## wp17114 - References

## Source details

**Canonical URL:** [wp17114 - References](https://www.imf.org/-/media/files/publications/wp/2017/wp17114.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2017/wp17114.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2017/wp17114.pdf.json)

---

### Context and motivation
- Despite waves of debt relief over more than three decades, many low-income countries (LICs) experienced prolonged periods of debt distress linked to sovereign debt crises beginning in the late 1970s and 1980s.
- Evolution of official creditor responses:
  - From 1976 to 1988 the Paris Club agreed upon 81 nonconcessional flow reschedulings with 27 of the countries now identified as HIPCs.
  - Toronto (1988), London (1990), and Naples (1994) terms reduced net present value of debt by 33, 50, and 67 percent respectively.
  - In 1996, the Heavily Indebted Poor Countries (HIPC) Initiative included debt relief on amounts owed to multilateral institutions.
  - PC creditors granted Lyon (1996) and Cologne (1999) terms (with 80 and 90 percent reduction in net present value of debt respectively) in the context of concerted action under the HIPC initiative.
- During 2000-2015, of the 39 countries eligible or potentially eligible for HIPC Initiative assistance, 36 received full and irrevocable reductions in their external debt.
- After HIPC debt relief, public debt ratios declined considerably in LICs (IMF (2014) cited).

### Research objective and contribution
- Objective: identify determinants of debt distress in LICs beyond traditional debt burden and liquidity indicators by examining factors closely associated with long-term growth performance: economic institutions, macroeconomic policies, and vulnerability to external shocks.
- Key contributions:
  - Disentangles impact of highly persistent factors (quality of institutions, policy record, export concentration, exposure to external shocks) from transitory factors (contemporaneous external environment, recent policies, debt burden, liquidity indicators).
  - Uses a wide set of covariates for more country-specific and granular analysis of debt sustainability with higher predictive power than earlier studies.
  - Models unobserved country heterogeneity and its potential correlation with covariates to obtain consistent and efficient coefficient estimates.
  - First to examine impact of common exogenous shocks originating from commodity prices on LIC debt distress clustering.

### Main findings (empirical insights)
- Determinants and transmission channels:
  - Protection of private property rights (a highly persistent institutional variable) is among the most influential factors explaining the likelihood of debt distress.
  - LICs experiencing debt distress tend to have weaker macroeconomic policies and higher exposure to adverse external shocks over the sample period compared to LICs without debt distress.
  - Common, relatively persistent, adverse shocks to real non-oil commodity prices could lead to clustering of LIC debt distress episodes, conditional on initial debt burdens, policies, and institutions.
- Comparison with Kraay and Nehru (2006, KN):
  - KN used CPIA, real GDP growth, and an indicator of debt burden; CPIA as a composite prevents separating policy and institutional quality.
  - This study measures institutions via private property rights to identify their separate role.
  - KN found growth effects significant only for mixed samples; this study finds specific roles for commodity-price-related common shocks.
- Methodological improvements over KN:
  - Models unobserved country heterogeneity and its correlation with covariates.
  - Extends set of LICs and refines definition of debt distress episodes (drops some episodes identified solely by EFF).
  - Allows overlapping five-year “normal” episodes and models correlation in error terms from unobserved country effects.

### Hypotheses tested
- Hypothesis 1:
  - H.1.1 Structural variables determined by institutions are highly influential in predicting debt distress in LICs.
  - H.1.2 Weak economic institutions lead to higher probability of debt distress through persistently poor policies and high vulnerability to adverse external shocks.
  - Private property rights used as core measure of institutional quality.
- Hypothesis 2:
  - Adverse external shocks raise default probability in LICs short-term and long-term; persistent adverse global shocks could lead to clustering of debt defaults.
- Hypothesis 3:
  - Liquidity conditions impact timing of default; after controlling for institutions, shocks, and policies, liquidity conditions have a marginal effect and mainly predict timing for countries with weak fundamentals.

### Identification and empirical methodology
- Dependent variable: panel dummy = 1 for the start year of a debt distress episode and 0 for a normal episode (initial year of five consecutive years without debt distress).
  - Normal episodes defined as five years distress-free; three-year robustness check yields qualitatively similar results.
- Debt distress defined if any of:
  - (i) sum of interest and principal arrears exceeds five percent of total long term PPG external debt;
  - (ii) Paris Club rescheduling and/or debt reduction granted: year of debt relief and two subsequent years classified as distress;
  - (iii) IMF lending for immediate balance of payments needs exceeds 80 percent of the quota.
- IMF financing definition: actual purchases/drawn down under SBAs, CFF, and Oil Facility included; EFF drawings excluded.
- Country coverage and sample:
  - Broad sample: forty-two debt distress episodes for forty countries.
  - Estimation sample: thirty-one distress episodes with explanatory variables available.
  - Average duration of distress episodes: slightly longer than 20 years.
- Explanatory variable blocks: institutional quality, exogenous shocks, macroeconomic policies, debt burden, liquidity.
  - Debt burden measures: PV of PPG external debt scaled by GDP and by exports; PPG external debt service scaled by exports.
  - Institutional measure: private property rights index; alternative: weighted average of ICRG components.
  - Liquidity measures: reserve coverage (months of imports) and net resource transfers (nrt) with GDF formula: nrt = D + grant + port + fdi + profit + short - P - I.
  - Exogenous shocks: terms of trade shocks, real export growth, natural disaster damage indicator; time-specific: real oil and non-oil commodity prices, real U.S. interest rates, cyclical world trade; share of exports of primary commodities tested.
- Econometric approach:
  - Binary response models for panel data (probit class): pooled probit, random effects (RE) probit, fixed effects (FE) probit discussed, Chamberlain’s correlated RE (includes time averages of covariates), Chamberlain’s probit, GEE with exchangeable correlation.
  - Six estimators used: pooled probit, Chamberlain’s probit, RE probit, Chamberlain’s RE probit, GEE, Chamberlain’s GEE. Choice based on tests of ρ and time averages significance.
  - Marginal effect notation preserved.

