R)̂ to weight observations (AIPW) and estimate local projections for horizons h = 1..5.
- Local projection specification for cumulative GDP growth:
  - 푔
    푖,푡+ℎ
    =훼
    푖
    푅,ℎ
    +훬
    푅,ℎ
    퐷
    푖,푡+1
    푅
    +푿
    푖,푡
    휷
    푅.ℎ
    +휖
    푖,푡+ℎ
    푅
- Interpretation: estimated coefficients viewed as average treatment effects (ATE) comparing restructuring observations to weighted non-restructuring observations (formula (4) for ATE).
- AIPW results (Figure 7 summary):
  - Post-default restructurings: more severe and prolonged declines in GDP and investment than weakly preemptive restructurings; strictly preemptive restructurings have mildest impacts.
  - Private credit: severe and prolonged decline after post-default restructurings; weakly preemptive negative effects milder.
  - Lending interest rates: large and protracted rise after post-default restructurings (lending-rate estimates for weakly and strictly merged due to limited observations).
  - Capital inflows: severe decline after post-default episodes; mild decline after weakly preemptive episodes.
- OLS local projections (Appendix C.1) broadly confirm AIPW patterns; differences arise because AIPW reweights observations predicted with low probability.

### Role of capital inflow–credit channel and severe sudden stops
- Definition vi. A ‘severe sudden stop’ is an event where the percentage point difference in net capital inflows-to-GDP ratio from the pre-restructuring year to the start of restructuring (year 0 to year 1),
  푑
  푖,푡+1
  푁퐶퐼
  =100×(푁퐶퐼
  푖,푡+1
  /퐺퐷푃
  푖,푡+1
  −푁퐶퐼
  푖,푡
  /퐺퐷푃
  푖,푡
  ), for R ∈ {PD, Weakly, Strictly}, is less than the 25th or 50th percentile among observations experiencing that type of restructuring.
- Empirical findings (Figures 8 and 9, and related AIPW estimates):
  - Private credit, investment, and GDP decline more severely when restructurings coincide with a severe sudden stop (purple lines) than when they do not (brown lines).
  - The severe sudden stop amplifies declines in private credit for all restructuring types; the decline is deeper for post-default than for preemptive restructurings.
  - Investment and GDP mirror private credit dynamics: sustained declines when restructurings are accompanied by a severe sudden stop, deeper for post-default than preemptive restructurings.
- Local projection with interaction (equation (5)) estimates separate impacts for restructurings with and without severe sudden stops using a two-type first-stage model and AIPW weighting.

### Key quantitative summary points (preserve exact reported figures)
- Counts and rates:
  - 48 severe credit declines (1978–2010) coincide with restructurings.
  - 40 (34) severe protracted net capital inflow declines coincide with restructurings (variant counts reported).
  - Under 5-year criterion: 49 and 35 (32) episodes identified, respectively.
  - Banking crises after post-default restructurings: 30 cases out of 91 (33 percent).
  - Severe credit declines in post-default restructurings: 36 cases out of 73 (around 49 percent).
  - Net capital inflow declines in post-default restructurings: 38 (33) percent on average; in weakly preemptive and strictly preemptive: 35 and 20 (31 and 10) percent, respectively.
- Durations:
  - Average duration of restructuring strategies: post-default 5.0 years, weakly preemptive 1.1 years, strictly preemptive 0.9 years (used to set horizons in severe-sudden-stop analysis).
- Sample and period:
  - Sample period: 1970–2013.
  - Benchmark sample: 68 countries experiencing at least one restructuring.

*Source: Appendix A.3, wpiea2019069.*

### Appendix D.1.

### Appendix D.1.

### Overview
- Focus: whether duration and haircuts of debt restructurings explain differences in post-default output dynamics.
- Approach: comparison of post-default and weakly preemptive restructurings, subsamples matched on duration and/or haircuts, local projection estimates using information available one year prior to restructurings, AIPW (augmented inverse probability weighting) estimates, and complementary logit models.

### Definitions and classification rules (used for subsamples and estimations)
- Definition vii. A ‘post-default restructuring with high (small) GDP growth forecasts’, where expected GDP growth is above (below) the mean in post-default restructurings.
- Definition viii. A ‘post-default restructuring with long (short) expected duration’, where expected duration is above (below) the mean in post-default restructurings.
- Definition ix. A ‘post-default restructuring with high (small) expected haircuts’, where expected NPV haircuts are above (below) the mean in post-default restructurings.
- Prediction variables: short-run GDP growth forecast g_{i,t}^{F,t+1} and long-run forecast g_{i,t}^{F,Long} = sum_{s=t+2}^{t+6} g_{i,t}^{F,s} /5.
- Dummy construction for expected duration/haircuts: predicted values above the 50th percentile set to unity.

### Empirical findings — duration, haircuts, and GDP dynamics
- Matching post-default episodes to weakly preemptive episodes by duration and/or haircut produces intermediate GDP trajectories:
  - Short-duration post-default restructurings: sharp decline over first two years, then quick recovery by year 3 (brown line, Panel A described).
  - Small-haircut post-default restructurings: severe decline first two years, gradual recovery from year 3 (brown line, Panel B described).
  - Short-duration and small-haircut post-default subsample: dynamics lie between post-default and weakly preemptive cases, with sizable differences remaining in the 3–5 year horizon.
- Specific summary statistics cited:
  - Average duration: weakly preemptive restructurings 1.1 years vs post-default restructurings 5 years.
  - Average haircuts: weakly preemptive restructurings 18 percent vs post-default restructurings 48 percent.
- Interpretation: shorter duration appears more important than smaller haircuts for better post-default outcomes; however, both matter and are related.

### Predicting duration and haircuts (Table 4, Poisson / Negative Binomial regressions)
- Main coefficient estimates (selected, exactly as reported):
  - GDP growth forecasts, short-run: -0.030 (0.026), -0.028 (0.026), -0.058 (0.102) in columns (1)-(3); -0.034*** (0.006), -0.039*** (0.006), 0.003 (0.160) in columns (4)-(6).
  - GDP growth forecasts, long-run: -0.484*** (0.056), -0.547*** (0.064), -0.474** (0.199) in columns (1)-(3); -0.493*** (0.014), -0.552*** (0.016), -0.757** (0.377) in columns (4)-(6).
  - Cyclical component of log of GDP per capita: 2.307** (1.176) and 0.262 (0.302) in reported columns.
  - # of past post-default episodes: 0.219*** (0.057), 0.166** (0.071), 0.338 (0.303) and 0.194*** (0.015), 0.148*** (0.018), 0.312 (0.454).
  - # of past preemptive episodes: -0.307*** (0.090), -0.315*** (0.094), -0.498** (0.215) and -0.072*** (0.016), -0.075*** (0.016), -0.058 (0.237).
  - Constant terms: -0.406* (0.221), -0.0874 (0.255), -0.391 (0.992) and 2.309*** (0.056), 2.644*** (0.064), 3.164* (1.768).
- Sample/estimation notes:
  - Observations: 1,193; 1,056; 1,193 in columns (1)-(3) and 1,193; 1,056; 1,193 in columns (4)-(6) as presented.
  - Pseudo R-sq.: 0.125, 0.133, 0.039 and 0.127, 0.143, 0.010 respectively.
  - Models: Poisson and Negative Binomial specifications reported.
- Key inference:
  - High long-run GDP growth forecasts are associated with shorter expected duration (statistically significant).
  - High short-run and long-run GDP growth forecasts both lead to smaller expected haircuts.
  - Past post-default episodes increase expected duration and haircuts; past preemptive episodes decrease both.

### AIPW local projection estimates — heterogeneity by expected outcomes (Figure 11 summary)
- AIPW estimates show:
  - GDP declines are more severe and prolonged for post-default restructurings with:
    - low GDP growth forecasts (Panel A),
    - long expected duration (Panel B),
    - high expected haircuts (Panel C).
- Robustness: Figure D3 (Appendix D.2) contrasts baseline AIPW with one-type model AIPW and OLS local projection; results are qualitatively similar.

### Logit evidence on banking and financial sector events (Table 5)
- Model specifications:
  - Banking crisis probability at t=1 conditional on presence of restructuring types in t=-3..1.
  - Severe credit declines and net capital inflow declines defined as cumulative 5-year changes falling below the 25th percentile (-1.65 and -7.03, respectively).
  - Controls: current account-to-GDP ratio, foreign reserves-to-GDP ratio, real exchange rate (log), GDP growth rate (lag).
- Selected coefficient estimates and marginal effects (exactly as reported):
  - Banking Crises at t = 1:
    - Post-default 1{t = -3, -2, -1, 0, 1}: Coef. 1.051*** (0.279); dy/dx 0.204*** (0.075).
    - Weakly preemptive 1{t = -3, -2, -1, 0, 1}: Coef. 0.910** (0.368); dy/dx 0.176** (0.074).
    - Strictly preemptive 1{t = -3, -2, -1, 0, 1}: Coef. 0.270 (0.514); dy/dx 0.052 (0.100).
  - At t = 1 (start year):
    - Post-default 1{t=1}: Coef. 0.812*** (0.299); dy/dx 0.179*** (0.072).
    - Weakly preemptive 1{t=1}: Coef. 0.298 (0.409); dy/dx 0.066 (0.090).
    - Strictly preemptive 1{t=1}: Coef. -0.419 (0.636); dy/dx -0.092 (0.142).
  - Severe Credit Declines (cum. t=0..5):
    - Post-default 1{t=1}: Coef. 0.704** (0.300); dy/dx 0.143** (0.065).
    - Weakly preemptive 1{t=1}: Coef. 0.671* (0.389); dy/dx 0.137* (0.080).
    - Controls: GDP growth rate, lag: Coef. -6.453*** (1.226); dy/dx -1.423*** (0.330).
  - Severe Net Capital Inflow Declines (cum. t=0..5):
    - Post-default 1{t=1}: Coef. 0.704**? (table reports 0.704** in column grouping for severe credit declines; for net capital inflow declines the reported entry is 0.704**0.143** in the same row structure) — the table shows that post-default events are associated with increased probability of severe net capital inflow declines with dy/dx 0.179*** and 0.143** across columns as reported.
    - Controls with significance: current account-to-GDP ratio and foreign reserves-to-GDP ratio and real exchange rate show expected signs in various specifications; foreign reserves-to-GDP ratio: Coef. 4.506*** (1.568); dy/dx 0.917*** (0.338) in the net capital inflow decline regression.
- Aggregate interpretations:
  - Post-default restructurings in prior years are associated with higher likelihoods of banking crises and severe credit and net capital inflow declines.
    - Quantitative effects reported: banking crises more likely by 20 percent in column (1’); severe credit decline more likely by 18 percent; severe net capital inflow decline more likely by 14 percent (as summarized in text).
  - Control variables align with literature: large current account deficits and real exchange rate depreciations associated with severe net capital inflow declines; banking crises and severe credit declines followed by real exchange rate depreciation; GDP growth negative sign indicates crises during recessions.

