## Appendix A (wpiea2026161-source-pdf)

## Source details

**Canonical URL:** [Appendix A (wpiea2026161-source-pdf)](https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026161-source-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2026/english/wpiea2026161-source-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2026/english/wpiea2026161-source-pdf.pdf.json)

---

### Contents and scope
- Table A1: Overview of control variables
- Table A2: Descriptive statistics
- Table A3: Share of Senior Debt and the Sovereign Debt Crisis
- Table A4: Share of Senior Debt and Borrowing Costs
- Figure A1: Analysis of Creditors’ RPIDs in Countries with an IMF Program
- Figure A2: Analysis of Creditors’ RPIDs in Countries During Financial Crises
- Figure A3: Percentile distribution of external debt stock as % of GNI
- Sample and period: Annual data 1980 to 2022 covering 119 emerging market and developing economies.
- Creditor categories (World Bank IDS classification): (i) bilateral creditors; (ii) multilateral institutions, excluding the IMF; (iii) bondholders; (iv) long-term commercial bank loans and credits; (v) the International Monetary Fund (IMF).
- Key data sources: World Bank IDS; BoC–BoE sovereign default database; Laeven and Valencia (2020) debt-crisis dummy; J.P. Morgan EMBIG; sovereign ratings averaged from Moody’s, Fitch, and S&P standardized to a 21 point scale.

### Data and measurement (control variables; Table A1)
- IMF Credit Outstanding (% of GNI): IDS
- Multilateral Debt (excl. IMF) (% of GNI): IDS
- Bilateral Debt (% of GNI): IDS
- Bondholders Debt (% of GNI): IDS
- Commercial Debt (% of GNI): IDS
- External Debt to GNI: "Total external debt is debt owed to nonresidents repayable in foreign currency, goods, or services. It is the sum of public, publicly guaranteed, and private nonguaranteed long-term debt, short-term debt, and use of IMF credit." Source: IDS
- Sovereign Default: "Total amount of defaults with each creditor" Source: BoC-BoE sovereign default database
- IMF Liquidity: "IMF liquidity ratio equals liquid resources divided by liquid liabilities. We use the average IMF liquidity ratio from the past five years." Source: IMF’s Annual Reports
- Log of GNI: Logarithm of Gross National Income. Source: WDI
- Growth: Annual GDP Growth Rate (%). Source: WDI
- Debt Restructuring_Past: "Total Number of Debt Restructurings in the Last 5 Years" Authors' calculation based on Asonuma and Trebesch (2016), Paris Club on Private and Official debt restructuring
- Banking Crises: Binary =1 if Banking Crisis is Ongoing. Source: Reinhart & Rogoff (2009)
- Debt Crises: Binary =1 if Sovereign Debt Crisis is Ongoing. Source: Reinhart & Rogoff (2009)
- Market Access: Access to International Financial Markets (1=yes, 0=no). Calculated by the authors based on data from Moody's, Fitch, Standard & Poor's (S&P)
- EMBIG Spreads: The J.P. Morgan Emerging Market Bond Index Global (EMBIG) index. Source: J.P. Morgan
- IMF Program: Participation in an IMF Program (1=yes, 0=no). Source: IMF, Dreher (2012)
- Trade openness: "Trade openness is defined as the sum of a country’s exports and imports as a fraction of GNI in percent." Source: WDI
- VIX: Market Vulnerability Index (VIX). Source: Chicago Board Options Exchange

### Stylized facts on de facto seniority (RPID)
- RPID definition: difference between default-to-debt ratio of creditor group k and that of the entire creditor portfolio; range is [-1, 1]. Negative RPID: preferential treatment; Positive RPID: less preferential treatment.
- Main stylized facts:
  - The IMF and the World Bank display the most negative RPID values, confirming their seniority.
  - Official bilateral creditors display negative RPID values on average, ranking behind multilateral creditors but ahead of private creditors.
  - Private creditors — bondholders and commercial banks — display positive RPID values, indicating they bear a higher-than-average share of defaults.
  - On average commercial banks are treated as less senior compared to bondholders.
- Historical patterns:
  - First wave: 1980s–early 1990s (e.g., Peru 1985–1993; Vietnam 1985–1993; Zambia 1986–1994).
  - Second wave: 1990s–early 2000s (e.g., Bosnia and Herzegovina 1992–2000; Iraq 1990–2005; Liberia 1998–2008).
  - Recent phase: 2010s–2020s (e.g., Sudan 1981–2021; Somalia 2005–2020; Zimbabwe 2003–2022).
- Robustness: Across subsamples (IMF program participation; financial-crisis episodes) IMF and World Bank consistently lowest RPIDs; bilateral creditors privileged relative to private creditors.

### Empirical strategy
- Debt-crisis likelihood: Probit / linear probability models using creditor-specific debt shares (% of GNI).
- Endogeneity for IMF lending addressed by IV:
  - IVi,t = Historical_IMF_Loani,t × IMF_Liquidityt
  - Historical_IMF_Loan: average annual IMF loan-to-GDP ratio over previous 10 years.
  - IMF_Liquidity: ratio of the Fund’s liquid assets to its liquid liabilities.
- Market-access analysis: OLS and Two-Stage Least Squares (2SLS) for sovereign bond spreads (EMBIG).
- Baseline illustrative linear probability model:
  - Pr(Debt Crisis = 1) = β1 IMFi,t + β2 Multilaterali,t + β3 Bilateriali,t + β4 Bondholdersi,t + β5 CommercialBanki,t + δ X i,t + ηi + μt + εi,t
- Creditor shares expressed as share of GNI (robustness checks use creditor shares in total debt).

