## wpiea2021020-print-pdf

## Source details

**Canonical URL:** [wpiea2021020-print-pdf](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021020-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021020-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021020-print-pdf.pdf.json)

---

### I. Introduction — scope and purpose
- Global financial integration: sum of gross external assets and liabilities has more than tripled as a share of GDP since the early 1990s.
- Objective: examine how size and structure of the international investment position (IIP) affect the likelihood of:
  - an external crisis (sovereign default or restructuring, or recourse to IMF financing), or
  - a sudden stop with high adverse growth impact.
- Structural IIP features analyzed:
  - gross versus net imbalances,
  - equity versus debt positions,
  - currency composition of external debt,
  - role of official and non-official foreign assets.
- Other factors examined: current account balance, exchange rate overvaluation, fiscal position, credit cycle, global risk aversion.
- Sample and methods:
  - Sample of 73 advanced and emerging economies.
  - Period 1991–2018 (currency composition data available from 1991).
  - Empirical approaches: event analysis and pooled probit models.
- Data sources:
  - IIP, foreign official reserves, current account, nominal GDP: updated External Wealth of Nations database (EWN; Lane and Milesi-Ferretti, 2007).
  - Foreign currency shares of external liabilities: Bénétrix et al. (2019).

### II. Key empirical findings — summary
- Predictive role of net IIP:
  - High negative predictive power of net IIP for external crises/sudden stops is predominantly driven by gross debt liability position.
  - Equity counterparts and non-official debt assets are not statistically significant or robust across specifications.
- Currency composition:
  - Likelihood of external crisis or sudden stop increases with the size of foreign currency denominated debt, especially for EMDEs.
- Foreign official reserves:
  - Reserves mitigate external crisis risk with a non-linear impact and diminishing returns.
- Macro and global factors:
  - Higher current account deficits increase external crisis likelihood.
  - Overvalued exchange rates, fiscal deficits, and excessive credit increase sudden stop likelihood.
  - Higher global risk aversion increases likelihood of both external crises and sudden stops.
- Interaction of vulnerabilities:
  - Large FX-denominated debt, large current account deficits, and low reserves together markedly increase crisis/sudden-stop likelihood, especially during higher global risk aversion.
- Robustness:
  - Main findings survive checks across country sample changes, crisis definitions, and additional controls.

### III. Event-study evidence — pre- and post-crisis IIP dynamics
- Event-study specification includes country and time fixed effects.
- Pre-crisis dynamics:
  - Deterioration of net IIP ahead of external crises is driven mainly by sharp increases in foreign-currency-denominated external debt liabilities (valuation effects from currency depreciation and lower real GDP growth).
  - Foreign-currency external debt assets also rise ahead of crisis (private capital flight, valuation effects).
  - Foreign exchange reserves decline sharply just ahead of crisis.
  - Equity liabilities fall ahead of crisis (mitigating role).
  - Domestic-currency external indebtedness rises much less.
  - Net FX-denominated assets (assets minus liabilities) tend to decrease before crisis.
- Post-crisis dynamics:
  - Net IIP typically rises after crisis, driven by a significant drop in FX-denominated external debt liabilities (deleverage, restructuring, valuation).
  - Reserves decline precedes crisis and continues afterward.
- Large external crises (IMF assistance exceeding 200 percent of quota):
  - Much larger drop in net IIP.
  - Larger rise in FX-denominated debt liabilities and decline in official reserve assets.
  - IIP components rebound more sharply but end below pre-crisis peaks.