### Policy implications and recommendations (extracted faithfully)
- Even if a sovereign must miss payments (default), adverse effects on GDP, investment, and banking sector can be mitigated by:
  - reaching an agreement with creditors quickly (short duration), and/or
  - inflicting on creditors the smallest possible losses (small haircuts) needed to restore solvency.
- Preemptive restructurings (without missing payments) are largely able to avoid or mitigate output and banking sector costs associated with restructuring.
- Constraints: sudden large shocks or "bad equilibrium" multiple-equilibria contexts may force sovereigns to stop servicing debt; still, quick suspension agreed with creditors is associated with better outcomes than unilateral suspension.
- Haircut design caveat: the haircut should be large enough to restore solvency; offering a small haircut that requires another restructuring is counter-productive and can lead to even higher costs.
- Official financing implications: where feasible, official financing and creditor cooperation that allow countries to restructure without missing payments can attenuate long-run costs; delaying adjustment and requests for official support until default becomes inevitable imposes costs.

*Source: Appendix D.1., wpiea2019069 - Appendix D.1.*

### References

### References

### Major themes covered
- Sovereign debt crises, defaults, and restructurings
- Links between sovereign risk and banking sector fragility
- Macroeconomic costs of sovereign defaults and banking crises (output, investment, credit)
- Debt renegotiation duration, delays, and creditor interactions
- Capital flows, sudden stops, and international credit supply shocks
- Measurement and databases: systemic banking crises, sovereign default databases, World Development Indicators, International Financial Statistics
- Methodologies: local projections, semiparametric estimates, general equilibrium modeling, empirical case studies

### Representative citations (selected from the list)
- Acharya, V. V., T. Eisert, C. Eufinger, and C. Hirsch, 2015, “Real Effects of the Sovereign Debt Crisis in Europe: Evidence from Syndicated Loans,” CEPR Discussion Paper 10108.
- Asonuma, T., 2016, “Serial Sovereign Defaults and Debt Restructurings,” IMF Working Paper 16/66.
- Asonuma, T., and C. Trebesch, 2016, “Sovereign Debt Restructurings: Preemptive or Post-default,” Journal of the European Economic Association, Vol.14(1), pp. 175–214.
- Bocola, L., 2015, “The Pass-Through of Sovereign Risk,” Federal Reserve Bank of Minneapolis Working Paper 722.
- Borensztein, E., and U. Panizza, 2009, “The Costs of Sovereign Default,” IMF Staff Papers, Vol. 56(4), pp. 683–741.
- Cruces, J., and C. Trebesch, 2013, “Sovereign Defaults: The Price of Haircuts,” American Economic Journal: Macroeconomics, Vol.5(3), pp. 85–117.
- Das, U., M. Papaioannou, and C. Trebesch, 2012, “Sovereign Debt Restructurings 1950–2010: Literature Survey, Data, and Stylized Facts,” IMF Working Paper 12/203.
- Gennaioli, N., A. Martin, and S. Rossi, 2018, “Banks, Government Bonds, and Default: What do the Data Say?” Journal of Monetary Economics, Vol.98, pp. 98–113.
- Jordá, Ó., M. Schularick, and A. M. Taylor,2016, “Sovereigns versus Banks: Credit, Crises, and Consequences,” Journal of the European Economic Association, Vol.14(1), pp. 45–79.
- Laeven, L., and F. Valencia, 2013, “Systemic Banking Crises Database” IMF Economic Review, Vol.61(2), pp. 225–270.
- Reinhart, M. C., and K. Rogoff, 2009, This Time is Different: Eight Centuries of Financial Folly, Princeton University Press.
- Sturzenegger, F., and J. Zettelmeyer, 2006, Debt Defaults and Lessons from a Decade of Crises, MIT Press.
- Trebesch, C., and M. Zabel, 2017, “The Output Costs of Hard and Soft Sovereign Default,” European Economic Review, Vol.92, pp.416–432.
- World Bank (WB), 2016a, Global Financial Development Database.
- International Monetary Fund (IMF), 2016, International Financial Statistics, 1948–2016.

### Data and methodological sources cited
- IMF: International Financial Statistics, 1948–2016; World Economic Outlook, April 2017; Policy Papers on The Fund’s Lending Framework and Sovereign Debt (2014, 2015).
- World Bank: Global Financial Development Database (2016a); World Development Indicators (2016b).
- Databases and working-paper collections: Systemic Banking Crises Database; sovereign debt/restructuring datasets and surveys spanning 1950–2015.

*References section of the source document.*

### Appendix A. Data

### Appendix A. Data

### A.1. Sample Countries
- Dataset includes only countries that experienced debt restructurings.
- 60 countries experienced 111 episodes of post-default debt restructuring between 1978 and 2010: Albania, Algeria, Argentina, Bolivia, Bosnia and Herzegovina, Brazil, Bulgaria, Cameroon, Congo Rep., Costa Rica, Cote d’Ivoire, Croatia, Cuba, Dominican Republic, Ecuador, Ethiopia, Gabon, the Gambia, Guinea, Guyana, Honduras, Iraq, Jamaica, Jordan, Kenya, Liberia, Macedonia (FYR), Madagascar, Malawi, Mauritania, Moldova, Morocco, Mozambique, Nicaragua, Niger, Nigeria, Pakistan, Panama, Paraguay, Peru, the Philippines, Poland, Romania, Russian Federation, Sao Tome and Principe, Senegal, Serbia and Montenegro, Seychelles, Sierra Leone, Slovenia, Sudan, Tanzania, Togo, Turkey, Uganda, Venezuela, Vietnam, Yemen, Rep. of, Zaire, and Zambia.
- 26 countries experienced 46 episodes of weakly preemptive debt restructuring: Argentina, Belize, Brazil, Chile, Dominica, Dominican Republic, Ecuador, Grenada, Jamaica, Nigeria, Malawi, Mexico, Morocco, Niger, Panama, Peru, the Philippines, Romania, Senegal, South Africa, Trinidad and Tobago, Turkey, Ukraine, Uruguay, Venezuela, and Yugoslavia.
- 13 countries experienced 23 episodes of strictly preemptive debt restructuring: Algeria, Chile, Dominican Republic, Jamaica, Mexico, Moldova, Nicaragua, Pakistan, Peru, South Africa, Ukraine, Uruguay, and Yugoslavia.

### A.2. Data Sources
- Private external debt restructuring data: Asonuma and Trebesch (2016)
- Banking crisis data: Laeven and Valencia (2013)
- Terms of trade: PWT 8.0
- Population: WDI (World Bank, 2016a)
- Openness: Authors’ calculation based on PWT 8.0 data
- Investment: WDI (World Bank, 2016a)
- Real exchange rates: IMF International Financial Statistics
- Real GDP, US-dollar denominated: WDI (World Bank, 2016a)
- Real GDP per capita: WDI (World Bank, 2016a)
- Real GDP growth rates: Authors’ calculation based on WDI (World Bank, 2016a)
- Cyclical component of real GDP per capita: Authors’ calculation based on WDI (World Bank, 2016a)
- Private credit (real): WDI (World Bank, 2016a)
- Lending interest rates: IMF International Financial Statistics
- Net capital inflows: IMF World Economic Outlook
- Exchange rate regime classification: Ilzetzki, Reinhart and Rogoff (2015)
- S&P sovereign default data: Standard and Poor’s
- Official external (Paris Club) debt restructuring data: Das et al. (2012) and Paris Club
- Public debt-to-GDP ratio: World Bank’s Global Financial Development Database
- Countries’ credit ratings: the Institutional Investor magazine
- Political stability index (civil liberties): FreedomHouse
- Total interest payments-to-GDP ratio: World Bank’s Global Financial Development Database
- Short-term interest payments-to-GDP ratio: World Bank’s Global Financial Development Database
- Haircuts on private external debt restructurings: Cruces and Trebesch (2013)
- Coerciveness index on private external debt restructurings: Enderlein et al. (2012)
- Legislative and executive indices of electoral competitiveness (LIEC and EIEC): Inter-American Development Bank’s Database of Political Institutions 2015 (DIP2015)
- Notes: PWT and WDI stand for the Penn World Table and the World Development Indicators, respectively.

### A.3. Data Statistics (selected summary statistics from Table A1)
- Observations and summary statistics (Obs, Mean, Std. Dev., Min, Max) for baseline variables:
  - GDP (billion USD, contant 2010 prices): Obs = 2,501; Mean = 98.47; Std. Dev. = 251.01; Min = 0.16; Max = 2400
  - Investment-to-GDP ratio (percent): Obs = 2,205; Mean = 22.17; Std. Dev. = 8.39; Min = -5.74; Max = 73.49
  - Private credit-to-GDP ratio (percent): Obs = 2,345; Mean = 27.47; Std. Dev. = 18.27; Min = 1.66; Max = 105.92
  - Lending interest rate (percent): Obs = 1,361; Mean = 19.27; Std. Dev. = 13.50; Min = 3.00; Max = 95.08
  - Net capital inflows-to-GDP ratio (percent): Obs = 2,069; Mean = 0.08; Std. Dev. = 0.15; Min = -1.15; Max = 1.24
  - Cyclical component of log of GDP per capita: Obs = 2,159; Mean = 0.00; Std. Dev. = 0.07; Min = -0.65; Max = 0.52
  - Public debt-to-GDP ratio (percent): Obs = 1,535; Mean = 73.49; Std. Dev. = 47.47; Min = 1.03; Max = 316.08
  - Current account-to-GDP ratio (percent): Obs = 1,986; Mean = -6.62; Std. Dev. = 17.60; Min = -248.13; Max = 49.16
  - Foreign reserves-to-GDP ratio (percent): Obs = 1,949; Mean = 0.00; Std. Dev. = 0.10; Min = -0.65; Max = 2.61
  - ln(Real exchange rate): Obs = 2,525; Mean = 3.71; Std. Dev. = 2.56; Min = -1.84; Max = 10.53
  - GDP growth rate (percent): Obs = 2,437; Mean = 3.66; Std. Dev. = 7.08; Min = -63.38; Max = 107.41
- Additional variables (Obs, Mean, Std. Dev., Min, Max):
  - # of past post-defaults: Obs = 2,816; Mean = 0.87; Std. Dev. = 1.09; Min = 0; Max = 6
  - # of past weakly preemptive: Obs = 2,816; Mean = 0.37; Std. Dev. = 0.81; Min = 0; Max = 4
  - # of past strictly preemptive: Obs = 2,816; Mean = 0.19; Std. Dev. = 0.68; Min = 0; Max = 5
  - # of past banking crises: Obs = 2,816; Mean = 1.56; Std. Dev. = 2.36; Min = 0; Max = 10
  - Total external debt-to-GDP ratio (percent): Obs = 2,107; Mean = 0.76; Std. Dev. = 1.09; Min = 0; Max = 18.47
  - Interest payments-to-GDP ratio (percent): Obs = 2,107; Mean = 0.02; Std. Dev. = 0.03; Min = 0; Max = 0.64
  - Credit ratings, change: Obs = 1,503; Mean = 0.24; Std. Dev. = 3.15; Min = -22.28; Max = 10.48
  - Freedom house index, civil liberty: Obs = 2,467; Mean = 4.11; Std. Dev. = 1.64; Min = 1; Max = 7
  - # of years from the last debt restructurings: Obs = 1,791; Mean = 14.42; Std. Dev. = 9.20; Min = 0; Max = 38
  - LIEC (Political stability index): Obs = 2,296; Mean = 5.54; Std. Dev. = 2.03; Min = 1; Max = 7
  - EIEC (Political stability index): Obs = 2,296; Mean = 5.28; Std. Dev. = 2.13; Min = 1; Max = 7
  - Coerciveness index: Obs = 713; Mean = 1.10; Std. Dev. = 1.98; Min = 0; Max = 9
  - Forecastted GDP growth rate, long-run: Obs = 1,472; Mean = 4.71; Std. Dev. = 1.98; Min = -1.82; Max = 30.03
  - Forecastted GDP growth rate, short-run: Obs = 1,481; Mean = 4.58; Std. Dev. = 2.72; Min = -6.66; Max = 55.56