### Main estimation results (benchmark specifications and robustness)
- Three models estimated using alternative debt burden indicators:
  - Model I: PV of PPG external debt to GDP.
  - Model II: PV of PPG external debt to exports.
  - Model III: PPG external debt service to exports.
- Two benchmark specifications for each model:
  - Benchmark specification 1: includes protection of private property rights (smaller sample).
  - Benchmark specification 2: excludes private property rights and uses country-specific averages to capture institutional effects indirectly (larger sample).
- Sample and estimation notes:
  - Overall sample: fifty-seven PRGF-eligible countries, period 1975–2010; estimation restricted to 1975–2006.
  - Example estimation sizes:
    - # of observations = 257 (benchmark 1 Model I).
    - # of observations = 388 (benchmark 2 Model I).
    - # of countries = 43 (benchmark 1 Model I).
    - # of countries = 53 (benchmark 2 Model I).
  - Sample probability of default: 0.12 (benchmark 1 Model I), 0.10 (benchmark 2 Model I).
- Key statistically significant determinants (across models and specifications):
  - Institutional quality:
    - Private property rights significant and negative where included (example coefficient: -0.583***(0.208)).
  - External shocks:
    - Cumulative three year change in real non-oil commodity prices significant and negative (example coefficient: -0.0361***(0.00780)).
    - Cumulative three year change in terms of trade often negative and significant (example: -0.0134**(0.00571)).
    - Cumulative three year change in real oil prices: for oil exporters negative and significant (example: -0.159**(0.0742)); for oil importers positive/insignificant in some specs.
  - Debt burden:
    - Higher PV of PPG external debt to GDP increases probability of default (example present value coefficient: 0.0277***(0.0105)).
    - PV of PPG external debt to exports and debt service to exports also positively significant.
  - Policy and liquidity:
    - Reserve coverage in months of imports significant and negative (example reserve coverage (t-1): -0.178**(0.0850)).
    - Current account balance to GDP (t-1) negative and significant in several specs (example: -0.0455**(0.0177)).
    - Macroeconomic (in)stability indicator significant where property rights excluded (example country-specific macro instability 0.0935**(0.0426)).
  - Liquidity via creditors/donors:
    - Net resource transfers and its components generally not significant.
- Magnitudes and relative influence:
  - Benchmark 1 marginal effects highlight most influential variables: protection of private property rights, real non-oil commodity prices, and the debt burden.
  - Benchmark 2 marginal effects: sample averages of current account balance and reserve coverage, followed by real non-oil commodity prices, are most influential.
  - Simulations illustrate predicted probabilities for median LIC versus LIC with weak fundamentals; predicted probability for debt distress is zero for the median LIC under specified initial values (simulation discussion continued).

### Goodness of fit and predictive performance
- Pseudo R2 for Model I reported values: 0.52, 0.38, 0.40, 0.40, 0.40 indicating about 40 to 50 percent of uncertainty explained by benchmark specifications.
- Likelihood ratio tests significantly reject joint null that explanatory coefficients are zero.
- Predicted probabilities distinguish crisis events from normal episodes:
  - Example Model I benchmark specification 1:
    - Median predicted probability for crisis events = 0.61; for normal episodes = 0.01.
    - 75th percentile for crisis events = 0.88; 75th percentile for normal episodes = 0.08.
- Overall: models with different debt burden indicators differentiate well between distress and normal episode distributions.

### Threshold probability analysis and forecasting framework
- Threshold probabilities derived by minimizing asymmetrically-weighted and equally-weighted loss functions (weights for type I and II errors: 0.65 and 0.35 respectively).
- Selected threshold and error figures (Model I–III summaries from Table 9):
  - Model I (PV PPG external debt/GDP), Benchmark 1:
    - Primary threshold = 0.14; Secondary threshold = 0.14.
    - Missed crises (Type I) = 0.097; False alarms (Type II) = 0.133; Overall error = 0.128.
  - Model I, Benchmark 2:
    - Primary threshold = 0.04; Secondary threshold = 0.137.
    - Missed crises = 0.026; False alarms = 0.332; Overall error = 0.302.
  - Model II Benchmark 1:
    - Primary threshold = 0.23; Missed crises = 0.133; False alarms = 0.071; Overall error = 0.078.
  - Model III Benchmark 1:
    - Primary threshold = 0.06; Missed crises = 0.034; False alarms = 0.230; Overall error = 0.208.
- Interpretation:
  - Predicted probabilities > threshold → classify country as moderate to high likelihood of debt distress.
  - Secondary (higher) threshold → classify as high risk of debt distress.
  - Benchmark specification 1 (including private property rights) generally has lower overall error than specification 2.
- Comparison to KN-based DSF: models in this study have better predictive power versus corresponding KN regressions.

### Counterfactual simulations: setup and key parameter values
- Simulations are counterfactual: all other variables remain unchanged while a single variable is allowed to change over its entire sample range.
- Exact values used in counterfactuals:
  - Present value of PPG external debt to GDP (%) — 23.8; 34.8
  - Reserve coverage in months of imports — 2.8; 1.6
  - Cumulative three year change in terms of trade — 0.8; -9.3
  - Cumulative three year change in real non-oil commodity prices — 0.0; 0.0
  - Cumulative three year change in real oil prices for oil importers — 0.0; 0.0
  - Private property rights — 6.3; 5.2
  - Current account balance to GDP (%) — -2.4; -7.0
  - Share of primary commodities in total exports (%) — 69.5; 92.6
  - Threshold probability — 0.14
  - Predicted probability — 0.007; 0.267
  - 95% confidence band — [0.001 0.033]; [0.143 0.465]
- Note: Threshold and predicted probabilities are reported for benchmark specification 1 for model I with present value of PPG external debt to GDP (%).