### IV. Probit estimation — determinants of external crisis likelihood
- Model: pooled probit; dependent variable = external crisis occurrence (country-year).
- Key explanatory variables: IIP components (net and gross), current account (lagged two-year average), real effective exchange rate gap (5-year MA deviation), relative income per capita, VXO (global risk indicator), credit gap, financial development index interacted with relative income per capita.
- Main findings (selected, preserving coefficients/significance as reported):
  - Net IIP, share of GDP: -0.34* and -0.27* (standard errors (0.19)(0.15))
  - Net debt, share of GDP: -0.55*** and -0.50*** (0.21)(0.16)
  - Net foreign currency debt: -0.61** and -0.52*** (0.25)(0.17)
  - Foreign currency debt liabilities: 0.62** and 0.44** (0.28)(0.21)
  - Foreign exchange reserves, share of GDP: -7.22***, -7.21***, -7.02***, -6.13***, -6.20***, -5.22*** (1.43)(1.44)(1.65)(1.65)(1.70)(1.81)
  - Current account/GDP, 2yr MA: -5.32***, -5.71***, -5.73***, -6.89*** (1.52)(1.54)(1.53)(1.50)
  - VXO: 0.02** (0.01) across multiple specifications
  - Financial development index and its interaction with income per capita show significance in some specifications (examples: -4.96*** and Fin. develop. * income per capita 5.76***).
- Model fit and diagnostics:
  - Observations: 1,828 in main specifications.
  - Number of ifs_code: 73.
  - Log pseudolikelihood examples: -410.38 to -333.29 across specifications.
  - AUROC estimates: 0.66, 0.67, 0.72, 0.82, 0.83, 0.84, 0.84, 0.87.
- EMDE subsample highlights:
  - Impact of FX-denominated debt liabilities increases (example coefficient 1.78*** (0.33) in EMDEs-only column).
  - Reserves play a stronger mitigating role.
  - Crisis risks rise more with higher current account deficits and higher VXO.
  - Financial development and income per capita not significant in EMDE-only sample.
- Robustness:
  - FX-denominated debt liabilities remain statistically significant with stable magnitude across many robustness checks (samples excluding oil exporters, alternative crisis definitions, additional controls).
  - Additional controls tested include exchange rate regime, capital account openness, commodity price changes, fiscal balance; only commodity prices significant among these.

### V. Probit margins — nonlinear effects and economic significance
- Predicted probabilities computed holding covariates at sample means and varying variable of interest.
- EMDE sample:
  - Increase in FX-denominated debt liabilities from 40 percent of GDP (near EMDE median) to 60 percent of GDP → 5-percentage point increase in predicted probability of external crisis.
  - Decline in current account balance from surplus of 5 percent of GDP to deficit of 5 percent of GDP → 5.3 percentage points increase in predicted probability of external crisis for EMDEs.
  - Reserve thresholds: predicted external stress probability near zero when reserves >30 percent of GDP.
  - Decline in FX reserves from 20 percent to 10 percent of GDP → 6.5 percentage points increase in crisis likelihood.
  - Decline from 10 percent to 0 percent of GDP → additional 12.6 percentage points increase.
  - Doubling global risk aversion from historical mean → increases predicted crisis probabilities by 6.5 percentage points; a further doubling to peaks seen in early GFC and Covid-19 → increases by 23.5 percentage points.
- Full sample:
  - Same FX-debt rise from 40 to 60 percent of GDP → 0.2 percentage points increase in crisis probability.
  - Current account swing 5 percent surplus to 5 percent deficit → 1.1 percentage points rise in predicted probability for full sample.
  - Reserve-decline impacts: 0.7 and 2.1 percentage points for declines described above.
  - Global risk aversion increases: 0.7 and 4 percentage points for the two doublings described above.
- Nonlinearities: reserve effects and interactions with global risk aversion are markedly nonlinear, with larger marginal effects at low reserve levels in EMDEs.

### VI. Interaction of vulnerabilities and global risk shocks
- Multiple vulnerabilities amplify crisis probability nonlinearly in EMDEs.
  - Example: With FX debt at 40 percent of GDP, predicted probability ranges from 2–13 percent depending on reserves and current account (75th vs 25th percentiles).
  - For current account deficit of 5 percent: predicted crisis probabilities range from 2.5 to 21 percent depending on vulnerability configuration (reserves and FX debt at quartiles).
- Global risk shock example:
  - When VXO reaches peak values seen in initial phase of Covid-19 pandemic (about 80, a value also reached during the Global Financial Crisis), external crisis probability for an EMDE with median preexisting vulnerabilities rises to about 40 percent.
  - This is more than double the estimated probability for less vulnerable EMDEs.