### A.3. Debt Restructurings, Private Credit, and Net Capital Inflows (selected results)
- Table A2: 3-Year Horizon (3-year cumulative percentage point differences from Year 0)
  - Panel A: Private Credit (푑푖,푡+3퐶 = 100×(Creditᵢ,ₜ+3/GDPᵢ,ₜ+3 − Creditᵢ,ₜ/GDPᵢ,ₜ))
    - Post-default: Obs = 73; Mean = -1.63; p(5) = -13.67; p(25) = -4.94; p(50) = -1.21; p(75) = 0.80; p(95) = 6.56
    - Weakly preemptive: Obs = 26; Mean = 0.47; p(5) = -8.51; p(25) = -2.62; p(50) = -1.02; p(75) = 4.52; p(95) = 8.73
    - Strictly preemptive: Obs = 8; Mean = 0.31; p(5) = -9.54; p(25) = -2.00; p(50) = -0.28; p(75) = 2.52; p(95) = 11.51
    - All obs. from restructuring countries: Obs = 1,895; Mean = 1.56; p(5) = -8.13; p(25) = -1.22; p(50) = 1.53; p(75) = 4.65; p(95) = 12.17
  - Panel B: Net Capital Inflows (sum over years 1–3)
    - Post-default: Obs = 77; Mean = -10.18; p(5) = -77.06; p(25) = -31.54; p(50) = -4.35; p(75) = 9.78; p(95) = 41.41
    - Weakly preemptive: Obs = 26; Mean = -4.80; p(5) = -53.12; p(25) = -16.71; p(50) = -0.94; p(75) = 11.42; p(95) = 29.59
    - Strictly preemptive: Obs = 10; Mean = 11.02; p(5) = -31.33; p(25) = 1.90; p(50) = 6.49; p(75) = 17.14; p(95) = 91.91
    - All obs. from restructuring countries: Obs = 1,625; Mean = 1.48; p(5) = -51.47; p(25) = -13.40; p(50) = 0.83; p(75) = 15.69; p(95) = 53.76
  - Panel C: Net Capital Inflows (3-year cumulative difference)
    - Post-default: Obs = 77; Mean = -2.42; p(5) = -31.50; p(25) = -8.77; p(50) = -1.88; p(75) = 3.78; p(95) = 16.83
    - Weakly preemptive: Obs = 26; Mean = 0.10; p(5) = -23.76; p(25) = -8.53; p(50) = 1.35; p(75) = 6.94; p(95) = 26.25
    - Strictly preemptive: Obs = 10; Mean = 6.04; p(5) = -13.95; p(25) = -0.01; p(50) = 5.49; p(75) = 8.12; p(95) = 38.65
    - All obs. from restructuring countries: Obs = 1,625; Mean = 0.53; p(5) = -22.00; p(25) = -5.90; p(50) = 0.48; p(75) = 6.81; p(95) = 23.33
- Table A3: 5-Year Horizon (5-year cumulative percentage point differences from Year 0)
  - Panel A: Private Credit (푑푖,푡+5퐶 = 100×(Creditᵢ,ₜ+5/GDPᵢ,ₜ+5 − Creditᵢ,ₜ/GDPᵢ,ₜ))
    - Post-default: Obs = 73; Mean = -1.72; p(5) = -14.79; p(25) = -6.63; p(50) = -2.35; p(75) = 1.99; p(95) = 10.15
    - Weakly preemptive: Obs = 26; Mean = 2.75; p(5) = -10.04; p(25) = -4.15; p(50) = 0.23; p(75) = 10.86; p(95) = 19.82
    - Strictly preemptive: Obs = 9; Mean = 1.11; p(5) = -13.72; p(25) = -1.48; p(50) = 2.00; p(75) = 2.63; p(95) = 12.92
    - All obs. from restructuring countries: Obs = 1,888; Mean = 2.74; p(5) = -9.78; p(25) = -1.65; p(50) = 2.54; p(75) = 7.21; p(95) = 17.19
  - Panel B: Net Capital Inflows (sum over years 1–5)
    - Post-default: Obs = 73; Mean = -16.72; p(5) = -123.09; p(25) = -42.56; p(50) = -8.22; p(75) = 14.77; p(95) = 70.11
    - Weakly preemptive: Obs = 26; Mean = -4.34; p(5) = -88.06; p(25) = -25.93; p(50) = -0.33; p(75) = 32.13; p(95) = 62.27
    - Strictly preemptive: Obs = 10; Mean = 31.56; p(5) = -45.85; p(25) = 7.65; p(50) = 24.34; p(75) = 39.67; p(95) = 185.32
    - All obs. from restructuring countries: Obs = 1,599; Mean = 2.34; p(5) = -83.49; p(25) = -22.48; p(50) = 1.70; p(75) = 27.13; p(95) = 94.16
  - Panel C: Net Capital Inflows (5-year cumulative difference)
    - Post-default: Obs = 73; Mean = -3.97; p(5) = -31.03; p(25) = -12.26; p(50) = -3.21; p(75) = 5.83; p(95) = 19.00
    - Weakly preemptive: Obs = 26; Mean = 0.98; p(5) = -22.05; p(25) = -7.00; p(50) = 0.98; p(75) = 9.02; p(95) = 18.84
    - Strictly preemptive: Obs = 10; Mean = 11.54; p(5) = -5.32; p(25) = 2.18; p(50) = 8.59; p(75) = 18.12; p(95) = 44.88
    - All obs. from restructuring countries: Obs = 1,599; Mean = 0.49; p(5) = -24.13; p(25) = -7.05; p(50) = 0.57; p(75) = 7.77; p(95) = 25.96

### A.4. Severe Credit Declines and Severe Protracted Net Capital Inflow Declines (Table A4, 5-year Horizon)
- Panel A: Debt Restructuring Sample
  - Post-default: # of episodes = 111; # of countries = 60; Average duration (years) = 5.0; Average haircuts (percent) = 48.0; Representative episodes: Argentina 2001-05, Ukraine in 1999-2010
  - Weakly preemptive: # of episodes = 45; # of countries = 26; Average duration (years) = 1.1; Average haircuts (percent) = 18.2; Representative episodes: Pakistan (Global Exch.) 2000, (Ext. bonds) 1999, Belize 2006-07
  - Strictly preemptive: # of episodes = 23; # of countries = 13; Average duration (years) = 0.9; Average haircuts (percent) = 18.9; Representative episodes: Uruguay 2003
- Panel B: Banking Crisis Sample (banking crises over five years since the start of debt restructurings, year from t + 1 to t + 5)
  - All debt restructurings: # of available episodes = 127; # of banking crises = 39; Conditional probability = 30.7%; Representative episodes: Argentina 2001-05
  - Post-default: # of available episodes = 91; # of banking crises = 30; Conditional probability = 33.0%; Representative episodes: Turkey 1981
  - Weakly preemptive: # of available episodes = 39; # of banking crises = 7; Conditional probability = 17.9%; Representative episodes: Algeria 1990
  - Strictly preemptive: # of available episodes = 21; # of banking crises = 2; Conditional probability = 9.5%; Representative episodes: Russia 1998-2000, Niger 1983, Ukraine 1998
  - All obs. 1/: # of available episodes = 1,126; # of banking crises = 71; Conditional probability = 6.3%
- Panel C: Severe Credit Declines (defined as cumulative 5-year credit-to-GDP difference less than 25th percentile)
  - All debt restructurings: # of available episodes = 107; # of severe credit declines = 49; Conditional probability = 45.8%; Representative episodes: Turkey 1976
  - Post-default: # of available episodes = 73; # of severe credit declines = 39; Conditional probability = 53.4%; Representative episodes: Brazil 1983
  - Weakly preemptive: # of available episodes = 26; # of severe credit declines = 8; Conditional probability = 30.8%; Representative episodes: Algeria 1990
  - Strictly preemptive: # of available episodes = 8; # of severe credit declines = 2; Conditional probability = 25.0%; Representative episodes: Togo 1991
  - All obs. 1/: # of available episodes = 1,427; # of severe credit declines = 314; Conditional probability = 22.0%
- Panel D: Severe Protracted Net Capital Inflow Declines (Sum) (defined as sum of cumulative 5-year net capital inflows differences less than 25th percentile)
  - All debt restructurings: # of available episodes = 109; # of severe net cap. inf. dec. = 35; Conditional probability = 32.1%; Representative episodes: Costa Rica 1981
  - Post-default: # of available episodes = 73; # of severe net cap. inf. dec. = 25; Conditional probability = 34.2%; Representative episodes: Brazil 1982
  - Weakly preemptive: # of available episodes = 26; # of severe net cap. inf. dec. = 9; Conditional probability = 34.6%; Representative episodes: Ukraine 1998
  - Strictly preemptive: # of available episodes = 10; # of severe net cap. inf. dec. = 1; Conditional probability = 10.0%; Representative episodes: Moldova 2001
  - All obs. 1/: # of available episodes = 1,161; # of severe net cap. inf. dec. = 303; Conditional probability = 26.1%
- Panel E: Severe Protracted Net Capital Inflow Declines (Difference) (defined as 5-year cumulative net capital inflows difference less than 25th percentile)
  - All debt restructurings: # of available episodes = 109; # of severe net cap. inf. dec. = 32; Conditional probability = 29.4%
  - Post-default: # of available episodes = 73; # of severe net cap. inf. dec. = 26; Conditional probability = 35.6%
  - Weakly preemptive: # of available episodes = 26; # of severe net cap. inf. dec. = 6; Conditional probability = 23.1%
  - Strictly preemptive: # of available episodes = 10; # of severe net cap. inf. dec. = 0; Conditional probability = 0.0%
  - All obs. 1/: # of available episodes = 1,161; # of severe net cap. inf. dec. = 303; Conditional probability = 26.1%
- Notes:
  - 1/ Including episodes in countries without any experience of private external debt restructurings.
  - 2/ Banking crises over five years since the start of debt restructurings, year from t + 1 to t + 5.
  - 3/ Severe credit declines defined as cumulative 5-year credit-to-GDP difference less than the 25th percentile.
  - 4/ Severe protracted net capital inflow declines (sum) defined as sum of cumulative 5-year net capital inflows differences less than the 25th percentile.
  - 5/ Severe protracted net capital inflow declines (difference) defined as 5-year cumulative net capital inflows difference less than the 25th percentile.
  - 6/ Sample after 2008 is dropped because macroeconomic variables in the period are largely influenced by the Global Financial Crisis. Debt restructuring episodes with more than one strategy within a window of [−2, +3] years around the start year of a debt restructuring t = 0 are coded with the worst strategy only.