### Main simulation findings (marginal effects and heterogeneity)
- LICs with weak fundamentals respond much more strongly to isolated changes in explanatory variables than the median LIC; predicted probability of debt distress can swing from below the threshold to close to one.
- Most influential variables on predicted probability:
  - NPV of PPG external debt to GDP
  - Protection of private property rights
  - Real non-oil commodity prices
- Adverse shocks to NPV of PPG external debt to GDP and protection of private property rights can raise predicted probability above the threshold even for the median LIC.
- Extreme adverse shocks to current account balance to GDP and change in real non-oil commodity prices also predict distress for the median LIC.
- For covariates other than change in real oil prices, sufficiently large improvements could reduce predicted probability below the threshold for the LIC with weak fundamentals.
- Combined correlated deterioration of covariates would push predicted probability above the threshold faster.

### Debt thresholds and safe-debt levels (exact reported values)
- For the median LIC:
  - Predicted probability goes above the threshold when NPV of PPG external debt exceeds 66 percent of GDP.
  - Can accommodate an isolated decline in real non-oil commodity prices of 10 percent without exceeding the probability threshold if pre-shock NPV of debt is below 53 percent of GDP.
  - Can accommodate an isolated decline in real non-oil commodity prices of 20 percent without exceeding the probability threshold if pre-shock NPV of debt is about 40 percent of GDP.
- For the LIC with weak fundamentals:
  - Exceeds the threshold probability at a much lower NPV of debt of only about 18 percent of GDP.
  - A debt distress event would be predicted with greater confidence at NPV of debt exceeding 21 percent of GDP (the lower bound of the confidence interval for the predicted probability exceeds the threshold).
- Comparative DSF reference values:
  - The IMF-WB DSF NPV of debt threshold for LICs with strong institutions is 50 percent of GDP.
  - The IMF-WB DSF NPV of debt threshold for LICs with weak institutions is 30 percent of GDP.

### Policy and environment scenarios that alter probabilities (exact reported changes)
- Liquidity buffers and policy/institution improvements needed to reduce predicted probability for LIC with weak institutions:
  - Increase reserve coverage from 1.6 to 5.7 months of imports would bring predicted probability below the threshold for a LIC with weak institutions.
  - Very favorable external environment examples that delay predicted default for LIC with weak fundamentals:
    - Improve cumulative three year change in terms of trade from -9.3 to 42 percent.
    - Improve real non-oil commodity prices from 0 to about 20 percent.
  - Institutional improvement example:
    - Increase protection of private property rights from 5.2 to 6.5 would bring predicted probability below the threshold for a LIC with higher institutional quality.
- Conversely, significant deterioration in policies and the external environment may still not produce a predicted debt distress episode for a LIC with higher institutional quality.

### Interpretation and concluding empirical findings
- Initial conditions determined by policies and institutions profoundly influence the level of debt a LIC can safely carry.
- Exposure to adverse exogenous shocks increases the likelihood of debt distress and reduces safe debt levels.
- A plausible risk scenario should include simultaneous, correlated deteriorations (e.g., external shocks increasing current account deficits and eroding reserves), which would further reduce safe debt levels.
- Hypothesis tests and model implications summary:
  - Hypothesis 1 validated: weak protection of private property rights and high share of primary commodities in exports significantly raise probability of debt distress.
  - Hypothesis 2 validated: persistent declines in real non-oil commodity prices, increases in real oil prices for oil importers, and adverse terms of trade shocks significantly increase default probability and can lead to clustering.
  - Hypothesis 3 validated: widening current account deficits and declining reserve coverage significantly increase likelihood and timing of default; large reserve buffers can delay default.
  - Excluding protection of private property rights makes country-specific averages of policy and shock variables highly significant, indicating that private property rights explain unobserved country heterogeneity.
- Policy emphasis:
  - Strengthen persistent institutional factors (e.g., private property rights).
  - Improve macroeconomic policy records consistently over time (inflation control, fiscal balance, exchange rate management, reserve accumulation).
  - Account for exposure to commodity-price shocks in debt sustainability analysis and policy frameworks.

*Source: wp17114 - References (excerpted content).*

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

### wp17114 - References

### Context and motivation
- Despite waves of debt relief over more than three decades, many low-income countries (LICs) experienced prolonged periods of debt distress, linked to sovereign debt crises beginning in the late 1970s and 1980s.
- Official creditor responses evolved from mainly nonconcessional “flow reschedulings” to increasingly concessional rescheduling and debt relief mechanisms:
  - From 1976 to 1988 the Paris Club agreed upon 81 nonconcessional flow reschedulings with 27 of the countries now identified as HIPCs.
  - Toronto (1988), London (1990), and Naples (1994) terms reduced net present value of debt by 33, 50, and 67 percent respectively.
  - In 1996, the Heavily Indebted Poor Countries (HIPC) Initiative included debt relief on amounts owed to multilateral institutions.
  - PC creditors granted Lyon (1996) and Cologne (1999) terms (with 80 and 90 percent reduction in net present value of debt respectively) in the context of concerted action under the HIPC initiative.
- During 2000-2015, of the 39 countries eligible or potentially eligible for HIPC Initiative assistance, 36 received full and irrevocable reductions in their external debt.
- After HIPC debt relief, public debt ratios declined considerably in LICs (IMF (2014) cited).

### Research objective and contribution
- Objective: identify determinants of debt distress in LICs going beyond traditional debt burden and liquidity indicators by examining factors closely associated with long-term growth performance: economic institutions, macroeconomic policies, and vulnerability to external shocks.
- Key contributions:
  - Disentangles impact of highly persistent factors (quality of institutions, policy record, export concentration, exposure to external shocks) from transitory factors (contemporaneous external environment, recent policies, debt burden, liquidity indicators).
  - Uses a wide set of covariates to enable more country-specific and granular analysis of debt sustainability with higher predictive power than earlier studies.
  - Models unobserved country heterogeneity and its potential correlation with covariates to obtain consistent and efficient coefficient estimates.
  - First to examine impact of common exogenous shocks originating from commodity prices on LIC debt distress clustering.