### Key empirical results — debt crises (Table 1, Table 2, Table 3 summaries)
- Baseline Probit marginal-effect highlights (creditor shares as % of GNI; robust SEs in parentheses):
  - IMF: -0.00815*** (0.00209) and -0.01315*** (0.00333) in specifications shown.
  - Multilateral Loans (excl. IMF): 0.00333*** (0.00067) and 0.00295*** (0.00074).
  - Bilateral Loans: 0.00437*** (0.00098) and 0.00433*** (0.00108).
  - Bond Loans: 0.00239** (0.00107) and 0.00286** (0.00123).
  - Commercial Bank Loans: -0.00039 (0.00125) and -0.00115 (0.00152) — not significant.
- Textual marginal-effect interpretations:
  - A one percentage point increase in the share of bonds (as % of GNI) associated with a rise of between 0.24 and 0.29 percentage points in crisis probability.
  - A one percent increase in multilateral loans associated with a 0.30 percentage points increase in crisis probability.
  - A one percent increase in bilateral loans associated with a 0.43 percentage points increase in crisis probability.
  - IMF credit outstanding associated with a reduction in default probability with estimated effects ranging between -0.81 and -1.32 percentage points (text summary of columns (1) and (6) in baseline discussion).
- IV-Probit / IV estimates (IMF instrumented; robust SEs in parentheses):
  - IMF (IV results): -0.0654* (0.0380), -0.173*** (0.0488), -0.147** (0.0602) across columns (1)–(3).
  - Multilateral Loans (excl. IMF): 0.0319*** (0.00328) and 0.0217*** (0.00574) where reported.
  - Bilateral Loans: 0.0262*** (0.00454) and 0.0222*** (0.00484).
  - Bonds: 0.00689 (0.00625) and 0.0163** (0.00799).
  - Commercial Bank Loans: 0.0166* (0.00867) and -0.0101 (0.00785).
  - Selected control variables (IV results): Log of GNI -1.153*** (0.353) and -0.942*** (0.306); VIX Index 1.106*** (0.263) and 1.076*** (0.242); Short-Term Debt % 0.0127*** (0.00344) and 0.0178*** (0.00420); Debt Restructuring_Past 0.620*** (0.0525) and 0.590*** (0.0490); Banking Crises 0.372** (0.166) and 0.408** (0.167); Foreign Direct Investment -0.0135** (0.00584) and -0.0150** (0.00625).
- Non-linear / indebtedness-threshold findings:
  - Debt-stock percentiles and ranges:
    - Lower percentiles (10 to 40): debt burden between 18% and 40%.
    - Intermediate percentiles (40th–75th): external debt between 43% and 68% of GDP.
    - Upper percentiles (75–100): debt stocks ranging from 68% to 159% of GDP.
  - IMF % of GNI coefficients (selected; standard errors in parentheses):
    - IMF % of GNI: -0.0489*** (0.0161), -0.135*** (0.0337), -0.125*** (0.0258), -0.130*** (0.0236), -0.0840*** (0.0232) across columns.
    - IMF % of GNI × Stock Debt<70%: -0.0763*** (0.0295).
    - IMF % of GNI × Stock Debt>70%: 0.0914** (0.0404).
    - IMF % of GNI × Stock Debt>90%: 0.0919*** (0.0342).
    - IMF % of GNI × Stock Debt>100%: 0.110*** (0.0312).
    - IMF % of GNI × Stock Debt>120%: 0.0884** (0.0451).
  - Interpretation: IMF lending reduces crisis probability across subsamples but the marginal effect diminishes as debt stock rises; stabilizing effect strongest for indebtedness below 70 percent of GDP, marginal effects decline at higher debt thresholds.

### Key empirical results — sovereign bond spreads (EMBIG; Table 4 and robustness)
- Estimation: 2SLS with IMF loans instrumented by Historical_IMF_Loan × IMF_Liquidity.
- Baseline OLS and preferred 2SLS coefficients (robust SEs in parentheses):
  - IMF:
    - OLS: -0.995** (0.392)
    - 2SLS: -2.317** (0.938)
  - Multilateral Loans (excl. IMF):
    - OLS: 0.514*** (0.115)
    - 2SLS: 0.618*** (0.156)
  - Bilateral Loans:
    - OLS: -0.0432 (0.0898)
    - 2SLS: 0.581*** (0.216)
  - Bond_Loans:
    - OLS: 0.00308 (0.0540)
    - 2SLS: 0.147* (0.0769)
  - Commercials Bank Loans:
    - OLS: -0.0586 (0.0868)
    - 2SLS: 0.248** (0.111)
- Other notable 2SLS controls (robust SEs in parentheses):
  - Log_GNI: -5.641** (2.413)
  - GDP Growth: -0.00341 (0.170)
  - Debt Crises: 4.823*** (1.048)
  - Currency Crises: 2.346** (0.987)
  - Capital Account Openness: -1.609*** (0.475)
  - Short-Term Debt % GNI: 0.407*** (0.121)
- Instrument relevance: Kleibergen-Paap rk 12.829; P-val 0.001.
- Interpretation:
  - IMF credit outstanding is negatively and significantly associated with sovereign spreads (2SLS larger magnitude than OLS).
  - Coefficients on non-IMF multilateral and bilateral loans are positive and significant in 2SLS — interpreted as adverse-selection (countries relying more on official lenders often have higher underlying risk), not necessarily causal.
  - Bond and commercial-bank loan coefficients become positive and significant in 2SLS, consistent with private-creditor exposure associating with higher spreads.

### Descriptive statistics (Table A2 selected)
- IMF (% of GNI): Observations 4,098; Mean 2.564239; Std. Dev. 4.977649; Min 0; Max 69.97369
- Multilateral Loans (excl. IMF) (% of GNI): Observations 4,349; Mean 18.58857; Std. Dev. 21.33003; Min 0; Max 319.9574
- Bilateral Loans (% of GNI): Observations 4,357; Mean 18.35796; Std. Dev. 33.52719; Min 0; Max 652.4404
- Bond Loans (% of GNI): Observations 4,625; Mean 2.92994; Std. Dev. 7.56168; Min 0; Max 144.8
- Commercials Bank loans (% of GNI): Observations 4,998; Mean 2.835316; Std. Dev. 7.682194; Min 0; Max 162.8461
- Log of GNI: Observations 4,436; Mean 22.96832; Std. Dev. 2.141282; Min 17.94961; Max 30.50858
- Growth: Observations 4,620; Mean 3.530515; Std. Dev. 6.127119; Min -64.0471; Max 88.95766
- Debt Restructuring_Past: Observations 4,403; Mean .5210084; Std. Dev. .9955383; Min 0; Max 5
- Banking Crises: Observations 4,920; Mean .0567073; Std. Dev. .231306; Min 0; Max 1
- Debt Crises: Observations 4,663; Mean .4509972; Std. Dev. .4976463; Min 0; Max 1
- EMBIG Spreads: Observations 988; Mean 599.6267; Std. Dev. 1272.046; Min 0; Max 31310.52
- IMF Program: Observations 4,998; Mean .2933173; Std. Dev. .4553282; Min 0; Max 1
- Openness: Observations 3,953; Mean 34.94235; Std. Dev. 17.47341; Min .0104996; Max 173.9983
- VIX: Observations 6,460; Mean 19.60173; Std. Dev. 5.762154; Min 11.09024; Max 32.69553