### Figure A1 (notes summary)
- Figure panels show probability of banking crises (Panel A), severe credit declines (Panel B), and severe protracted net capital inflow declines (Panels C and D) at the 5-year horizon.
- Definitions are consistent with Table A4 notes.
- Sample after 2008 is dropped for consistency with Table A4.

### Appendix B. Further Empirical Analysis — Support for Stylized Facts
- Objective: Test whether GDP and investment are more tightly associated during debt restructuring periods than non-restructuring periods.
- Method: Ordinary least square (OLS) estimator on GDP growth rates with contemporaneous investment growth rates across samples (non-restructuring, countries that experienced at least one debt restructuring, post-default, weakly preemptive, strictly preemptive).
- Key findings (Table B1: Dep. Var. = 100×(GDPᵢ,ₜ+1 − GDPᵢ,ₜ)/GDPᵢ,ₜ; explanatory variable: 100×(Investmentᵢ,ₜ+1 − Investmentᵢ,ₜ)/GDPᵢ,ₜ):
  - All countries (column (1)):
    - Investment growth rates coefficient = 0.431*** (standard error = 0.074)
    - Constant = 3.483*** (standard error = 0.097)
    - Observations = 4,230
    - R-squared = 0.264
  - Countries that experienced at least one debt restructuring (column (2)):
    - Investment growth rates coefficient = 0.377*** (standard error = 0.137)
    - Constant = 3.459*** (standard error = 0.180)
    - Observations = 1,407
    - R-squared = 0.189
  - During post-default (the entire period) (column (3)):
    - Investment growth rates coefficient = 0.736*** (standard error = 0.094)
    - Constant = 1.963*** (standard error = 0.272)
    - Observations = 297
    - R-squared = 0.324
  - During post-default (the first half period) (column (4)):
    - Investment growth rates coefficient = 0.768*** (standard error = 0.076)
    - Constant = 1.340*** (standard error = 0.378)
    - Observations = 187
    - R-squared = 0.353
  - During weakly preemptive (column (5)):
    - Investment growth rates coefficient = 0.763*** (standard error = 0.176)
    - Constant = 2.384*** (standard error = 0.348)
    - Observations = 65
    - R-squared = 0.41
  - During strictly preemptive (column (6)):
    - Investment growth rates coefficient = 0.486** (standard error = 0.182)
    - Constant = 2.286*** (standard error = 0.624)
    - Observations = 18
    - R-squared = 0.321
- In all regressions: Country fixed effects included. Standard errors are reported in parentheses. ***, **, and * indicate statistical significance at 1 percent, 5 percent, and 10 percent, respectively.
- Conclusion: Estimated coefficients and explanatory power (푅2) of investment growth rates are higher in debt restructuring periods (post-default, weakly preemptive, strictly preemptive) than in non-restructuring periods and the sample of all countries, indicating GDP growth is more tightly linked with investment growth during debt restructuring periods.

*Source: Appendix A and Appendix B (data and tables) from the provided content unit.*

### Appendix C. Estimation Results from Local Projections

### Appendix C. Estimation Results from Local Projections

### C.1. Ordinary Least Square (OLS) Estimator — Key Estimation Results (Panels A–E)

- Notes: Local projections (OLS) for h = 1,2,..,5. Regressions include a constant term, lagged dependent variables (lag 1 and lag 2), cyclical component of log of GDP (HP filter, smoothing parameter = 100), and country fixed effects. Robust standard errors, clustered at the country-level, are in parentheses. ***, **, and * indicate statistical significance at 1 percent, 5 percent, and 10 percent level, respectively. Sample period: 1970–2013.

- Panel A: Dep. Var. = 100×(GDP_{i,t+h} − GDP_{i,t})/GDP_{i,t}
  - h = 1, h = 2, h = 3, h = 4, h = 5, Sum
  - Post-default: -1.930*** (0.707), -3.047*** (0.777), -3.477*** (0.672), -3.579*** (0.916), -3.605** (1.495), Sum = -15.10*** (2.874)
  - Weakly preemptive: -1.741** (0.665), -1.909** (0.895), -1.322 (1.218), -1.375 (1.683), -3.036 (1.853), Sum = -9.197* (4.914)
  - Strictly preemptive: -1.790*** (0.594), -1.273 (1.089), -0.773 (1.469), -0.923 (1.539), -1.359 (2.221), Sum = -5.862 (4.002)
  - R-squared: 0.412, 0.622, 0.701, 0.702, 0.664, Sum R-squared = 0.768
  - # of obs.: 2,044, 2,044, 1,985, 1,926, 1,867, Sum obs = 1,867
  - # of countries: 59, 59, 59, 59, 59, 59
  - # of post-default obs.: 65, 65, 65, 65, 65, 65
  - # of weakly preemptive obs.: 31, 31, 31, 31, 31, 31
  - # of strictly preemptive obs.: 11, 11, 11, 11, 11, 11

- Panel B: Dep. Var. = 100×(Investment_{i,t+h}/GDP_{i,t+h} − Investment_{i,t}/GDP_{i,t})
  - h = 1, h = 2, h = 3, h = 4, h = 5, Sum
  - Post-default: -2.196** (0.847), -3.669*** (1.035), -2.314** (1.013), -1.856* (0.933), -2.618** (1.024), Sum = -12.35*** (4.064)
  - Weakly preemptive: -1.726** (0.690), -2.557** (1.064), -2.045** (0.949), -1.950** (0.927), -2.620** (1.290), Sum = -11.03** (4.201)
  - Strictly preemptive: -0.639 (0.412), 0.702 (0.992), 2.086* (1.096), 2.166* (1.173), 1.822* (0.984), Sum = 6.055 (3.783)
  - R-squared: 0.054, 0.138, 0.179, 0.191, 0.199, Sum R-squared = 0.235
  - # of obs.: 1,757, 1,757, 1,702, 1,647, 1,592, Sum obs = 1,592
  - # of countries: 55 (all horizons)
  - # of post-default obs.: 56 (all horizons)
  - # of weakly preemptive obs.: 28 (all horizons)
  - # of strictly preemptive obs.: 11 (all horizons)

- Panel C: Dep. Var. = 100×(Credit_{i,t+h}/GDP_{i,t+h} − Credit_{i,t}/GDP_{i,t})
  - h = 1, h = 2, h = 3, h = 4, h = 5, Sum
  - Post-default: -0.066 (0.370), -1.956*** (0.527), -2.800*** (0.653), -3.252*** (0.857), -3.832*** (0.805), Sum = -12.06*** (2.949)
  - Weakly preemptive: -0.151 (0.460), -1.506** (0.670), -2.083** (1.005), -2.472* (1.409), -2.959* (1.579), Sum = -9.340** (4.558)
  - Strictly preemptive: 0.287 (0.341), -0.865** (0.425), -0.906 (1.578), -0.829 (2.344), -1.225 (3.447), Sum = -3.327 (8.296)
  - R-squared: 0.073, 0.049, 0.038, 0.029, 0.026, Sum R-squared = 0.046
  - # of observations: 1,855, 1,848, 1,787, 1,728, 1,669, Sum obs = 1,646
  - # of countries: 59, 59, 59, 59, 58, 58
  - # of post-default obs.: 57, 56, 56, 56, 56, 56
  - # of weakly preemptive obs.: 31, 31, 31, 31, 31, 31
  - # of strictly preemptive obs.: 10, 10, 9, 10, 10, 9

- Panel D: Dep. Var. = 100×(LendingRate_{i,t+h} − LendingRate_{i,t})
  - h = 1, h = 2, h = 3, h = 4, h = 5, Sum
  - Post-default: 4.471 (3.45), 0.099 (7.15), 9.919 (5.97), 9.451 (6.35), 11.12* (6.57), Sum = 44.56 (29.61)
  - Weakly preemptive: 12.78 (9.04), 19.90 (17.52), 17.81 (14.21), 20.87 (15.67), 24.74 (16.58), Sum = 17.80 (87.96)
  - Strictly preemptive: 5.68 (4.02), 10.86 (9.99), 6.27 (6.43), 1.97 (5.41), 0.40 (6.45), Sum = 29.90 (31.73)
  - R-squared: 0.12, 0.226, 0.498, 0.519, 0.524, Sum R-squared = 0.419
  - # of observations: 1,031, 1,018, 965, 914, 864, Sum obs = 847
  - # of countries: 52, 51, 51, 51, 51, 51
  - # of post-default obs.: 24, 24, 23, 23, 23, 23
  - # of weakly preemptive obs.: 15, 15, 15, 14, 14, 14
  - # of strictly preemptive obs.: 9, 9, 9, 9, 9, 9

- Panel E: Dep. Var. = 100×(NCI_{i,t+h}/GDP_{i,t+h} − NCI_{i,t}/GDP_{i,t})
  - h = 1, h = 2, h = 3, h = 4, h = 5, Sum
  - Post-default: -1.934 (1.225), -6.094*** (2.220), -0.387 (2.435), -1.197 (1.781), -3.898*** (1.307), Sum = -13.78*** (5.108)
  - Weakly preemptive: -6.394*** (1.597), -5.142** (2.234), -3.443 (2.387), -4.671* (2.537), -4.427 (3.254), Sum = -22.87** (9.941)
  - Strictly preemptive: 1.412 (1.326), 3.215 (3.094), 5.149** (2.194), 4.438* (2.382), 9.384*** (1.842), Sum = 23.07*** (3.756)
  - R-squared: 0.165, 0.234, 0.214, 0.187, 0.192, Sum R-squared = 0.328
  - # of observations: 1,662, 1,651, 1,588, 1,527, 1,466, Sum obs = 1,454
  - # of countries: 58 (all horizons)
  - # of post-default obs.: 62, 62, 62, 62, 60, 60
  - # of weakly preemptive obs.: 31 (all horizons)
  - # of strictly preemptive obs.: 10 (all horizons)

### C.2. Augmented Inverse Probability Weighted (AIPW) Estimator — Key Estimation Results (Panels A–E)

- Notes: Local projections (AIPW) for h = 1,2,..,5. Regressions include a constant term, lagged dependent variable (lag 1), cyclical component of log of GDP (HP filter, smoothing parameter = 100), and country fixed effects. Robust standard errors, clustered at the country-level, are in parentheses. ***, **, and * indicate statistical significance at 1 percent, 5 percent, and 10 percent level, respectively.