### Main findings (empirical insights)
- Determinants and transmission channels:
  - Protection of private property rights (a highly persistent institutional variable) is among the most influential factors explaining the likelihood of debt distress.
  - LICs experiencing debt distress tend to have a record of weaker macroeconomic policies and higher exposure to adverse external shocks over the sample period compared to LICs without debt distress.
  - Results indicate that common, relatively persistent, adverse shocks to real non-oil commodity prices could lead to clustering of LIC debt distress episodes, conditional on initial debt burdens, policies, and institutions.
- Comparison with Kraay and Nehru (2006, KN):
  - KN used parsimonious probit regressions with three explanatory variables: the Country Policy and Institutional Assessments indicator (CPIA), real GDP growth, and an indicator of debt burden.
  - When CPIA is used as a single composite indicator, it is not possible to separate the relative significance of policy quality and institutional quality; this study measures institutions via private property rights to identify their separate role.
  - KN find growth effects significant only for mixed samples (LICs and MICs), not for LIC-only samples; this study comprehensively investigates multiple exogenous shocks and finds specific roles for commodity-price-related common shocks.
- Methodological improvements over KN:
  - Models unobserved country heterogeneity and its correlation with covariates.
  - Extends the set of LICs included in KN and refines KN’s definition of debt distress episodes by dropping some episodes identified solely by the IMF’s Extended Fund Facility (EFF).
  - Allows overlapping five-year “normal” episodes (KN defined normal episodes as non-overlapping five-year periods without distress), modeling correlation in error terms from unobserved country effects—raising the time-dimension coverage crucial to capture external shock impacts.

### Theoretical and policy implications
- Debt relief alone is not a panacea for restoring debt sustainability if underlying impediments to growth (weak institutions, poor policy records, exposure to shocks) persist.
- Possible reverse causality considerations:
  - Debt-overhang theory: high debt may reduce supply of new loans and deter investment (Krugman (1988), Sachs (1986)).
  - Alternative view for LICs: impediments such as poor business climate and uneven regulation may make causality run from low growth to high debt (Rajan (2005)); official concessional lenders often continue to provide financing, moderating a debt-driven cut in flows.
  - One-time debt relief is effective only if resources are allocated to total factor productivity enhancing uses (Koeda (2008)); absent such allocation and minimum institutional quality, debt relief may not lead to sustained growth (Asideu (2003)).
- Policy emphasis:
  - Strengthen persistent institutional factors (e.g., private property rights) that underpin long-term growth and reduce vulnerability to debt distress.
  - Improve macroeconomic policy records consistently over time (inflation control, fiscal balance, exchange rate management, reserve accumulation) to lower probability of distress.
  - Account for exposure to commodity-price shocks in debt sustainability analysis and policy frameworks to avoid clustering of distress episodes.

*Source: wp17114 - References (excerpted content).*

### conclusions and policy implications are summarized in Section V.

### wp17114 - conclusions and policy implications are summarized in Section V

### Hypotheses Tested
- Hypothesis 1: H.1.1 Structural variables determined by underlying economic and political institutions are highly influential in predicting debt distress in LICs. H.1.2 Weak economic institutions lead to higher probability of debt distress through persistently poor policies and high vulnerability to adverse external shocks.
  - Weak institutions → low growth, high macroeconomic instability → higher debt burden relative to repayment capacity.
  - Weak institutions → less diversified, commodity-dependent economies → higher vulnerability to external shocks.
  - Extractive institutions → high discount rate against the future → excessive debt accumulation and/or extraction of government assets.
  - Private property rights used as core measure of institutional quality.
- Hypothesis 2: Adverse external shocks raise the default probability in LICs both in the short-term and in the long-term; persistent adverse global shocks could lead to clustering of debt defaults.
  - Short-term: shocks help predict timing of default for countries with weak fundamentals.
  - Long-term: persistent/repeated shocks erode repayment capacity and accelerate debt accumulation; adverse shocks translate into substantial persistent output losses on average.
- Hypothesis 3: Liquidity conditions have an impact on the timing of default.
  - After controlling for institutions, shocks, and policies, liquidity conditions likely have a marginal effect and mainly predict timing for countries with weak fundamentals.

### Identification and Empirical Methodology
- Dependent variable: panel dummy = 1 for the start year of a debt distress episode and 0 for a normal episode (initial year of five consecutive years without debt distress).
  - Normal episodes defined as five years distress-free; three-year robustness check yields qualitatively similar results.
- Debt distress defined if any of:
  - (i) sum of interest and principal arrears exceeds five percent of total long term PPG external debt;
  - (ii) Paris Club rescheduling and/or debt reduction granted: year of debt relief and two subsequent years classified as distress;
  - (iii) IMF lending for immediate balance of payments needs exceeds 80 percent of the quota.
- IMF financing definition: actual purchases/drawn down under SBAs, CFF, and Oil Facility included; EFF drawings excluded.
- Country coverage and sample:
  - Broad sample: forty-two debt distress episodes for forty countries (Appendix II).
  - Estimation sample: thirty-one distress episodes with explanatory variables available.
  - Average duration of distress episodes: slightly longer than 20 years.
- Explanatory variable blocks: institutional quality, exogenous shocks, macroeconomic policies, debt burden, liquidity.
  - Debt burden measures: present value (PV) of PPG external debt scaled by GDP and by exports; PPG external debt service scaled by exports.
  - Institutional measure: private property rights index; alternative: weighted average of ICRG components (bureaucratic quality, corruption, rule of law, protection of property rights).
  - Macroeconomic (in)stability (MI) composite indicator formula described; higher mitot indicates increased instability.
  - Liquidity measures: reserve coverage (months of imports) and net resource transfers (nrt) as per GDF formula: nrt = D + grant + port + fdi + profit + short - P - I (as defined).
  - Exogenous shocks: terms of trade shocks, real export growth, natural disaster damage indicator; time-specific: real oil and non-oil commodity prices, real U.S. interest rates, cyclical world trade; share of exports of primary commodities tested.
- Econometric approach:
  - Binary response models for panel data (probit class): pooled probit, random effects (RE) probit, fixed effects (FE) probit discussed, Chamberlain’s correlated RE (includes time averages of covariates), Chamberlain’s probit, GEE with exchangeable correlation.
  - Six estimators used: pooled probit, Chamberlain’s probit, RE probit, Chamberlain’s RE probit, GEE, Chamberlain’s GEE. Choice based on tests of ρ and time averages significance.
  - Marginal effect expression preserved in notation: marginal effect of kth covariate = β_k φ(β̂' x_it + ĉ_i) / ...