### Regression results — Share of Senior Debt and the Sovereign Debt Crisis (Table A3 selected)
- Dependent variable: Debt Crises. Observations: 999. Year FE: Yes. Country FE: Yes.
- Share of IMF Debt:
  - Column (1): -0.00802*** (0.00184)
  - Column (2): -0.00680*** (0.00169)
- Share of Senior_Creditors Debt (Excl. IMF):
  - Column (2): 0.00269** (0.00109)
  - Column (3): 0.00221** (0.00109)
- Share of Senior_creditors_debt (Incl. IMF):
  - Column (4): 0.0102 (0.00653)
- Selected controls (robust SEs in parentheses):
  - Log_of_GNI: Column (1): -0.29585*** (0.06136); Column (4): -1.458*** (0.426)
  - GDP Growth: Column (1): -0.00778** (0.00319)
  - VIX index: Column (1): 0.34158*** (0.04064)
  - Short-Term Debt %: Column (1): 0.00498*** (0.00127)
  - Debt Restructuring_Past: Column (1): 0.15553*** (0.01959)

### Regression results — Share of Senior Debt and Borrowing Costs (Table A4)
- Dependent variable: EMBIG. Observations: 102. R-squared: 0.817. Kleibergen-Paap rk: 11.232. Kleibergen-Paap rk (P-val): 0.0036. Country FE: Yes. Year FE: Yes.
- Key coefficient estimates (robust SEs in parentheses):
  - Share of IMF Debt: -0.395* (0.216)
  - Share of Senior_Creditors Debt (Excl. IMF): -0.0684*** (0.0224)
  - Debt Crises: 5.496*** (1.791)
  - Currency Crises: 1.378* (0.750)
  - Capital Account Openness: -0.865*** (0.321)
  - Constant: -10.03 (36.62)

### Robustness checks (Appendix B highlights)
- Alternative samples: excluding 1980–1990; excluding Cold War years and USSR collapse; excluding 2008–2009; excluding Covid-19 years. Main patterns persist:
  - IMF lending remains negatively associated with debt-crisis probability.
  - Multilateral (excl. IMF) and bilateral loans remain positively associated with crisis risk.
  - EMBIG regressions: IMF credit outstanding negatively associated with spreads; other official exposures positively associated with spreads.
- Panel 5A (IVProbit) selected coefficients:
  - IMF: -0.245*** (0.0709), -0.171*** (0.0526), -0.111* (0.0580), -0.146** (0.0599) across columns.
  - Multilateral Loans (excl. IMF): 0.0298*** (0.00607), 0.0235*** (0.00542), 0.0189*** (0.00483), 0.0209*** (0.00567).
  - Bilateral Loans: 0.0222*** (0.00450), 0.0246*** (0.00579), 0.0304*** (0.00554), 0.0266*** (0.00453).
- Panel 5B (2SLS for EMBIG) selected coefficients:
  - IMF: -3.199** (1.381), -2.317** (1.055), -2.350** (1.022), -2.317** (1.055).
  - Multilateral Loans (excl. IMF): 0.594*** (0.177), 0.618*** (0.124), 0.686*** (0.137), 0.618*** (0.124).
  - Bilateral Loans: 0.676** (0.264), 0.581** (0.251), 0.591** (0.253), 0.581** (0.251).
  - Kleibergen-Paap rk values reported (e.g., 9.847; p-values 0.007, 0.005, 0.004, 0.005).

### Additional robustness (Tables B1–B3)
- Table B1 (Debt Crisis with additional controls): IMF coefficients across columns generally negative and significant (e.g., -0.147** (0.0602)); Multilateral Excl. IMF and Bilateral Loans positive and significant across specifications.
- Table B2 (Bond Spreads robustness): IMF negative and significant across several 2SLS columns (e.g., -2.317** (0.938)); Multilateral, Bilateral, Bond, and Commercial loan coefficients generally positive and significant in 2SLS columns.
- Table B3 (Alternative estimation methods): IV and 2SLS-GMM results qualitatively comparable to base model. Selected coefficients:
  - Debt Crisis dependent variable: IMF: -0.0255* (0.0137); -0.0296** (0.0139). Multilateral Loans (excl. IMF): 0.00244** (0.00112); Bond Loans: 0.00289** (0.00136).
  - Bonds Spreads dependent variable: IMF: -2.317*** (0.893); -2.861** (1.259). Multilateral Loans (excl. IMF): 0.618*** (0.207); Bilateral Loans: 0.581*** (0.123); Commercials Loans: 0.248*** (0.0740).
  - Kleibergen-Paap rk statistics and p-values reported across specifications (e.g., 12.829; P-val 0.001).

### Main conclusions and policy-relevant implications
- De facto seniority hierarchy via RPID:
  - Multilateral creditors (IMF, World Bank) are on average most senior.
  - Official bilateral creditors are next in seniority.
  - Private creditors (bondholders, commercial banks) are on average treated as junior.
- Creditor composition matters for sovereign risk:
  - Higher indebtedness to non-IMF creditors correlates with higher likelihood of debt crises and wider sovereign spreads.
  - IMF lending (as share of GNI) is robustly and negatively associated with the probability of a debt crisis and with sovereign bond spreads.
  - The stabilizing effect of IMF lending diminishes as external debt stocks rise; strongest effects observed for indebtedness below 70 percent of GDP, with marginal effects declining at higher debt thresholds.
- Interpretation caveats:
  - IV strategy for IMF lending provides credible identification of a stabilizing effect but residual confounding cannot be entirely ruled out.
  - For other creditor categories (multilateral excl. IMF, bilateral, private), endogeneity and adverse-selection concerns remain; estimates are conditional associations unless exogenous variation is exploited.
  - Positive association between non-IMF official lending and spreads likely reflects selection (countries in fragile conditions attract official financing), not necessarily a causal effect of such lending on market perceptions.
- Mechanisms:
  - Preferred creditor status of IMF and multilaterals can support liquidity provision, signaling, and policy credibility; IMF conditionality and program engagement can act as commitment devices affecting repayment dynamics, maturities, haircuts, and market re-entry.
  - Private creditors’ junior status feeds into higher ex ante risk premia, stricter lending conditions, and potential market withdrawal, amplifying sudden-stop dynamics.