- Panel A: Dep. Var. = 100×(GDP_{i,t+h} − GDP_{i,t})/GDP_{i,t}
  - h = 1, h = 2, h = 3, h = 4, h = 5, Sum
  - Post-default: -2.486*** (0.314), -3.141*** (0.461), -3.777*** (0.479), -2.422*** (0.607), -5.178*** (0.966), Sum = -16.21*** (2.237)
  - # of obs.: 825, 825, 825, 825, 796, Sum obs = 796
  - # of countries: 29 (all horizons)
  - # of debt restructuring obs.: 44 (all horizons)
  - Weakly preemptive: -2.270*** (0.340), -2.602*** (0.453), -1.557*** (0.455), -2.591*** (0.632), -4.147*** (0.841), Sum = -12.81*** (2.116)
  - # of obs.: 371, 371, 371, 371, 357, Sum obs = 357
  - # of countries: 14 (all horizons)
  - # of debt restructuring obs.: 21 (all horizons)
  - Strictly preemptive: -1.908*** (0.276), -1.069** (0.434), 1.237 (0.887), 2.197* (1.218), 3.124* (1.870), Sum = 3.921 (4.028)
  - # of obs.: 287, 287, 280, 273, 266, Sum obs = 266
  - # of countries: 7 (all horizons)
  - # of debt restructuring obs.: 11 (all horizons)

- Panel B: Dep. Var. = 100×(Investment_{i,t+h}/GDP_{i,t+h} − Investment_{i,t}/GDP_{i,t})
  - h = 1, h = 2, h = 3, h = 4, h = 5, Sum
  - Post-default: -1.930*** (0.396), -2.317*** (0.433), -0.798* (0.430), -0.186 (0.578), -0.679 (0.686), Sum = -5.714** (2.301)
  - # of obs.: 737, 737, 737, 737, 711, Sum obs = 711
  - # of countries: 26 (all horizons)
  - # of debt restructuring obs.: 40 (all horizons)
  - Weakly preemptive: -0.211 (0.338), -0.386 (0.447), 0.180 (0.387), -0.664 (0.420), -1.931*** (0.482), Sum = -2.726 (1.814)
  - # of obs.: 343, 343, 343, 343, 330, Sum obs = 330
  - # of countries: 13 (all horizons)
  - # of debt restructuring obs.: 20 (all horizons)
  - Strictly preemptive: -0.230 (0.470), 0.097 (0.530), 1.810*** (0.488), 2.333*** (0.528), 1.425*** (0.470), Sum = 5.628*** (1.837)
  - # of obs.: 287, 287, 280, 273, 266, Sum obs = 266
  - # of countries: 7 (all horizons)
  - # of debt restructuring obs.: 11 (all horizons)

- Panel C: Dep. Var. = 100×(Credit_{i,t+h}/GDP_{i,t+h} − Credit_{i,t}/GDP_{i,t})
  - h = 1, h = 2, h = 3, h = 4, h = 5, Sum
  - Post-default: -0.230 (0.217), -2.038*** (0.347), -2.420*** (0.437), -2.381*** (0.516), -2.563*** (0.588), Sum = -9.883*** (1.897)
  - # of obs.: 752, 750, 748, 747, 720, Sum obs = 716
  - # of countries: 27 (all horizons)
  - # of debt restructuring obs.: 41, 40, 40, 40, 40, 40
  - Weakly preemptive: 0.136 (0.192), -1.157*** (0.276), -1.617*** (0.358), -1.824*** (0.471), -2.418*** (0.550), Sum = -7.375*** (1.590)
  - # of obs.: 371, 371, 371, 371, 357, Sum obs = 357
  - # of countries: 14 (all horizons)
  - # of debt restructuring obs.: 21 (all horizons)
  - Strictly preemptive: 0.881*** (0.331), 0.433 (0.505), 0.275 (0.508), -0.452 (0.611), -1.733** (0.736), Sum = 0.715 (2.131)
  - # of obs.: 284, 284, 277, 270, 263, Sum obs = 259
  - # of countries: 7 (all horizons)
  - # of debt restructuring obs.: 10, 10, 9, 10, 10, 9

- Panel D: Dep. Var. = 100×(LendingRate_{i,t+h} − LendingRate_{i,t})
  - h = 1, h = 2, h = 3, h = 4, h = 5, Sum
  - Post-default: 2.633*** (0.410), 7.469*** (0.692), 9.323*** (0.893), 7.412*** (0.906), 10.59*** (1.137), Sum = 38.20*** (3.475)
  - # of obs.: 356, 350, 345, 340, 325, Sum obs = 320
  - # of countries: 15, 15, 14, 14, 14, 14
  - # of debt restructuring obs.: 22, 22, 21, 21, 21, 21
  - Weakly & strictly preemptive (combined): 1.507 (1.357), 0.997 (1.671), 5.140*** (1.914), 5.829*** (1.985), 5.233** (2.100), Sum = 19.20** (9.185)
  - # of obs.: 135, 133, 132, 130, 122, Sum obs = 121
  - # of countries: 7 (all horizons)
  - # of debt restructuring obs.: 2 (all horizons)

- Panel E: Dep. Var. = 100×(NCI_{i,t+h}/GDP_{i,t+h} − NCI_{i,t}/GDP_{i,t})
  - h = 1, h = 2, h = 3, h = 4, h = 5, Sum
  - Post-default: -1.365 (1.056), -4.904*** (1.326), 0.034 (1.218), -1.926* (1.024), -2.972*** (1.017), Sum = -10.93*** (4.078)
  - # of obs.: 792, 787, 783, 778, 747, Sum obs = 744
  - # of countries: 29 (all horizons)
  - # of debt restructuring obs.: 44, 44, 44, 44, 44, 43
  - Weakly preemptive: -3.753*** (0.782), -1.112 (0.925), 0.627 (0.933), -2.060* (1.069), -2.024* (1.151), Sum = -7.759** (3.950)
  - # of obs.: 360, 358, 357, 357, 343, Sum obs = 338
  - # of countries: 14 (all horizons)
  - # of debt restructuring obs.: 21 (all horizons)
  - Strictly preemptive: 0.5717 (1.152), 0.169 (1.297), 2.041 (1.594), 2.597** (1.281), 7.287*** (1.607), Sum = 12.71** (5.784)
  - # of obs.: 251, 251, 244, 237, 230, Sum obs = 230
  - # of countries: 7 (all horizons)
  - # of debt restructuring obs.: 10 (all horizons)

*Source: Appendix C. Estimation Results from Local Projections (tables and notes as provided).*

### Appendix D. Further Regression Results - Support for Main Results

### Appendix D. Further Regression Results - Support for Main Results

### D.1. Capital Inflow-Credit Channel — key regression findings
- OLS local projections (Panel A: Private Credit)
  - Post-default, with sudden stops: Sum = -122.7*** (h = 1..5 coefficients: -4.027, -18.19***, -27.21***, -31.52***, -37.39**; robust SEs shown in table).
  - Post-default, without sudden stops: Sum = -85.97*** (h = 1..5 coefficients: -2.879, -14.39***, -20.10***, -21.39***, -26.45***).
  - Weakly preemptive, with sudden stops: Sum = -77.32* (h = 1..5: -8.771**, -16.11**, -18.13**, -13.78, -23.03*).
  - Weakly & strictly preemptive, with sudden stops: Sum = -100.3** (h = 1..5: -8.285**, -16.50***, -19.77**, -20.68*, -34.59**).
  - R-squared by horizon: 0.079, 0.039, 0.028, 0.029, 0.039; Sum R-squared = 0.049.
  - Sample sizes: # of observations by horizon = 1,904; 1,898; 1,836; 1,775; 1,716; Sum sample = 1,692. # of countries = 59.
- OLS local projections (Panel B: Investment)
  - Post-default, with sudden stops: h = 1..5 coefficients (AIPW table excerpt) include negative and statistically significant short-run declines (examples in tables and figures).
  - In AIPW two-type model (Table D2, Panel B): Post-default, with sudden stops Sum = -10.70*** (h = 1..5: -4.072***, -3.607***, -2.156***, -0.434, -0.401).
  - Weakly preemptive, with sudden stops (AIPW): Sum = -18.04*** (h = 1..5: -3.059***, -4.517***, -2.861***, -3.504***, -4.084***).
  - Weakly preemptive, without sudden stops (AIPW): Sum = 7.278*** (h = 1..5: 1.596***, 2.342***, 2.134***, 1.216***, -0.427).
- OLS local projections (Panel C: GDP)
  - OLS estimates (Table D1, Panel C):
    - Post-default, with sudden stops: Sum = -23.63*** (h = 1..5: -3.496***, -4.940***, -5.669***, -5.061***, -5.156**).
    - Post-default, without sudden stops: Sum = -15.76*** (h = 1..5: -1.382, -2.804**, -3.261**, -4.393***, -4.509**).
    - Weakly preemptive, with sudden stops: Sum = -10.66 (h = 1..5: -3.024***, -3.809**, -1.787, -0.337, -1.676).
    - Weakly preemptive, without sudden stops: Sum = -20.54*** (h = 1..5: -1.572*, -2.064**, -3.519**, -5.762***, -7.853***).
  - AIPW one-type model (Figure D2 / Table D3 summary):
    - Post-default, with sudden stops (AIPW one-type, Panel C): Sum = -24.89*** (h = 1..5: -4.049***, -4.839***, -5.462***, -4.845***, -6.507***).
    - Post-default, without sudden stops (AIPW one-type, Panel C): Sum = -10.17*** (h = 1..5: -1.371***, -2.412***, -2.718***, -0.710, -3.727***).
- Panel OLS & IV (Table D4) — annualized growth around restructurings
  - Panel A: Annualized GDP growth (100×[ln(GDP_{i,t+2}) − ln(GDP_{i,t})]/2)
    - Post-default, d.NCI < 25th: coefficient = -3.302*** (columns (1)–(6) show robustness; e.g., column (1) (OLS) = -3.302*** (1.070) etc.).
    - Post-default, d.NCI ≥ 25th: coefficient = -1.183**.
    - Weakly preemptive, d.NCI < 25th: -2.500*.
    - Weakly preemptive, d.NCI ≥ 25th: 1.370*.
    - Difference between strategies (Post-default difference): -2.120* (SE = 1.185) in column (1); other columns report similar negative and often significant differences (e.g., column (5): -2.701**).
    - # of observations: across specifications 2,309; 1,671; 1,671; 2,245; 1,614; 1,614. # of countries: 64; 53; 53; 64; 52; 52.
  - Panel B: Annualized Investment change (100×[Invest_{i,t+2}/GDP_{i,t+2} − Invest_{i,t}/GDP_{i,t}]/2)
    - Post-default, d.NCI < 25th: -3.894***.
    - Post-default, d.NCI ≥ 25th: -1.244***.
    - Difference (Post-default): -2.650*** (column (1)); other columns show consistent negative and significant differences.
    - # of observations: e.g., 2,017; 1,657; 1,657; 1,955; 1,601; 1,601.
  - Panel C: Annualized rate of change in private credit (100×[ln(Credit_{i,t+2}) − ln(Credit_{i,t})]/2)
    - Post-default, d.NCI < 25th: -6.114**.
    - Post-default, d.NCI ≥ 25th: -5.546***.
    - Control variables and significance patterns reported in table (e.g., GDP growth rate, lag = 0.614*** in some specs).
    - R-squared up to 0.478 in reported specifications; # of obs. examples: 2,047; 1,561; 1,561; 1,974; 1,500; 1,500.
- General notes on estimation methods and sample
  - Local projections reported for h = 1,2,..,5 horizons.
  - Regressions include constant term, lagged dependent variable (lag 1), cyclical component of log of GDP (HP filter smoothing parameter = 100), and country fixed effects.
  - Robust standard errors clustered at country-level where applicable; ***, **, * denote 1 percent, 5 percent, 10 percent significance.
  - Definition: Severe sudden stops defined using percentage point difference in net capital inflows-to-GDP ratio from year 0 to year 1, d_{i,t+1}^{NCI} = 100 × (NCI_{i,t+1}/GDP_{i,t+1} − NCI_{i,t}/GDP_{i,t}), with cutoff at the 25th percentile among observations experiencing that type of restructurings (25th percentile cutoff for Panels B and C to have enough observations).