### VII. Output costs and macro dynamics after crises (EMDEs)
- Method: local projections (Jordà, 2005).
- Vulnerability classification: high FX debt (above EMDE median), high current account deficit (below median), low reserves (below median) = highly vulnerable.
- Post-crisis outcomes for vulnerable EMDEs:
  - Initial output loss is over 4 percent; output remains below pre-crisis levels after five years.
  - Initial exchange rate depreciation averages 10 percent.
  - Current account adjustment averages 2 percent of GDP.
- Less vulnerable EMDEs:
  - Initial losses ≈ 1 percent.
  - Output rises above pre-crisis level after four years.
- Conclusion: Post-crisis adjustment larger and recovery slower when multiple pre-existing external vulnerabilities exist.

### VIII. Sudden stops — magnitudes, drivers, distinctions from external crises
- Event analysis:
  - Sudden stops preceded by net IIP declines driven mainly by increases in FX-denominated external debt liabilities, currency depreciation, and negative real growth.
  - Reserves decline and continue falling afterward, indicating reserve sales buffer sudden stops.
  - Domestic-currency external debt liabilities are flat around sudden stops; equities follow similar path as crises.
- Quantitative magnitudes:
  - Net IIP declines by more than 40 percentage points around a sudden stop episode.
  - Net IIP declines by about 3 percentage points for all external crises and about 14 percentage points for large external crises (IMF assistance >200 percent of quota).
  - FX-denominated external debt liabilities increase by almost 80 percentage points for sudden stops, compared with 12 percentage points for external crises and 22 percentage points for large external crises.
- Probit results for sudden stops (selected coefficients preserved):
  - Net IIP, share of GDP: -0.26*** (0.07)
  - Net foreign currency debt: -0.33*** (0.09)
  - Foreign currency debt liabilities: 0.34*** (0.08)
  - Domestic currency debt liabilities: -1.40*** (0.51)
  - Income per capita, relative to US: positive and significant (examples: 4.46***; standard errors ~1.39–1.58)
  - REER gap and credit gap: positive and significant (examples: REER gap 2.89*** (0.63); credit gap 4.31*** (1.23))
  - VXO: 0.09*** (0.01)
  - Fiscal balance becomes significant: higher fiscal deficits increase sudden stop probability in some specifications (example fiscal coefficient -7.01** (3.38) in additional controls panel, noting sign conventions in table).
- Predictive margins:
  - Predicted probit margins for sudden stops are generally much lower than for external crises; sudden stops are rarer and harder to predict.
  - Increase of FX-debt liabilities from 40 to 60 percent of GDP associated with only a marginal increase in sudden-stop probability, though effects are nonlinear and larger when combined with other vulnerabilities.
  - Global risk aversion effects for sudden stops can be quantitatively similar to external crises.

### IX. Policy-relevant implications and priorities
- Size and currency composition of the IIP matter, especially for EMDEs:
  - Higher external debt increases external crisis likelihood.
  - Higher FX-denominated debt increases sudden-stop likelihood.
- Official reserves:
  - Reserves mitigate external crisis probability, with diminishing returns and markedly nonlinear effects concentrated at low reserve levels.
- Macroeconomic monitoring priorities:
  - Monitor flow and stock imbalances, including FX-denominated debt and current account positions.
  - Watch for overvalued exchange rates, excessive credit, and fiscal deficits as drivers of sudden stops.
- Policy priorities:
  - Guard against buildup of external debt, especially in foreign currency.
  - Maintain external buffers (official reserves) to manage global risk shocks.
  - Recognize that macroeconomic consequences (lost real GDP, sharp current account and REER adjustments) are larger for economies with greater preexisting vulnerabilities.
- Low-income countries:
  - Vulnerabilities relevant for LICs that have borrowed in foreign currency; data constraints limited LIC coverage and future research should study these countries.