*Source: Appendix A–B, "Creditors’ Seniority and Sovereign Risk: Evidence from Debt Composition," Working Paper No. WP/2026/161.*

### Appendix A .............................................................................................................

### Appendix A

### Contents of Appendix A
- Table A1: Overview of control variables
- Table A2: Descriptive statistics
- Table A3: Share of Senior Debt and the Sovereign Debt Crisis
- Table A4: Share of Senior Debt and Borrowing Costs
- Figure A1: Analysis of Creditors’ RPIDs in Countries with an IMF Program
- Figure A2: Analysis of Creditors’ RPIDs in Countries During Financial Crises
- Figure A3: Percentile distribution of external debt stock as % of GNI

### Context and scope (from main paper introduction)
- Global debt as share of GDP peaked at 258 percent of GDP in 2020 during the COVID emergency, and decreased afterwards, remaining however at levels higher than pre-COVID.
- The study exploits disaggregated debt data from 119 emerging markets and developing economies from 1980 to 2022.
- The appendix supports empirical analysis on the de facto seniority structure of debt and how debt composition by creditor type impacts:
  - Countries’ likelihood to experience debt crises.
  - Countries’ ability to tap international capital markets.

### Key empirical findings referenced in the appendix material
- On average the IMF and the World Bank are the most senior creditors.
- Sovereigns facing repayment difficulties are more likely to default on private claims, such as bondholders or commercial banks, than on official bilateral creditors.
- Commercial banks are the least likely to see their repayments prioritized.
- The seniority of bilateral creditors aligns with Paris Club comparability of treatment principles, the G20 Common Framework, and their role during crises.
- A positive conditional correlation exists between debt from other creditor groups and the likelihood of a debt crisis, but:
  - This is not true for an increase in debt outstanding to the IMF as share of Gross National Income (GNI).
  - The mitigating effect of IMF loans on the likelihood of debt crisis remains significant for countries with debt stock surpassing 100 percent.
- An increase in borrowing from the IMF is associated with an improvement in market access proxies as measured by bond spreads.
- An increase in borrowing from more junior creditors is associated with an increase in bond spreads.
- Estimation of these impacts may be affected by selection bias; an instrumental variable strategy is used for IMF borrowing but addressing bias for other multilateral creditors and lending types is more complex and left for further research.

### Role of Appendix A tables and figures in supporting analysis
- Tables A1–A4 provide:
  - Control variables, descriptive statistics, and empirical results linking share of senior debt with sovereign debt crises and borrowing costs.
- Figures A1–A3 provide:
  - Visual analysis of creditors’ RPIDs in IMF program countries, during financial crises, and the percentile distribution of external debt stock as percent of GNI.

*Source: Appendix A (contents and supporting information) of the IMF working paper "Creditors Composition and Sovereign Risk: Evidence from Debt Composition."*

### Section 6 concludes by summarizing our main findings.

### Section 6 concludes by summarizing our main findings.

### Data and measurement
- Sample and period:
  - Annual data spanning 1980 to 2022 covering 119 emerging market and developing economies.
- Creditor categories (World Bank IDS classification):
  - (i) bilateral creditors
  - (ii) multilateral institutions, excluding the IMF
  - (iii) bondholders
  - (iv) long-term commercial bank loans and credits (including syndicated bank loans and supplier credits)
  - (v) the International Monetary Fund (IMF)
- Key data sources:
  - Debt stocks: World Bank’s International Debt Statistics (IDS).
  - Debt defaults: BoC–BoE sovereign default database.
  - Sovereign risk measures: Laeven and Valencia (2020) debt-crisis dummy; J.P. Morgan EMBIG bond-yield spreads; sovereign ratings averaged from Moody’s, Fitch, and S&P standardized to a 21 point scale.
- Variable controls included:
  - GDP per capita (log), GDP growth, VIX index, short-term debt to total external debt ratio, dummy for prior debt restructurings, banking-crisis dummy, FDI flows, among others.
- Units and normalization:
  - Baseline results use creditor-specific debt as a share of GNI. Robustness checks use creditor shares in total debt.

### Seniority structure (RPID) — measurement and stylized facts
- RPID definition and interpretation:
  - RPID measures, for country i at time t, the difference between default-to-debt ratio of creditor group k and that of the entire creditor portfolio; range is [-1, 1].
  - Negative RPID: preferential treatment (favored creditor, fewer defaults than average).
  - Positive RPID: less preferential treatment (higher default rates than average).
- Main stylized facts:
  - The IMF and the World Bank display the most negative RPID values, confirming their seniority.
  - Official bilateral creditors display negative RPID values on average, ranking behind multilateral creditors but ahead of private creditors.
  - Private creditors — bondholders and commercial banks — display positive RPID values, indicating they bear a higher-than-average share of defaults.
  - On average commercial banks are treated as less senior compared to bondholders.
- Historical and episodic patterns:
  - First wave of overlaps in creditor treatment: 1980s–early 1990s (Latin America and some low-income countries; e.g., Peru 1985–1993, Vietnam 1985–1993, Zambia 1986–1994).
  - Second sequence: 1990s–early 2000s (political transition / post-conflict; e.g., Bosnia and Herzegovina 1992–2000, Iraq 1990–2005, Liberia 1998–2008).
  - Recent phase: 2010s–2020s with prolonged distress (e.g., Sudan 1981–2021, Somalia 2005–2020, Zimbabwe 2003–2022).
- Robustness by subsample:
  - Across subsamples (participation in IMF programs; episodes of financial crises), IMF and World Bank consistently have the lowest RPIDs; bilateral creditors remain privileged relative to private creditors.