### Main Estimation Results (Benchmark Specifications and Robustness)
- Three models estimated using alternative debt burden indicators:
  - Model I: PV of PPG external debt to GDP.
  - Model II: PV of PPG external debt to exports.
  - Model III: PPG external debt service to exports.
- Two benchmark specifications for each model:
  - Benchmark specification 1: includes protection of private property rights (smaller sample).
  - Benchmark specification 2: excludes private property rights and uses country-specific averages to capture institutional effects indirectly (larger sample).
- Sample and estimation notes:
  - Overall sample: fifty-seven PRGF-eligible countries, period 1975–2010; estimation restricted to 1975–2006 due to episode definition.
  - Estimation sample sizes vary by specification: examples include # of observations = 257 (benchmark 1 Model I) and 388 (benchmark 2 Model I); # of countries = 43 (benchmark 1 Model I) and 53 (benchmark 2 Model I).
  - Sample probability of default: 0.12 (benchmark 1 Model I), 0.10 (benchmark 2 Model I).
- Key statistically significant determinants (across models and specifications):
  - Institutional quality:
    - Private property rights significant and negative where included (e.g., coefficient in Table 1 Benchmark 1: -0.583***(0.208); Benchmark 2 similar when available).
    - Conclusion: weak protection of private property rights significantly increases probability of debt distress → supports H.1.1 and H.1.2.
  - External shocks:
    - Cumulative three year change in real non-oil commodity prices significant and negative (e.g., Table 1: coefficients -0.0361***(0.00780) etc.), indicating declines in non-oil commodity prices increase default risk.
    - Cumulative three year change in terms of trade often negative and significant (e.g., Table 1: -0.0134**(0.00571)).
    - Cumulative three year change in real oil prices: for oil importers positive/insignificant in some specs; for oil exporters negative and significant (e.g., Table 1: -0.159**(0.0742)).
    - Finding: shocks matter both contemporaneously and via country-specific sample averages → supports Hypothesis 2.
  - Debt burden:
    - Higher PV of PPG external debt to GDP increases probability of default (Model I: Present value coefficient examples: 0.0277***(0.0105) in Table 1).
    - PV of PPG external debt to exports and debt service to exports also positively significant in Models II and III (Tables 4 and 5).
  - Policy and liquidity:
    - Reserve coverage in months of imports significant and negative in many specifications (e.g., Table 1 reserve coverage (t-1): -0.178**(0.0850); Table 4 similar).
    - Current account balance to GDP (t-1) negative and significant in several specs (Table 1: -0.0455**(0.0177)).
    - Macroeconomic (in)stability indicator significant where property rights excluded (evidence that institutions account for heterogeneity): e.g., Table 1 country-specific macro instability 0.0935**(0.0426).
    - Inflation and change in reserve coverage significant in some component specifications.
  - Liquidity via creditors/donors:
    - Net resource transfers and its components generally not significant.
- Magnitudes and relative influence (marginal effects and simulations):
  - Benchmark 1 marginal effects highlight the most influential variables: protection of private property rights, real non-oil commodity prices, and the debt burden.
  - Benchmark 2 marginal effects: sample averages of current account balance and reserve coverage, followed by real non-oil commodity prices, are most influential.
  - Simulations: illustrate predicted probabilities for median LIC versus LIC with weak fundamentals; predicted probability for debt distress is zero for the median LIC under specified initial values (simulation discussion continued in text beyond provided excerpt).

### Goodness of Fit and Predictive Performance
- Pseudo R2 and likelihood tests:
  - Pseudo R2 for Model I reported values: 0.52, 0.38, 0.40, 0.40, 0.40 (Table 1 columns) indicating about 40 to 50 percent of uncertainty explained by benchmark specifications.
  - Likelihood ratio tests significantly reject joint null that explanatory coefficients are zero.
- Predicted probabilities distributions (percentiles) distinguish crisis events from normal episodes:
  - Example Model I benchmark specification 1 (Table 6):
    - Median predicted probability for crisis events = 0.61; for normal episodes = 0.01.
    - 75th percentile for crisis events = 0.88; 75th percentile for normal episodes = 0.08.
  - Similar discrimination reported for Models II and III (Tables 7 and 8).
- Overall: models with different debt burden indicators differentiate well between distress and normal episode distributions.