### D.2. Duration, Haircuts and Missed Payments — first stage, ROC, and GDP dynamics
- First-stage AIPW prediction performance (Table D5 and Figure D3)
  - First-stage regressions predict: GDP Forecasts, Expected Durations, Expected Haircuts.
  - Pseudo R-squared and fit:
    - Column (1) (GDP Forecasts): Pseudo R-sq. = 0.417; Area under ROC curve = 0.95.
    - Column (2) (Expected Durations): Pseudo R-sq. = 0.180; Area under ROC = 0.84.
    - Column (3) (Expected Haircuts): Pseudo R-sq. = 0.241; Area under ROC = 0.87.
    - Column (4): Pseudo R-sq. = 0.606; Area under ROC = 0.98.
    - Column (5): Pseudo R-sq. = 0.235; Area under ROC = 0.87.
    - Column (6): Pseudo R-sq. = 0.543; Area under ROC = 0.97.
  - Selected first-stage coefficient examples (preserved exactly from table):
    - GDP growth rate, lag: column (2) = -0.049***; column (4) = -0.145*; column (6) = -0.129**.
    - Cyclical component of ln(GDP per capita), lag: column (1) = -7.765*; column (4) = 36.12***; column (6) = 27.17***.
    - # of past post-default cases: column (1) = 1.769***; column (4) = 6.604***; column (6) = 5.964***.
    - # of past banking crises: column (1) = -1.271***; column (2) = -0.348***; column (6) = -1.339**.
    - Interest payments-to-GDP ratio, lag: column (1) = 58.14***; column (2) = 11.06***; column (3) = 12.47***.
  - Observations in first-stage regressions: examples include 456; 1,260; 1,276; 481; 1,235; 481. Log-likelihood ratio p-values = 0.000 for reported columns.
- GDP dynamics after post-default restructurings (Figures D4; Tables D6–D8)
  - Two-type AIPW results (Table D6: allowing first-stage parameters to differ)
    - Panel A (GDP Forecasts)
      - Post-default, with lower GDP forecasts: Sum = -22.85*** (h = 1..5 coefficients: -5.866***, -3.713***, -6.078***, -5.301***, -2.809***).
      - Post-default, with higher GDP forecasts: Sum = 2.826 (h = 1..5: -2.023***, -0.520, -0.833**, 4.202***, 1.693**).
    - Panel B (Duration)
      - Post-default, with longer expected duration: Sum = -26.09*** (h = 1..5: -6.656***, -5.093***, -6.432***, -5.908***, -3.058***).
      - Post-default, with shorter expected duration: Sum = 2.826 (same coefficient pattern as higher GDP forecasts in table excerpt).
    - Panel C (Haircuts)
      - Post-default, with higher expected haircuts: Sum = -34.08*** (h = 1..5: -7.207***, -7.652***, -7.357***, -7.519***, -5.292***).
      - Post-default, with lower expected haircuts: Sum = 2.620 (h = 1..5: -2.129***, 0.326, -1.799***, 3.155***, 2.554***).
  - One-type AIPW results (Table D7: assuming identical first-stage parameters)
    - Panel A (GDP Forecasts)
      - Post-default, with lower GDP forecasts: Sum = -30.70*** (h = 1..5: -4.144***, -4.254***, -7.297***, -8.446***, -7.700***).
      - Post-default, with higher GDP forecasts: Sum = 1.148 (h = 1..5: -1.366***, -2.249***, -1.458***, 2.532***, 2.925***).
    - Panel B (Duration)
      - Post-default, with longer expected duration: Sum = -34.50*** (h = 1..5: -4.972***, -5.313***, -8.048***, -9.144***, -8.039***).
      - Post-default, with shorter expected duration: Sum = 1.148 (same coefficients as higher forecasts group).
    - Panel C (Haircuts)
      - Post-default, with higher expected haircuts: Sum = -37.89*** (h = 1..5: -5.019***, -7.922***, -8.416***, -9.145***, -8.322***).
      - Post-default, with lower expected haircuts: Sum = 2.729** (h = 1..5: -1.366***, -0.777***, -1.327***, 2.404***, 2.970***).
  - OLS local projections (Table D8)
    - Panel A (GDP Forecasts)
      - Post-default, with lower GDP forecasts: Sum = -27.77** (h = 1..5: -4.661***, -4.161***, -6.214***, -6.845**, -6.417).
      - Post-default, with higher GDP forecasts: Sum = -2.496 (h = 1..5: -1.148**, -2.661**, -2.735, 1.216, 1.993).
    - Panel B (Duration)
      - Post-default, with longer expected duration: Sum = -27.77** (same as lower GDP forecasts above in OLS table).
      - Post-default, with shorter expected duration: Sum = -2.496 (same as higher GDP forecasts above).
    - Panel C (Haircuts)
      - Post-default, with higher expected haircuts: Sum = -29.31*** (h = 1..5: -4.671***, -5.589***, -6.378***, -6.763**, -6.404).
      - Post-default, with lower expected haircuts: Sum = -1.08 (h = 1..5: 1.139** , -1.348, -2.585, 1.139, 1.98).
- Notes on interpretation across D.2
  - Classification into high/low GDP forecasts, long/short expected durations, and high/low expected haircuts is based on first-stage estimates (columns in Table D5).
  - AIPW estimates presented under both two-type and one-type first-stage parameter assumptions; figures display point estimates and 95 percent confidence intervals for horizons h = 1..5 over sample period 1970–2013.
  - Across methods, post-default restructurings associated with severe sudden stops, lower GDP forecasts, longer expected durations, and higher expected haircuts tend to show larger negative GDP and credit dynamics (statistically significant negative sums reported above).

*Source: Appendix D. Further Regression Results — wpiea2019069 (selected tables and figures).*

### Appendix E. Robustness Checks

### Appendix E. Robustness Checks (wpiea2019069)

### E.1. Large Sample Including Countries without Restructurings
- Sample construction and scope:
  - Excluded high income countries where PPP-adjusted GDP per capita is higher than the 80th percentile of the entire sample in 2000, leaving 122 countries.
  - Sample period: 1970–2013.
  - Because the expanded sample includes countries with restructuring dummies equal to 0, estimation is by OLS (local projections) rather than AIPW.
- Main OLS findings (local projections for h = 1,2,..,5; regressions include constant, lagged dependent variables (lag 1 and lag 2), cyclical component of log GDP from HP filter with smoothing parameter 100, and country fixed effects; robust SEs clustered at country level):
  - Panel A (GDP): Post-default coefficients:
    - h = 1: -1.978***
    - h = 2: -3.181***
    - h = 3: -3.337***
    - h = 4: -2.859**
    - h = 5: -2.285
    - Sum: -13.22***
  - Panel B (Investment): Post-default coefficients (reported block with Sum):
    - h = 1: -1.806**
    - h = 2: -3.179***
    - h = 3: -1.920**
    - h = 4: -1.455
    - h = 5: -2.160**
    - Sum: -10.30***
  - Panel C (Private credit): Post-default coefficients:
    - h = 1: -0.031
    - h = 2: -1.943***
    - h = 3: -2.778***
    - h = 4: -3.218***
    - h = 5: -3.788***
    - Sum: -11.97***
  - Panel D (Lending rates): Post-default coefficients:
    - h = 1: 4.54**
    - h = 2: 10.55**
    - h = 3: 10.34***
    - h = 4: 9.88**
    - h = 5: 12.09**
    - Sum: 48.77***
  - Panel E (Net capital inflows as percent of GDP): Post-default coefficients:
    - h = 1: -2.458**
    - h = 2: -4.784***
    - h = 3: -0.258
    - h = 4: -2.029
    - h = 5: -3.754***
    - Sum: -14.16***
- Robustness conclusion: Results are quantitatively similar to baseline AIPW estimates (Figure 7 and Table C1), confirming robustness.

*Source: Appendix E. Robustness Checks, wpiea2019069 (sample period 1970–2013).*

### E.2. Standard & Poor’s Default Episodes
- Data and comparability:
  - S&P default data cover only 88 “default” episodes in countries where ratings are assigned (S&P does not differentiate restructuring strategies; S&P covers (i) post-default restructurings and (ii) exchange events for preemptive restructurings).
  - S&P default sample is similar to post-default restructurings in the main sample.
  - Sample period: 1970–2013.
- First-stage (probit) and instrument performance:
  - Table E2 reports probit results predicting start of S&P defaults using past restructuring counts, past banking crises, external debt and interest payment ratios, change in credit ratings, cyclical component of log GDP per capita, GDP growth rate (lag), institutional indices, years since previous debt restructuring, and country fixed effects.
  - Reported model statistics (selected):
    - # of observations: 2,068; 815; 415 (across columns)
    - # of countries: 74; 39; 39
    - # of S&P default episodes: 46; 29; 25
    - Likelihood ratio (p-value): 34.53 (0.953); 59.599 (0.014); 91.03 (0.000)
    - Pseudo R-squared: 0.054; 0.190; 0.529
    - ROC: 0.70; 0.87; 0.96
- AIPW estimates for S&P defaults (local projections, controls as in Table E3 notes):
  - Panel A (GDP), S&P default with x1:
    - h = 1: -2.859***
    - h = 2: -5.453***
    - h = 3: -1.244*
    - h = 4: 0.192
    - h = 5: -0.670
    - Sum: -9.216***
    - # of observations: 372 (h = 1..4), 359 (h = 5)
    - # of countries: 16
    - # of S&P defaults: 21
  - Panel A (GDP), S&P default with x2:
    - h = 1: -3.383***
    - h = 2: -4.366***
    - h = 3: -3.634***
    - h = 4: -1.950
    - h = 5: -4.057**
    - Sum: -16.68***
    - # of observations: 786 (h = 1..4), 758 (h = 5)
    - # of countries: 28
    - # of S&P defaults: 38
  - Panel B (Investment) (selected):
    - With x1 Sum: -15.65***
    - With x2 Sum: -17.31***
  - Panel C (Credit) (selected):
    - With x1 Sum: -15.37***
    - With x2 Sum: -6.474***
  - Panel D (Net capital inflows) (selected):
    - With x1 Sum: -15.37***
    - With x2 Sum: -6.474***
- Robustness conclusion: AIPW estimates for GDP, investment, private credit, and net capital inflows in S&P defaults largely mirror those in post-default restructurings.