### Creditor composition and sovereign risk — empirical strategy
- Main econometric approaches:
  - Probit / linear probability models for debt-crisis likelihood using creditor-specific debt shares (% of GNI).
  - Instrumental-variable (IV) strategy for IMF lending to address endogeneity.
  - Market-access analysis: OLS and Two-Stage Least Squares (2SLS) for sovereign bond spreads (EMBIG).
- Baseline linear probability model (illustrative form):
  - Pr(Debt Crisis = 1) = β1 IMFi,t + β2 Multilaterali,t + β3 Bilateriali,t + β4 Bondholdersi,t + β5 CommercialBanki,t + δ X i,t + ηi + μt + εi,t
  - Creditor shares expressed as share of GNI (note in Table A4 complementary specification expresses exposures as share of total debt).
- Instrument for IMF lending (IV):
  - IVi,t = Historical_IMF_Loani,t × IMF_Liquidityt
  - Historical_IMF_Loan: average annual IMF loan-to-GDP ratio over previous 10 years.
  - IMF_Liquidity: ratio of the Fund’s liquid assets to its liquid liabilities.

### Key empirical results — debt crises (Table 1 and IV results)
- Baseline Probit marginal-effect highlights (Table 1; creditor shares as % of GNI; robust SEs in parentheses; significance levels preserved):
  - IMF: -0.00815*** (0.00209)  and  -0.01315*** (0.00333) in specifications shown.
  - Multilateral Loans (excl. IMF): 0.00333*** (0.00067) and 0.00295*** (0.00074).
  - Bilateral Loans: 0.00437*** (0.00098) and 0.00433*** (0.00108).
  - Bond Loans: 0.00239** (0.00107) and 0.00286** (0.00123).
  - Commercial Bank Loans: -0.00039 (0.00125) and -0.00115 (0.00152) — not significant.
- Interpretation of baseline associations:
  - A one percentage point increase in the share of bonds (as % of GNI) associated with a rise of between 0.24 and 0.29 percentage points in crisis probability (textual marginal-effect summary).
  - A one percent increase in multilateral loans associated with a 0.30 percentage points increase in crisis probability (text).
  - A one percent increase in bilateral loans associated with a 0.43 percentage points increase in crisis probability (text).
  - IMF credit outstanding associated with a reduction in default probability with estimated effects ranging between -0.81 and -1.32 percentage points (textual summary of columns (1) and (6) in baseline discussion).
- IV-Probit / IV estimates (Table 2; IMF instrumented; robust SEs in parentheses):
  - IMF (IV results): -0.0654* (0.0380), -0.173*** (0.0488), -0.147** (0.0602) across columns (1)–(3).
  - Multilateral Loans (excl. IMF): 0.0319*** (0.00328) and 0.0217*** (0.00574) where reported.
  - Bilateral Loans: 0.0262*** (0.00454) and 0.0222*** (0.00484).
  - Bonds: 0.00689 (0.00625) and 0.0163** (0.00799).
  - Commercial Bank Loans: 0.0166* (0.00867) and -0.0101 (0.00785).
  - Control variables (IV results): Log of GNI -1.153*** (0.353) and -0.942*** (0.306); VIX Index 1.106*** (0.263) and 1.076*** (0.242); Short-Term Debt % 0.0127*** (0.00344) and 0.0178*** (0.00420); Debt Restructuring_Past 0.620*** (0.0525) and 0.590*** (0.0490); Banking Crises 0.372** (0.166) and 0.408** (0.167); Foreign Direct Investment -0.0135** (0.00584) and -0.0150** (0.00625).
- Non-linear / indebtedness-threshold findings (Table 3 and associated discussion):
  - IMF lending as share of GNI reduces crisis probability across subsamples but the marginal effect diminishes as debt stock rises.
  - Debt-stock percentiles and ranges used:
    - Lower percentiles (10 to 40): debt burden between 18% and 40%.
    - Intermediate percentiles (40th–75th): external debt between 43% and 68% of GDP.
    - Upper percentiles (75–100): debt stocks ranging from 68% to 159% of GDP.
  - Interaction terms in Table reporting (selected coefficients, preserved):
    - IMF % of GNI: -0.0489*** (0.0161), -0.135*** (0.0337), -0.125*** (0.0258), -0.130*** (0.0236), -0.0840*** (0.0232) across columns.
    - IMF % of GNI × Stock Debt<70%: -0.0763*** (0.0295).
    - IMF % of GNI × Stock Debt>70%: 0.0914** (0.0404).
    - IMF % of GNI × Stock Debt>90%: 0.0919*** (0.0342).
    - IMF % of GNI × Stock Debt>100%: 0.110*** (0.0312).
    - IMF % of GNI × Stock Debt>120%: 0.0884** (0.0451).

### Key empirical results — sovereign bond spreads (EMBIG)
- Estimation framework:
  - First-stage: IMF_loansi,t = α0 + α1 IVi,t + α X′i,t + ηi + μt + εi,t
  - Second-stage: EMBIGit = β0 + β1 IMF_loans_hatit + Σc θc Creditori,t(c) + β X′i,t + ηi + μt + εit
- Baseline OLS and preferred 2SLS (Table 4; robust SEs in parentheses; significance preserved):
  - IMF:
    - OLS: -0.995** (0.392)
    - 2SLS: -2.317** (0.938)
  - Multilateral Loans (excl. IMF):
    - OLS: 0.514*** (0.115)
    - 2SLS: 0.618*** (0.156)
  - Bilateral Loans:
    - OLS: -0.0432 (0.0898)
    - 2SLS: 0.581*** (0.216)
  - Bond_Loans:
    - OLS: 0.00308 (0.0540)
    - 2SLS: 0.147* (0.0769)
  - Commercials Bank Loans:
    - OLS: -0.0586 (0.0868)
    - 2SLS: 0.248** (0.111)
  - Other notable controls (2SLS column):
    - Log_GNI: -5.641** (2.413)
    - GDP Growth: -0.00341 (0.170)
    - Debt Crises: 4.823*** (1.048)
    - Currency Crises: 2.346** (0.987)
    - Capital Account Openness: -1.609*** (0.475)
    - Short-Term Debt % GNI: 0.407*** (0.121)
  - Kleibergen-Paap rk (instrument relevance): 12.829; P-val 0.001.
- Interpretation:
  - IMF credit outstanding is negatively and significantly associated with sovereign spreads (2SLS larger in magnitude than OLS).
  - Coefficients on non-IMF multilateral and bilateral loans are positive and significant in 2SLS — interpreted as adverse-selection (countries relying more on official lenders often have higher underlying risk), not necessarily causal.
  - Bond and commercial-bank loan coefficients become positive and significant in 2SLS, consistent with private-creditor exposure associating with higher spreads.