### Threshold Probability Analysis and Forecasting Framework
- Threshold probabilities derived by minimizing asymmetrically-weighted and equally-weighted loss functions (weights for type I and II errors: 0.65 and 0.35 respectively).
- Primary and secondary thresholds reported for Models I-III and benchmark specifications (Table 9 summary):
  - Model I (PV PPG external debt/GDP), Benchmark 1:
    - Primary threshold = 0.14; Secondary threshold = 0.14.
    - Missed crises (Type I) = 0.097; False alarms (Type II) = 0.133; Overall error = 0.128.
  - Model I, Benchmark 2:
    - Primary threshold = 0.04; Secondary threshold = 0.137.
    - Primary misses only one crisis: Missed crises = 0.026; False alarms = 0.332; Overall error = 0.302.
  - Model II and III summarized similarly in Table 9 (selected figures):
    - Model II Benchmark 1 primary threshold = 0.23, Missed crises = 0.133, False alarms = 0.071, Overall error = 0.078.
    - Model III Benchmark 1 primary threshold = 0.06, Missed crises = 0.034, False alarms = 0.230, Overall error = 0.208.
- Interpretation:
  - Predicted probabilities > threshold → classify country as moderate to high likelihood of debt distress.
  - Secondary (higher) threshold → classify as high risk of debt distress.
  - Benchmark specification 1 (including private property rights) generally has lower overall error than specification 2.
- Comparison to KN-based DSF: models in this study have better predictive power versus corresponding KN regressions.

### Additional Findings and Robustness Notes
- Components of MI index tested separately—results broadly similar; inflation and change in reserve coverage significant in some component specifications.
- Government balance to GDP generally insignificant though sign consistent.
- U.S. real interest rate not significant in most specifications.
- Real GDP growth and real export growth not significant in LIC sample.
- Share of people affected by natural disasters not significant for immediate defaults.
- Robustness: alternative definitions of IMF financing and inclusion of S&P default classification tested; overall qualitative results robust.

*Source: Author's calculations presented in wp17114 - conclusions and policy implications are summarized in Section V.*

### 0.27 for the LIC with weaker fundamentals. In reference to the threshold probabilities

### wp17114 - 0.27 for the LIC with weaker fundamentals. In reference to the threshold probabilities

### Counterfactual simulations: setup and key parameter values
- Simulations are counterfactual: all other variables remain unchanged while a single variable is allowed to change over its entire sample range.
- Table (values used in counterfactuals; reported exactly):
  - Present value of PPG external debt to GDP (%) — 23.8; 34.8
  - Reserve coverage in months of imports — 2.8; 1.6
  - Cumulative three year change in terms of trade — 0.8; -9.3
  - Cumulative three year change in real non-oil commodity prices — 0.0; 0.0
  - Cumulative three year change in real oil prices for oil importers — 0.0; 0.0
  - Private property rights — 6.3; 5.2
  - Current account balance to GDP (%) — -2.4; -7.0
  - Share of primary commodities in total exports (%) — 69.5; 92.6
  - Threshold probability — 0.14
  - Predicted probability — 0.007; 0.267
  - 95% confidence band — [0.001 0.033]; [0.143 0.465]
- Note: Threshold and predicted probabilities are reported for benchmark specification 1 for model I with present value of PPG external debt to GDP (%).

### Main simulation findings (marginal effects and heterogeneity)
- LICs with weak fundamentals respond much more strongly to isolated changes in explanatory variables than the median LIC; in some cases predicted probability of debt distress swings from below the threshold to close to one.
- The most influential variables on predicted probability of debt distress:
  - NPV of PPG external debt to GDP
  - Protection of private property rights
  - Real non-oil commodity prices
- Adverse shocks to NPV of PPG external debt to GDP and protection of private property rights can raise predicted probability above the threshold even for the median LIC.
- Extreme adverse shocks to:
  - Current account balance to GDP
  - Change in real non-oil commodity prices
  also result in predicting debt distress for the median LIC.
- For covariates other than change in real oil prices, sufficiently large improvements could reduce predicted probability below the threshold for the LIC with weak fundamentals.
- Overall: predicted probability increases sharply for a large deterioration in any covariate for the LIC with weak fundamentals; combined correlated deterioration would push the predicted probability above the threshold faster.

### Debt thresholds and safe-debt levels (exact reported values)
- For the median LIC:
  - Predicted probability goes above the threshold when NPV of PPG external debt exceeds 66 percent of GDP.
  - Can accommodate an isolated decline in real non-oil commodity prices of 10 percent without exceeding the probability threshold if pre-shock NPV of debt is below 53 percent of GDP.
  - Can accommodate an isolated decline in real non-oil commodity prices of 20 percent without exceeding the probability threshold if pre-shock NPV of debt is about 40 percent of GDP.
- For the LIC with weak fundamentals:
  - Exceeds the threshold probability at a much lower NPV of debt of only about 18 percent of GDP.
  - A debt distress event would be predicted with greater confidence at NPV of debt exceeding 21 percent of GDP (the lower bound of the confidence interval for the predicted probability exceeds the threshold).
- Comparative DSF reference values cited:
  - The IMF-WB DSF NPV of debt threshold for LICs with strong institutions is 50 percent of GDP.
  - The IMF-WB DSF NPV of debt threshold for LICs with weak institutions is 30 percent of GDP.

### Policy and environment scenarios that alter probabilities (exact reported changes)
- Liquidity buffers and policy/institution improvements needed to reduce predicted probability for LIC with weak institutions:
  - Increase in reserve coverage from 1.6 to 5.7 months of imports would bring predicted probability below the threshold for a LIC with weak institutions.
  - Very favorable external environment examples that delay predicted default for LIC with weak fundamentals:
    - Improve cumulative three year change in terms of trade from -9.3 to 42 percent.
    - Improve real non-oil commodity prices from 0 to about 20 percent.
  - Institutional improvement example:
    - Increase protection of private property rights from 5.2 to 6.5 would bring predicted probability below the threshold for a LIC with higher institutional quality.
- Conversely, significant deterioration in policies and the external environment may still not produce a predicted debt distress episode for a LIC with higher institutional quality.

### Interpretation and inference
- Initial conditions determined by policies and institutions profoundly influence the level of debt a LIC can safely carry.
- Exposure to adverse exogenous shocks increases the likelihood of debt distress and reduces safe debt levels.
- A plausible risk scenario should include simultaneous, correlated deteriorations (e.g., external shocks increasing current account deficits and eroding reserves), which would further reduce safe debt levels.
- The study highlights difficulty in deriving clear universal debt thresholds from a rich model with many factors; probability thresholds better capture country-specific differences in policies, institutions, and external environment.