*Source: Appendix E. Robustness Checks, wpiea2019069 (sample period 1970–2013).*

### E.3. Restructurings with IMF-supported Programs and Paris Club Restructurings
- Motivation: Official financing (IMF-supported programs) and official debt treatment (Paris Club deals) may mitigate adverse effects on GDP and investment via a capital inflow → credit → investment channel.
- Estimation approach: Panel OLS regressions to capture influence of IMF-supported program or official debt restructuring over the forecast horizon (h = 1,2,..,5). Regressions control as in other panels (lagged dependent, HP-filter cyclical component, country fixed effects).
- Main finding:
  - Declines in GDP and investment following a post-default restructuring or a weakly preemptive restructuring are significantly moderated by official financing or official debt treatment (effects shown in Figure E4: GDP and Investment panels).
- Presentation details:
  - Point estimates shown for private debt restructurings with and without IMF-supported programs or official (Paris Club) debt restructurings; 95 percent confidence intervals reported.
  - Sample period: 1970–2013.

*Source: Appendix E. Robustness Checks, wpiea2019069 (sample period 1970–2013).*

### E.4. “Single” External Debt Restructurings (external only vs. domestic+external)
- Definitions:
  - "Single" external debt restructuring: no domestic debt restructuring occurs during the duration of an external debt restructuring.
  - "Dual" domestic-external debt restructuring: domestic debt restructuring occurs during the duration of an external debt restructuring.
  - Domestic debt restructuring dataset based on jurisdiction and creditor residence criterion from Asonuma and Papaioannou (2016); alternative definitions noted (currency denomination criteria).
- Summary counts and shares:
  - All debt restructuring episodes: 179
    - Post-default: 111
    - Weakly preemptive: 45
    - Strictly preemptive: 23
  - "Single" external debt restructurings (no domestic debt restructuring during external restructure): 156 total
    - Post-default: 95
    - Weakly preemptive: 41
    - Strictly preemptive: 20
  - Share of single external debt restructurings:
    - All: 87%
    - Post-default: 86%
    - Weakly preemptive: 91%
    - Strictly preemptive: 87%
- AIPW estimation results (Figure E5 and Table E5; regressions include constant, lagged dependent variable, HP-filter cyclical component, country fixed effects):
  - Panel A (GDP), Post-default (single external):
    - h = 1: -2.381***
    - h = 2: -2.889***
    - h = 3: -3.581***
    - h = 4: -1.813***
    - h = 5: -4.763***
    - Sum: -14.70***
    - # of observations: 825 (h = 1..4), 796 (h = 5)
    - # of countries: 29
    - # of debt restructuring obs.: 33
  - Panel A (GDP), Weakly preemptive (single external):
    - h = 1: -2.332***
    - h = 2: -2.543***
    - h = 3: -1.673***
    - h = 4: -2.824***
    - h = 5: -4.199***
    - Sum: -13.21***
    - # of observations: 371 (h = 1..4), 357 (h = 5)
    - # of countries: 14
    - # of debt restructuring obs.: 20
  - Panel B (Investment), Post-default (single external):
    - h = 1: -1.441***
    - h = 2: -1.828***
    - h = 3: -0.322
    - h = 4: 0.477
    - h = 5: 0.085
    - Sum: -2.884
  - Panel B (Investment), Weakly preemptive (single external):
    - h = 1: -0.046
    - h = 2: -0.235
    - h = 3: 0.328
    - h = 4: -0.605
    - h = 5: -1.813***
    - Sum: -2.062
  - Panel C (Credit), Post-default (single external):
    - h = 1: -0.151
    - h = 2: -1.542***
    - h = 3: -1.829***
    - h = 4: -2.060***
    - h = 5: -1.979***
    - Sum: -7.804***
  - Panel C (Credit), Weakly preemptive (single external):
    - h = 1: 0.244
    - h = 2: -0.990***
    - h = 3: -1.468***
    - h = 4: -1.627***
    - h = 5: -2.225***
    - Sum: -6.590***
  - Panel D (Net capital inflows), selected:
    - Post-default Sum: -7.088***
    - Weakly preemptive Sum: -5.635***
- Robustness conclusion:
  - Estimated impacts on GDP, investment, private credit, and net capital inflows for all external debt restructurings are quite similar and largely overlapping with those for single external debt restructurings.
  - This is consistent with single external debt restructurings comprising 87% of total external debt restructuring episodes.
  - Large costs of restructurings are dominantly associated with restructuring only external debt.

*Source: Appendix E. Robustness Checks, wpiea2019069 (sample period 1970–2013).*

### E.5. Non-overlapping and Sequential Restructurings
- Motivation: Address potential influence from overlapping restructuring durations.
- Alternative classifications considered:
  1. Individual restructurings without overlapping years: when durations overlap, duration of previous restructuring ends the year before the next restructuring starts; the next restructuring retains its defined duration.
  2. Sequential restructurings with the initial strategy: overlapping restructurings are combined into one sequential restructuring coded by the strategy of the initial restructuring.
  3. Sequential restructurings with the worst strategy: overlapping restructurings are combined into one sequential restructuring coded by the worst strategy among overlapping individual restructurings (e.g., weakly preemptive followed by post-default → coded as weakly preemptive under classification 2, coded as post-default under classification 3).
- AIPW results (Figure E6):
  - GDP dynamics under each alternative classification (individual non-overlapping, sequential initial strategy, sequential worst strategy) show similar patterns after each restructuring type.
  - Conclusion: Baseline AIPW estimates (as reported in Figure 7) are robust to alternative handling of overlapping durations.

*Source: Appendix E. Robustness Checks, wpiea2019069 (sample period 1970–2013).*

*Source: Appendix E. Robustness Checks (wpiea2019069), IMF working paper appendix.*

### Part I: GDP under Individual Restructurings without Overlapping Years

### Part I: GDP under Individual Restructurings without Overlapping Years

### Enlarged Set of Instruments for AIPW Estimation
- Motivation: Address concerns on influence by domestic political system, banking sector, past banking crises, government behavior, and pre-crisis GDP growth prediction.
- Augmented instruments include baseline selected instruments plus:
  - (i) years since the last restructurings (for any types of debt restructurings),
  - (ii) legislative and executive indices of electoral competitiveness (LIEC and EIEC from the Database of Political Institutions),
  - (iii) government coerciveness indicator during past restructurings (Enderlein et al., 2012).
- LIEC coding (excerpt): 1 if there is no legislature; 2 if unelected legislature; 3 if elected from one candidate; 4 if 1 party, multiple candidates; 5 if multiple parties are legal but only one party won seats; 6 if multiple parties win seats but the largest party received more than 75 percent of the seats; 7 if the largest party got less than 75 percent.
- EIEC coding (excerpt): 1 if rival chief executives in one country, particularly in the setting of armed conflicts, are counted as no executives; 2 if executives elected by small appointed juntas or by appointed electoral colleges; 3 if the chief executives of communist nations (the chairman of the Communist Party); 6 or 7 if competitively elected prime ministers.
- Limitation: Enlarged instrument set reduces the number of restructuring observations (samples substantially smaller than baseline).

### First-Stage Classification Power (ROC)
- ROC interpretation: takes a value between 0.50 and 1; 0.50 indicates no classification power, 1 indicates perfect classification power.
- ROC results (improvement with enlarged instruments):
  - Post-default case: ROC increases from 0.870 to 0.963.
  - Weakly preemptive case: ROC increases from 0.953 to 0.993.
  - “Weakly and strictly” preemptive case: ROC increases from 0.911 to 0.949.
- Trade-off: Improved classification power vs reduced sample of restructuring observations (baseline vs robustness analysis):
  - Baseline samples: 51 and 29 episodes for post-default and weakly preemptive cases.
  - Robustness analysis samples: 25 and 21 episodes for post-default and weakly preemptive cases.

### Predicting Start of Debt Restructurings (Probit) — Selected Coefficient Highlights (Table E6)
- Dependent variable: Dummy taking 1 at the start of debt restructurings. All regressions include country fixed effects. Sample period: 1970–2013.
- Selected coefficients (standard errors in parentheses):
  - # of past post-default: 0.755*** (0.182); 1.818*** (0.439); -1.557** (0.608); -7.329 (665.9); -0.619* (0.358); -1.246* (0.732).
  - # of past weakly preemptive: 0.101 (0.268); 0.582 (0.432); 0.502* (0.263); 5.418*** (1.724); 0.067 (0.221); -0.356 (0.311).
  - # of past strictly preemptive: -0.593 (0.605); -10.21 (802.3); -1.474* (0.800); -2.679 (2.528); 0.585 (0.448); 1.465* (0.860).
  - # of past banking crises: -0.376*** (0.084); -0.255 (0.172); -0.319 (0.195); 0.036 (0.699); -0.241* (0.142); -0.281 (0.281).
  - External debt-to-GDP ratio, lag: -0.926** (0.416); -5.312*** (1.734); 0.715 (1.386); 9.239* (4.824); -1.069 (1.097); 0.368 (1.763).
  - Total interest payments-to-GDP ratio, lag: 17.53*** (6.216); 40.17*** (13.02); 51.33*** (17.68); -112.14 (80.37); 0.18*** (14.27); 26.27 (21.86).
  - GDP growth rate, lag: -0.0370 (0.018); -0.0719** (0.034); 0.028 (0.034); -0.192* (0.105); -0.005 (0.027); -0.036 (0.042).
  - Freedom index, civil liberty, lag: 0.275** (0.110); 0.314 (0.253); -0.345 (0.281); -1.420 (1.044); -0.166 (0.203); -0.315 (0.291).
- Additional instruments coefficients (selected):
  - # of years from the previous debt restuct.: -0.211*** (0.064); -2.595*** (0.748); -0.083 (0.072).
  - Legislative index of electoral comp., lag: 0.280 (0.253); 0.881 (1.283); -0.172 (0.310).
  - Executive index of electoral comp., lag: -0.264 (0.280); 0.018 (0.669); 0.508* (0.272).
  - Coerciveness index, lag: -0.170* (0.087); 0.075 (0.296); -0.049 (0.116).
- Model fit and sample indicators (selected columns):
  - # of observations across columns: 825, 408, 400, 298, 467, 340.
  - # of countries: 49, 20, 49, 20, 49, 20 (as presented).
  - Log likelihood ratio (p-value): 87.21 (0.000); 109.62 (0.000); 88.92 (0.000); 118.83 (0.000); 84.37 (0.000); 89.00 (0.000).
  - Pseudo R-squared: 0.25; 0.54; 0.46; 0.78; 0.35; 0.50.
  - ROC reported: 0.86; 0.96; 0.95; 0.99; 0.92; 0.95.