### Robustness checks
- Alternative samples (Table 5):
  - Excluding 1980–1990; excluding Cold War years and USSR collapse; excluding 2008–2009 global financial crisis; excluding Covid-19 pandemic years.
  - Results remain broadly stable:
    - IMF lending remains negatively associated with debt-crisis probability.
    - Multilateral (excl. IMF) and bilateral loans remain positively associated with crisis risk.
    - EMBIG regressions: IMF credit outstanding negatively associated with spreads; other official exposures positively associated with spreads (caution on endogeneity).
  - Panel 5A (IVProbit) selected coefficients:
    - IMF: -0.245*** (0.0709), -0.171*** (0.0526), -0.111* (0.0580), -0.146** (0.0599) across columns.
    - Multilateral Loans (excl. IMF): 0.0298*** (0.00607), 0.0235*** (0.00542), 0.0189*** (0.00483), 0.0209*** (0.00567).
    - Bilateral Loans: 0.0222*** (0.00450), 0.0246*** (0.00579), 0.0304*** (0.00554), 0.0266*** (0.00453).
  - Panel 5B (2SLS for EMBIG) selected coefficients:
    - IMF: -3.199** (1.381), -2.317** (1.055), -2.350** (1.022), -2.317** (1.055).
    - Multilateral Loans (excl. IMF): 0.594*** (0.177), 0.618*** (0.124), 0.686*** (0.137), 0.618*** (0.124).
    - Bilateral Loans: 0.676** (0.264), 0.581** (0.251), 0.591** (0.253), 0.581** (0.251).
    - Kleibergen-Paap rk values reported (e.g., 9.847; p-values 0.007, 0.005, 0.004, 0.005).
- Additional controls appended in robustness (Table B1):
  - Institutional quality, international reserves, population density, debt service as % of GNI, terms of trade.
  - Results for creditor-category effects remain close to base model; institutional quality tends to reduce crisis risk (not always significant); international reserves positively influence financial stability; terms of trade matter for sovereign risk.
- Alternative estimation methods:
  - 2SLS-GMM with main instrument (Historical IMF Credit × IMF Liquidity) and lags of endogenous variables — results qualitatively comparable to base model (Table B3).

### Main conclusions and implications
- De facto seniority hierarchy documented via RPID:
  - Multilateral creditors (IMF, World Bank) are on average most senior.
  - Official bilateral creditors are next in seniority.
  - Private creditors (bondholders, commercial banks) are on average treated as junior.
- Creditor composition matters economically for sovereign risk:
  - Higher indebtedness to non-IMF creditors correlates with higher likelihood of debt crises and wider sovereign spreads.
  - IMF lending (as share of GNI) is robustly and negatively associated with the probability of a debt crisis and with sovereign bond spreads.
  - The stabilizing effect of IMF lending diminishes as external debt stocks rise; strongest effects observed for indebtedness below 70 percent of GDP, with marginal effects declining at higher debt thresholds.
- Interpretation caveats and limitations:
  - For IMF lending, the IV strategy provides a credible identification of a stabilizing effect, but residual confounding cannot be entirely ruled out.
  - For other creditor categories (multilateral excl. IMF, bilateral, private), endogeneity and adverse-selection concerns remain; estimates are conditional associations unless exogenous variation for these creditors can be exploited.
  - Positive association between non-IMF official lending and spreads likely reflects selection (countries in fragile conditions attract official financing), not necessarily a causal effect of such lending on market perceptions.
- Broader economic mechanisms highlighted:
  - Preferred creditor status of IMF and multilaterals can support liquidity provision, signaling, and policy credibility; IMF conditionality and program engagement can act as commitment devices that affect repayment dynamics, maturities, haircuts, and market re-entry.
  - Private creditors’ junior status feeds into higher ex ante risk premia, stricter lending conditions, and potential market withdrawal, amplifying sudden-stop dynamics.

*Source: wpiea2026161-source-pdf - Section 6 concludes by summarizing our main findings.*

### Appendix A

### Appendix A

### Control variables (Table A1)
- IMF Credit Outstanding (% of GNI): IDS
- Multilateral Debt (excl. IMF) (% of GNI): IDS
- Bilateral Debt (% of GNI): IDS
- Bondholders Debt (% of GNI): IDS
- Commercial Debt (% of GNI): IDS
- External Debt to GNI: "Total external debt is debt owed to nonresidents repayable in foreign currency, goods, or services. It is the sum of public, publicly guaranteed, and private nonguaranteed long-term debt, short-term debt, and use of IMF credit." Source: IDS
- Sovereign Default: "Total amount of defaults with each creditor" Source: BoC-BoE sovereign default database
- IMF Liquidity: "IMF liquidity ratio equals liquid resources divided by liquid liabilities. We use the average IMF liquidity ratio from the past five years." Source: IMF’s Annual Reports
- Log of GNI: Logarithm of Gross National Income. Source: WDI
- Growth: Annual GDP Growth Rate (%). Source: WDI
- Debt Restructuring_Past: "Total Number of Debt Restructurings in the Last 5 Years" Authors' calculation based on Asonuma and Trebesch (2016), Paris Club on Private and Official debt restructuring
- Banking Crises: Binary =1 if Banking Crisis is Ongoing. Source: Reinhart & Rogoff (2009)
- Debt Crises: Binary =1 if Sovereign Debt Crisis is Ongoing. Source: Reinhart & Rogoff (2009)
- Market Access: Access to International Financial Markets (1=yes, 0=no). Calculated by the authors based on data from Moody's, Fitch, Standard & Poor's (S&P)
- EMBIG Spreads: The J.P. Morgan Emerging Market Bond Index Global (EMBIG) index. Source: J.P. Morgan
- IMF Program: Participation in an IMF Program (1=yes, 0=no). Source: IMF, Dreher (2012)
- Trade openness: "Trade openness is defined as the sum of a country’s exports and imports as a fraction of GNI in percent." Source: WDI
- VIX: Market Vulnerability Index (VIX). Source: Chicago Board Options Exchange