### Concluding empirical findings (summary of hypothesis tests and model implications)
- Hypothesis 1 (role of structural variables/institutions): validated — weak protection of private property rights and high share of primary commodities in exports significantly raise probability of debt distress.
- Hypothesis 2 (external shocks): validated — persistent declines in real non-oil commodity prices, increases in real oil prices for oil importers, and adverse terms of trade shocks significantly increase default probability and can lead to clustering of defaults.
- Hypothesis 3 (liquidity/timing): validated — widening current account deficits and declining reserve coverage significantly increase likelihood and timing of default; large reserve buffers can delay default.
- When protection of private property rights is excluded from the model:
  - Country-specific averages of policy and shock variables become highly significant (they correlate with country heterogeneity).
  - Unobserved country heterogeneity becomes significant, indicating that protection of private property rights explains country heterogeneity in the model.
  - Weak institutions raise probability of debt distress through weak policies and higher vulnerability to external shocks.
- The most influential variables in predicting debt distress: protection of private property rights, real non-oil commodity prices, and the debt burden. When institutions are excluded, influential variables shift to averages of current account balance, reserve coverage, and real non-oil commodity prices.

*Source: wp17114 (selected excerpts and figures as provided).*

### REFERENCES

### REFERENCES

### Institutions and Growth
- Acemoglu Daron and James Robinson, 2008, “The Role of Institutions in Growth and Development,” Commission on Growth and Development Working Paper No. 10.
- Acemoglu Daron, Simon Johnson and James Robinson, 2005a, “Institutions as the Fundamental Cause of Long-Run Growth,” by P. Aghion and S. Durlauf (eds). Handbook of Economic Growth, North Holland.
- ———, 2005b, “The rise of Europe: Atlantic trade, institutional change and economic growth,” American Economic Review, 95, 546–79.
- Acemoglu, Daron and Simon Johnson, 2005, “Unbundling Institutions,” Journal of Political Economy, Vol. 113, No. 5, pp. 949–995.
- Acemoglu, Daron, Simon Johnson, and James A. Robinson, 2002, Reversal of fortune: Geography and institutions in the making of the modern world income distribution. The Quarterly journal of economics, 117(4), 1231-1294.
- Acemoglu, Daron, Simon Johnson, and James A. Robinson, 2001, “The Colonial Origins of Comparative Development: An Empirical Investigation,” American Economic Review, Vol. 91, No.5, pp. 1369–1401.
- Knack, S., and Keefer, P., 1995, “Institutions and economic performance: cross‐country tests using alternative institutional measures,” Economics & Politics, 7(3), pp. 207-227.
- North, Douglass, 1990, Institutions, Institutional change, and Economic Performance, New York: Cambridge University Press.
- Kohlscheen, E., 2010, “Sovereign risk: Constitutions rule,” Oxford Economic Papers, 62(1), pp. 62-85.
- Kohlscheen, E., 2007, “Why are there serial defaulters? Evidence from constitutions,” The Journal of Law and Economics, 50(4), pp. 713-730.
- Van Rijckeghem, C., and Weder, B., 2009, “Political institutions and debt crises,” Public Choice, 138(3-4), pp. 387-408.

### Debt, Debt Relief, and Sovereign Defaults
- Arslanalp, Serkan and Peter Blair Henry, 2006, “Debt Relief,” The Journal of Economic Perspectives, Vol. 20, No. 1 (Winter, 2006), pp. 207–220.
- Asiedu, Elizabeth, 2003, “Debt relief and institutional reform: a focus on Heavily Indebted Poor Countries”, The Quarterly Review of Economics and Finance, 43(4), 614-626.
- Barry Herman, 2005, “How well do measurements of an enabling domestic environment for development stand up,” The IMF and the World Bank at Sixty, 281.
- Berg, A., and Sachs, J., 1988, “The debt crisis structural explanations of country performance,” Journal of Development Economics, 29(3), 271-306.
- Brooks Ray, M. Cortes, F. Fornasari, B. Ketchekmen, Y. Metzgen, R. Powell, S. Rizavi, D. Ross and K. Ross, 1998, “External Debt Histories of Ten Low-Income Developing Countries: Lessons from Their Experience,” IMF Working Paper No. 98/72 (Washington: International Monetary Fund).
- Cline, William. R., 1985, “International debt: from crisis to recovery?” The American Economic Review, 75(2), 185-190.
- Daseking, M. C., and Powell, M. R., 1999, “From Toronto terms to the HIPC initiative: a brief history of debt relief for low-income countries,” IMF Working Paper No. 99/142 (Washington: International Monetary Fund).
- Chauvin, Depetris N., and Kraay, Aart., 2005, “What has 100 billion dollars worth of debt relief done for low-income countries?” http://ssrn.com/abstract=818504.
- Easterly William, 2002, “How Did Heavily Indebted Poor Countries Become Heavily Indebted? Reviewing Two Decades of Debt Relief,” World Development Vol. 30, No. 10, pp. 1677–1696.
- Kruger, Mark, and Miguel Messmacher. 2004. “Sovereign Debt Defaults and Financing Needs,” IMF Working Paper No. 04/53 (Washington: International Monetary Fund).
- Krugman, Paul, 1988, "Financing versus Forgiving a Debt Overhang,” Journal of Development Economics, Vol. 29 No.3, pp. 253–68.
- Manasse, Paolo, Nouriel Roubini, and Axel Schimmelpfennig, 2003, "Predicting sovereign debt crises," IMF Working Paper No. 03/221 (Washington: International Monetary Fund).
- Panizza Ugo, Federico Sturzenegger, and Jeromin Zettelmeyer, 2009, “The Economics and Law of Sovereign Debt and Default,” Journal of Economic Literature, Vol. 47, No. 3, pp. 651–698.
- Pescatori, Andrea, and Sy, Amadou N., 2007, “Are debt crises adequately defined?” IMF Staff Papers, 54(2), 306-337.
- Presbitero, Andrea F., 2009, “Debt‐Relief Effectiveness and Institution‐Building,” Development Policy Review, 27(5), pp. 529-559.
- Rajan Raghuram, 2005, “Debt Relief and Growth,” Finance and Development, June 2005, Volume 42, Number 2.
- Reinhart, Carmen M., Kenneth S. Rogoff, and Miguel A. Savastano, 2003, “Debt Intolerance,” Brookings Papers on Economic Activity, Vol.1, pp. 1–74.
- Roubini, N., Manasse, P., and Schimmelpfennig, A., 2003, Predicting sovereign debt crises (No. 03-221). IMF Working Paper No. 03/221 (Washington: International Monetary Fund).
- Sachs, Jeffrey, 1989, "The Debt Overhang of Developing Countries," in Debt Stabilization and Development, Guillermo A. Calvo, Ronald Findlay, Pentti Kouri and Jorge Braga de Macedo, eds. Oxford: Basil Blackwell, pp. 80–102.
- Powell, Robert, and Bird, G., 2010, “Aid and debt relief in Africa: have they been substitutes or complements?” World Development, 38(3), pp.219-227.
- Kraay, Aart, and Vikram Nehru, 2006, “When is external debt sustainable?” The World Bank Economic Review, 20(3), pp. 341-365.
- Pescatori, Andrea, and Sy, Amadou N., 2007, “Are debt crises adequately defined?” IMF Staff Papers, 54(2), 306-337.