### Estimation Results with Enlarged Set of Instruments (AIPW) — GDP, Investment, Credit, Net Capital Inflows (Table E7 and Figure E8)
- General result: Baseline AIPW estimates are robust to enlarged instruments. Post-default restructurings experience more severe and prolonged declines in GDP, investment, private credit, and net capital inflows than weakly preemptive restructurings.
- Panel A: Dep. Var. = 100 × (GDP_{i,t+h} − GDP_{i,t})/GDP_{i,t}. Local projections (h = 1..5). Selected point estimates and significance:
  - Post-default (h = 1 to 5; Sum):
    - h = 1: -4.371*** (0.561)
    - h = 2: -8.054*** (0.810)
    - h = 3: -5.653*** (0.560)
    - h = 4: -4.242*** (0.627)
    - h = 5: -4.502*** (0.725)
    - Sum: -26.28*** (2.858)
    - # of obs.: 400, 400, 400, 400, 386, 386
    - # of countries: 16 (across horizons)
    - # of debt restructuring obs.: 28 (across horizons)
  - Weakly preemptive:
    - h = 1: -1.503** (0.762)
    - h = 2: -1.600* (0.925)
    - h = 3: -0.523 (1.026)
    - h = 4: -3.118*** (1.023)
    - h = 5: -2.801** (1.206)
    - Sum: -9.544*** (3.428)
    - # of obs.: 96 (across horizons)
    - # of countries: 11
    - # of debt restructuring obs.: 21
  - Weakly & strictly preemptive:
    - h = 1: -2.646*** (0.304)
    - h = 2: -1.291*** (0.456)
    - h = 3: 1.127 (0.744)
    - h = 4: -0.253 (1.045)
    - h = 5: -2.997*** (1.053)
    - Sum: -5.861* (3.111)
    - # of obs.: 340 (across horizons)
    - # of countries: 13
    - # of debt restructuring obs.: 25
- Panel B: Investment (Dep. Var. = 100 × (Investment/GDP_{t+h} − Investment/GDP_{t})). Selected estimates:
  - Post-default:
    - h = 1: -0.891** (0.417)
    - h = 2: -1.144*** (0.443)
    - h = 3: -0.334 (0.414)
    - h = 4: 1.088** (0.465)
    - h = 5: 0.775 (0.622)
    - Sum: -0.240 (2.256)
    - # of obs.: 397, 397, 397, 397, 383, 383
    - # of debt restructuring obs.: 27 (reported)
  - Weakly preemptive:
    - h = 1: 1.103 (0.983)
    - h = 2: 1.081 (0.983)
    - h = 3: 1.330 (1.078)
    - h = 4: -1.258** (0.596)
    - h = 5: -2.167*** (0.699)
    - Sum: 0.089 (3.247)
    - # of obs.: 92 (across horizons reported)
  - Weakly & strictly preemptive:
    - h = 1: 0.613 (0.392)
    - h = 2: 1.157 (0.464)
    - h = 3: 1.226*** (0.423)
    - h = 4: 1.271** (0.516)
    - h = 5: 1.008* (0.609)
    - Sum: 5.456** (2.159)
    - # of obs.: 340 (across horizons)
- Panel C: Credit (Dep. Var. = 100 × (Credit/GDP_{t+h} − Credit/GDP_{t})). Selected estimates (from Figure E8 table excerpt):
  - Post-default (h = 1..5; Sum):
    - h = 1: -0.556*** (0.222)
    - h = 2: -4.184*** (0.389)
    - h = 3: -3.757*** (0.404)
    - h = 4: -3.552*** (0.547)
    - h = 5: -4.679*** (0.558)
    - Sum: -16.85*** (1.839)
    - # of obs.: 378, 378, 378, 378, 365, 365
    - # of countries: 15
    - # of debt restructuring obs.: 27
  - Weakly & strictly preemptive:
    - h = 1: -0.182 (0.188)
    - h = 2: -1.614*** (0.287)
    - h = 3: -1.505*** (0.379)
    - h = 4: -1.545*** (0.534)
    - h = 5: -1.510** (0.627)
    - Sum: -6.512*** (1.816)
    - # of obs.: 340
    - # of debt restructuring obs.: 25
- Panel D: Net Capital Inflows (NCI) (Dep. Var. = 100 × (NCI/GDP_{t+h} − NCI/GDP_{t})). Selected estimates:
  - Post-default:
    - h = 1: -5.067*** (1.047)
    - h = 2: -7.391*** (1.493)
    - h = 3: 0.1705 (1.111)
    - h = 4: 2.310** (1.091)
    - h = 5: -1.533 (1.020)
    - Sum: -10.50** (4.790)
    - # of obs.: 393, 392, 392, 392, 378, 375
    - # of countries: 16
    - # of debt restructuring obs.: 28 (varies by horizon)
  - Weakly preemptive:
    - h = 1: 0.174 (1.164)
    - h = 2: -0.138 (1.323)
    - h = 3: -2.467* (1.369)
    - h = 4: -5.727*** (1.486)
    - h = 5: 0.425 (1.587)
    - Sum: -6.188 (4.907)
    - # of debt restructuring obs.: 21
  - Weakly & strictly preemptive:
    - h = 1: 0.238 (0.782)
    - h = 2: 3.822*** (0.927)
    - h = 3: 3.180*** (0.867)
    - h = 4: 2.527** (1.010)
    - h = 5: 5.481*** (1.167)
    - Sum: 16.39*** (4.054)
    - # of obs.: 336, 335, 335, 335, 322, 319
    - # of debt restructuring obs.: 25

### Robustness Check: Controlling for GDP Forecasts (IMF WEO, 1990–2013)
- Method: Include GDP growth rate forecasts (g_{i,t}^{Forecast}) as an additional instrument in both first-stage and second-stage regressions. Forecast variable constructed as g_{i,t}^{Forecast} = (∑_{s=1}^{5} g_{i,t}^{Forecast,t+s})/5 (average of GDP growth rate forecasts over next five years in pre-restructuring year t).
- Sample reduction: GDP forecasts available 1990–2013, reducing sample of restructurings significantly.
- Comparative estimation: Baseline specification (without forecasts) also estimated on same reduced sample for comparability.

Selected results (Table E8 and Figure E9) — GDP (Dep. Var. = 100 × (GDP_{i,t+h} − GDP_{i,t})/GDP_{i,t}), h = 1..5:
- Post-default, with GDP forecasts:
  - h = 1: -5.858*** (0.606)
  - h = 2: -7.205*** (0.654)
  - h = 3: -6.374*** (0.683)
  - h = 4: -5.685*** (0.877)
  - h = 5: -4.382*** (0.863)
  - Sum: -28.26*** (3.304)
  - # of obs.: 145 (horizons 1–5), 138 (h = 5 sums)
  - # of countries: 8 (across reported rows)
  - # of debt restructuring obs.: 9 (reported across rows)
- Post-default, without GDP forecasts (same sample):
  - h = 1: -4.036*** (0.966)
  - h = 2: -5.350*** (1.071)
  - h = 3: -4.606*** (1.015)
  - h = 4: -3.349*** (1.193)
  - h = 5: -2.694*** (1.180)
  - Sum: -18.51*** (4.877)
  - # of obs.: 145 (horizons 1–5), 138 (h = 5 sums)
  - # of countries: 8
  - # of debt restructuring obs.: 9
- Conclusion: Baseline results for post-default restructurings remain robust when controlling for GDP forecasts.

### Consumer Price Index (CPI) Inflation Rates (E.7)
- Data: CPI inflation from WDI. Hyperinflation defined as year-to-year inflation rate > 500 percent (following Reinhart and Rogoff (2011b)); hyperinflation episodes reported in Table E9.
- Dynamics (Figure E10 summary):
  - Without hyperinflations:
    - Post-default restructurings: CPI inflation experiences a sharp rise following a post-default restructuring (red line in Panel B).
    - Weakly preemptive restructurings: Moderate rise in CPI inflation in the run-up (blue line in Panel B).
    - Strictly preemptive restructurings: Smaller and short-lived rise followed by recovery to pre-crisis level (green line in Panel B).
  - Across all three strategies, CPI inflation adjustments are smaller than those in domestic default episodes (per Reinhart and Rogoff (2009)).
- Cumulative responses (Panel C, non-hyperinflation episodes): Post-default restructurings experience a sharper rise in CPI inflation than weakly or strictly preemptive restructurings.

### Average CPI Inflation Rates (Table E9) — Selected cells
- Post-default Restructurings (Panel A):
  - Full Sample, h = 0..5 average inflation rates: 34.01, 64.81, 162.66, 110.31, 207.24, 185.37.
  - Full Sample standard deviations, h = 0..5: 80.75, 188.31, 523.91, 313.47, 1159.91, 677.25.
  - Sample sizes, h = 0..5: 71, 71, 73, 77, 79, 79.
  - Hyperinflation Sample averages (h = 0..5): 629.11, 1027.86, 1752.87, 985.29, 2788.20, 1865.94.
  - Non-Hyperinflation Sample averages (h = 0..5): 25.51, 36.90, 45.73, 36.37, 32.85, 21.99.
- Weakly Preemptive Restructurings (Panel B):
  - Full Sample averages, h = 0..5: 49.56, 54.16, 42.60, 43.22, 46.85, 133.40.
  - Full Sample standard deviations, h = 0..5: 105.25, 113.73, 55.10, 59.22, 76.79, 509.56.
  - Sample size across horizons: 37 (consistent).
  - Hyperinflation Sample averages (h = 0..5): 626.72, 672.18, NA, NA, NA, 1854.46 (sample size varying: 1,1,0,0,0,2).
  - Non-Hyperinflation Sample averages (h = 0..5): 33.53, 36.99, 42.60, 43.22, 46.85, 35.06.
- Strictly Preemptive Restructurings (Panel C):
  - Full Sample averages, h = 0..5: 27.07, 32.65, 32.27, 28.11, 33.94, 65.32.
  - Full Sample standard deviations, h = 0..5: 23.68, 30.87, 30.85, 26.39, 50.13, 150.39.
  - Sample size across horizons: 21 (consistent).
  - Hyperinflation Sample: NA for most horizons; only h = 5 shows 681.42 with sample size 1.
  - Non-Hyperinflation Sample averages, h = 0..5: 27.07, 32.65, 32.27, 28.11, 33.94, 34.51.

*Source: IMF Working Paper content (Part I: GDP under Individual Restructurings without Overlapping Years).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019069.pdf_