### Descriptive statistics (Table A2)
- IMF  (% of GNI): Observations 4,098; Mean 2.564239; Std. Dev. 4.977649; Min 0; Max 69.97369
- Multilateral Loans (excl. IMF) (% of GNI): Observations 4,349; Mean 18.58857; Std. Dev. 21.33003; Min 0; Max 319.9574
- Bilateral Loans (% of GNI): Observations 4,357; Mean 18.35796; Std. Dev. 33.52719; Min 0; Max 652.4404
- Bond Loans (% of GNI): Observations 4,625; Mean 2.92994; Std. Dev. 7.56168; Min 0; Max 144.8
- Commercials Bank loans (% of GNI): Observations 4,998; Mean 2.835316; Std. Dev. 7.682194; Min 0; Max 162.8461
- Log of GNI: Observations 4,436; Mean 22.96832; Std. Dev. 2.141282; Min 17.94961; Max 30.50858
- Growth: Observations 4,620; Mean 3.530515; Std. Dev. 6.127119; Min -64.0471; Max 88.95766
- Debt Restructuring_Past: Observations 4,403; Mean .5210084; Std. Dev. .9955383; Min 0; Max 5
- Banking Crises: Observations 4,920; Mean .0567073; Std. Dev. .231306; Min 0; Max 1
- Debt Crises: Observations 4,663; Mean .4509972; Std. Dev. .4976463; Min 0; Max 1
- EMBIG Spreads: Observations 988; Mean 599.6267; Std. Dev. 1272.046; Min 0; Max 31310.52
- IMF Program: Observations 4,998; Mean .2933173; Std. Dev. .4553282; Min 0; Max 1
- Openness: Observations 3,953; Mean 34.94235; Std. Dev. 17.47341; Min .0104996; Max 173.9983
- VIX: Observations 6,460; Mean 19.60173; Std. Dev. 5.762154; Min 11.09024; Max 32.69553

### Regression results — Share of Senior Debt and the Sovereign Debt Crisis (Table A3)
- Dependent variable: Debt Crises. Observations: 999 in each specification. Year FE: Yes. Country FE: Yes.
- Key coefficient estimates (robust standard errors in parentheses):
  - Share of IMF Debt:
    - Column (1): -0.00802*** (0.00184)
    - Column (2): -0.00680*** (0.00169)
  - Share of Senior_Creditors Debt (Excl. IMF):
    - Column (2): 0.00269** (0.00109)
    - Column (3): 0.00221** (0.00109)
  - Share of Senior_creditors_debt (Incl. IMF):
    - Column (4): 0.0102 (0.00653)
  - Log_of_GNI:
    - Column (1): -0.29585*** (0.06136)
    - Column (2): -0.24321*** (0.06084)
    - Column (3): -0.26464*** (0.05990)
    - Column (4): -1.458*** (0.426)
  - GDP Growth:
    - Column (1): -0.00778** (0.00319)
    - Column (2): -0.00787** (0.00313)
    - Column (3): -0.00821*** (0.00310)
    - Column (4): -0.0437*** (0.0165)
  - VIX index:
    - Column (1): 0.34158*** (0.04064)
    - Column (2): 0.29983*** (0.04269)
    - Column (3): 0.32367*** (0.04174)
    - Column (4): 1.743*** (0.317)
  - Short-Term Debt %:
    - Column (1): 0.00498*** (0.00127)
    - Column (2): 0.00430*** (0.00125)
    - Column (3): 0.00422*** (0.00127)
    - Column (4): 0.0273*** (0.00796)
  - Debt Restructuring_Past:
    - Column (1): 0.15553*** (0.01959)
    - Column (2): 0.14772*** (0.02070)
    - Column (3): 0.15285*** (0.02021)
    - Column (4): 0.855*** (0.0941)
  - Banking Crises:
    - Column (1): 0.00451 (0.04005)
    - Column (2): 0.00247 (0.04182)
    - Column (3): 0.01743 (0.04192)
    - Column (4): -0.0521 (0.236)
  - Foreign Direct Investment:
    - Column (1): 0.00088 (0.00301)
    - Column (2): 0.00028 (0.00301)
    - Column (3): 0.00039 (0.00305)
    - Column (4): 0.00271 (0.0172)

- Significance notation: *** p<0.01, ** p<0.05, * p<0.1.

### Regression results — Share of Senior Debt and Borrowing Costs (Table A4)
- Dependent variable: EMBIG. Observations: 102. R-squared: 0.817. Kleibergen-Paap rk: 11.232. Kleibergen-Paap rk (P-val): 0.0036. Country FE: Yes. Year FE: Yes.
- Key coefficient estimates (robust standard errors in parentheses):
  - Share of IMF Debt: -0.395* (0.216)
  - Share of Senior_Creditors Debt (Excl. IMF): -0.0684*** (0.0224)
  - Log_of_GNI: 1.097 (1.616)
  - GDP Growth: -0.150 (0.0994)
  - Banking Crises: -0.798 (1.080)
  - Debt Restructuring_Past: 0.291 (0.576)
  - Debt Crises: 5.496*** (1.791)
  - Currency Crises: 1.378* (0.750)
  - Trade Openness: 0.122 (0.119)
  - Capital Account Openness: -0.865*** (0.321)
  - Short-Term Debt % GNI: 0.0817 (0.129)
  - Broad Money % GDP: -0.0733 (0.0446)
  - Government Final Consumption: -0.0108 (0.0343)
  - Constant: -10.03 (36.62)

### Figures (Figures A1–A3)
- Figures A1–A2: "Analysis of Creditors’ RPIDs in Countries with an IMF Program" and "in Countries During Financial Crises." Source: Authors’ calculations; based on data from the World Bank’s International Debt Statistics (IDS) and the Bank of Canada–Bank of England Sovereign Default Database.
- Figure A3: "Percentile distribution of external debt stock as % of GNI." Source: Authors’ calculations. Data from the World Bank’s International Debt Statistics (IDS).