### IMF, World Bank, and Policy Frameworks
- International Monetary Fund (IMF), 2014, Macroeconomic Developments in Low-Income Developing Countries, IMF (Washington).
- ——— and World Bank, 2013, Staff Guidance Note on the Application of the Joint Bank-Fund Debt Sustainability Framework for Low-Income Countries, IMF and The World Bank, (Washington).
- ——— and World Bank, 2012, Revisiting the Debt Sustainability Framework for Low-Income Countries, IMF and The World Bank, (Washington).
- ——— and World Bank, 2010, Staff Guidance Note on the Application of the Joint Bank-Fund Debt Sustainability Framework for Low-Income Countries, IMF and The World Bank, (Washington).
- ——— and World Bank, 2004, Debt Sustainability in Low-Income Countries - Proposal for an Operational Framework and Policy Implications, IMF and The World Bank, (Washington).
- Berg, M. A., Berkes, M. E., Pattillo, M. C. A., Presbitero, A., and Yakhshilikov, M. Y., 2014, “Assessing Bias and Accuracy in the World Bank-IMF's Debt Sustainability Framework for Low-Income Countries” IMF Working Paper No. No. 14/48 Washington: International Monetary Fund).
- Becker Torbjorn, and Paolo Mauro, 2007, “Output Drops and the Shocks That Matter,” IMF Working Paper No.06/172 (Washington: International Monetary Fund).
- Brooks Ray, M. Cortes, F. Fornasari, B. Ketchekmen, Y. Metzgen, R. Powell, S. Rizavi, D. Ross and K. Ross, 1998, “External Debt Histories of Ten Low-Income Developing Countries: Lessons from Their Experience,” IMF Working Paper No. 98/72 (Washington: International Monetary Fund).
- Manasse, Paolo, Nouriel Roubini, and Axel Schimmelpfennig, 2003, "Predicting sovereign debt crises," IMF Working Paper No. 03/221 (Washington: International Monetary Fund).

### Methodology and Empirical Tools
- Cameron, A. Colin, Frank A.G. Windmeijer, 1997, “An R-squared measure of goodness of fit for some common nonlinear regression models”, Journal of Econometrics, vol. 77, pp. 329–342.
- Zeger, Scott L., and Kung-Yee Liang, 1986, "Longitudinal data analysis for discrete and continuous outcomes," Biometrics, 121-130.
- Demirgüc-Kunt, Asli and Enrica Detragiache, 1999, “Monitoring Banking Sector Fragility: A Multivariate Logit Approach,” IMF Working Paper No. 99/147 (Washington: International Monetary Fund).
- Detragiache, Enrica and Spilimbergo, Antonio, 2001, “Crises and Liquidity: Evidence and Interpretation,” IMF Working Paper No. 01/2 (Washington: International Monetary Fund).
- McFadden, Daniel, Richard Eckaus, Gershon Feder,Vassilis A. Hajivassiliou, and Stephen O’Connell, 1985, “Is There Life after Debt? An Econometric Analysis of the Creditworthiness of Developing Countries,” In International Debt and the Developing Countries, ed. Gordon W. Smith and John T. Cuddington, pp. 179–209. Washington, D.C.: World Bank.
- Fischer, Stanley, 1993, “The Role of Macroeconomic Factors in Growth”, NBER Working Paper Series, Working Paper No. 4565.
- Papageorgiou C., C. Pattillo, N. Spatafora, and A. Berg, 2010. "The End of an Era? The Medium- and Long-term Effects of the Global Crisis on Growth in Low-Income Countries," IMF Working Paper No. 10/205 (Washington: International Monetary Fund).
- Pattillo, Catherine A., Helene Poirson, and Luca Antonio Ricci, 2002, “External debt and growth,” No. 2002-2069. IMF Working Paper No. 02/69 (Washington: International Monetary Fund).
- Sahay, Ratna and Rishi Goyal, 2006, “Volatility and Growth in Latin America: An Episodic Approach”, IMF Working Paper No. 06/287 (Washington: International Monetary Fund).

*Source: wp17114 - REFERENCES*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2017/wp17114.pdf_