---

### Appendix B — Robustness tests

### Additional controls: Debt Crisis (Table B1)
- Dependent variable: Debt Crisis. Country Fixed Effect: Yes. Year Fixed Effect: Yes.
- Selected coefficient estimates (robust standard errors in parentheses):
  - IMF:
    - (1): -0.147** (0.0602)
    - (2): -0.147** (0.0599)
    - (3): -0.143** (0.0601)
    - (4): -0.138*** (0.0534)
    - (5): -0.140** (0.0605)
    - (6): -0.142** (0.0579)
  - Multilateral Excl. IMF:
    - (1): 0.0217*** (0.00574)
    - (2): 0.0217*** (0.00572)
    - (3): 0.0215*** (0.00578)
    - (4): 0.0218*** (0.00614)
    - (5): 0.0214*** (0.00551)
    - (6): 0.0224*** (0.00570)
  - Bilateral Loans:
    - (1): 0.0222*** (0.00484)
    - (2): 0.0223*** (0.00484)
    - (3): 0.0218*** (0.00474)
    - (4): 0.0225*** (0.00525)
    - (5): 0.0228*** (0.00513)
    - (6): 0.0217*** (0.00485)
  - Bond Loans:
    - (1): 0.0163** (0.00799)
    - (2): 0.0165** (0.00795)
    - (3): 0.0159** (0.00800)
    - (4): 0.0129 (0.00881)
    - (5): 0.0171** (0.00850)
    - (6): 0.0164** (0.00802)
  - Commercials Loans: coefficients around -0.0101 to -0.00876, none significant at conventional levels in these columns.
  - Additional controls included across specifications: Quality of Institutions (-0.0855 (0.0701) in one column), International Reserves (-0.000497* (0.000264)), Population density (0.00180 (0.00124)), Total debt service % GNI (-0.00146 (0.00983)), Terms of Trade (-0.00668** (0.00330)).
- Observations vary across columns: 2,869; 2,865; 2,682 depending on specification.
- Constants reported (e.g., -2.352 (1.502) in column (1); 33.79*** (12.57) in column (4)).

### Additional controls: Bond Spreads (JPM EMBIG) (Table B2)
- Dependent Variable: Bonds Spreads (JPM EMBIG).
- Selected coefficient estimates (robust standard errors in parentheses):
  - IMF:
    - (1): -2.317** (0.938)
    - (2): -2.263** (0.939)
    - (3): -2.882*** (1.005)
    - (4): -1.630** (0.703)
    - (5): -2.238*** (0.827)
    - (6): -2.719** (1.071)
  - Multilateral Loans (excl. IMF): coefficients positive and significant across columns, e.g., 0.618*** (0.156) in (1).
  - Bilateral Loans: positive and significant across columns, e.g., 0.581*** (0.216) in (1).
  - Bond Loans: positive, e.g., 0.147* (0.0769) in (1); up to 0.191** (0.0847) in (3).
  - Commercial Loans: positive and significant across columns, e.g., 0.248** (0.111) in (1).
  - Quality of Institutions: -3.070*** (1.181) in one specification.
  - Terms of Trade: 0.0904** (0.0396) in one specification.
  - Constant values reported, e.g., 135.1** (62.27) in (1); 290.9*** (103.0) in (6).

### Alternative estimation methods (Table B3)
- Observations: 210 across methods. Country FE: Yes. Year FE: Yes.
- R-squared reported across specifications: 0.602, 0.617, 0.549, 0.658, 0.609, 0.578.
- Kleibergen-Paap rk statistics and p-values reported: e.g., Kleibergen-Paap rk 12.829 with P-val 0.001; other values include 16.859 (0.000), 16.588 (0.000), 11.826 (0.002), 12.131 (0.002), 12.829 (0.001).
- Selected coefficient excerpts (standard errors in parentheses):
  - Debt Crisis dependent variable:
    - IMF: -0.0255* (0.0137) in one IV specification; -0.0296** (0.0139) in another.
    - Multilateral Loans (excl. IMF): 0.00244** (0.00112); 0.00356*** (0.00118).
    - Bond Loans: 0.00289** (0.00136); 0.00313* (0.00163).
    - Commercials Loans: -0.00274* (0.00161); -0.00214 (0.00185).
    - Log_GNI: -0.271*** (0.0426); -0.270*** (0.0436).
    - VIX Index: 0.279*** (0.0407); 0.282*** (0.0415).
    - Short-Term Debt: 0.00314*** (0.000977); 0.00344*** (0.00101).
    - Debt Restructuring_Past: 0.144*** (0.0125); 0.145*** (0.0126).
    - Banking Crises: 0.0807** (0.0321); 0.0803** (0.0325).
  - Bonds Spreads dependent variable:
    - IMF: -2.317*** (0.893) in one method; -2.861** (1.259) in another.
    - Multilateral Loans (excl. IMF): 0.618*** (0.207); 0.464** (0.188).
    - Bilateral Loans: 0.581*** (0.123); 0.652*** (0.161).
    - Bond Loans: 0.147** (0.0680); 0.219** (0.0997).
    - Commercials Loans: 0.248*** (0.0740); 0.297*** (0.0948).
    - Log_GNI: -5.641** (2.342); -5.725** (2.438).
    - Debt Restructuring_Past: -2.317*** and -2.861*** reported in table context as coefficients in some columns (standard errors reported elsewhere).

- Standard errors in parentheses. Significance notation: *** p<0.01, ** p<0.05, * p<0.1.

---

*Source: Appendix A–B and References, "Creditors’ Seniority and Sovereign Risk: Evidence from Debt Composition," Working Paper No. WP/2026/161.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026161-source-pdf.pdf_