*Source: wpiea2021020-print-pdf*

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

### wpiea2021020-print-pdf - References .............................................................................................................

### I. Introduction — scope and purpose
- Global financial integration, measured as the sum of gross external assets and liabilities, has more than tripled as a share of GDP since the early 1990s.
- The paper examines how the size and structure of the international investment position (IIP) affect the likelihood of:
  - an external crisis (defined as involving a sovereign default or restructuring, or recourse to IMF financing), or
  - a sudden stop with high adverse growth impact.
- Structural IIP features highlighted for analysis:
  - gross versus net imbalances,
  - equity versus debt positions,
  - currency composition of external debt,
  - role of official and non-official foreign assets.
- The paper also examines global and domestic factors: current account balance, exchange rate overvaluation, fiscal position, credit cycle.
- Sample and methods:
  - Sample of 73 advanced and emerging economies.
  - Period 1991–2018 (data constraints on currency composition).
  - Empirical approaches include event analysis and probit models.
- Data sources:
  - IIP (including foreign official reserves), current account, and nominal GDP from the updated External Wealth of Nations database (EWN; Lane and Milesi-Ferretti, 2007).
  - Foreign currency shares of external liabilities from Bénétrix et al. (2019).

### II. Key empirical findings
- Predictive power of net IIP:
  - The high negative predictive power of a country’s net IIP for external crises or sudden stops is predominantly driven by its gross debt liability position.
  - Equity counterparts and non-official debt assets are neither statistically significant nor robust across specifications.
- Currency composition:
  - The likelihood of an external crisis or a sudden stop increases with the size of foreign currency denominated debt, especially for emerging market and developing economies (EMDEs).
- Role of foreign official reserves:
  - Foreign official reserves act as a mitigating factor to external crisis, with a non-linear impact and diminishing returns.
- Other macro fundamentals and global factors:
  - Higher current account deficits increase the likelihood of an external crisis.
  - Overvalued exchange rates, fiscal deficits, and excessive credit increase the likelihood of a sudden stop.
  - Higher levels of global risk aversion increase the likelihood of both external crises and sudden stops.
- Interaction of vulnerabilities:
  - A combination of a large stock of foreign currency denominated debt, a large current account deficit, and low foreign official reserves can significantly increase the likelihood of an external crisis or sudden stop, particularly during periods of higher global risk aversion.
- Robustness:
  - Main findings survive robustness checks across changes in country sample, crisis definition, and inclusion of additional control variables.

### III. Contribution to literature
- Advances prior work on crisis early warning systems by:
  - Studying both external crises and sudden stops with high adverse growth impact.
  - Investigating the currency composition of external liabilities and other IIP components using a novel dataset on currency composition (Bénétrix et al., 2019).
  - Expanding the set of control variables and estimating output costs from external crises and their relationship to pre-existing vulnerabilities.
- Builds on and extends Catão and Milesi-Ferretti (2014) by:
  - Including sudden stop episodes,
  - Examining currency composition of external liabilities,
  - Disaggregating gross assets and liabilities.

### IV. Data and definitions (selection)
- Coverage:
  - 73 advanced and emerging and developing economies (see Table A.1).
  - Period: 1991–2018.
- Main datasets:
  - IIP, foreign official reserves, current account, nominal GDP: updated External Wealth of Nations database (EWN; Lane and Milesi-Ferretti, 2007).
  - Foreign currency shares of external liabilities: Bénétrix et al. (2019).
- Event definitions:
  - External crisis: sovereign default or restructuring, or recourse to IMF financing.
  - Sudden stops: sudden stops with high adverse growth impact (terms used interchangeably in the paper).

*Source: wpiea2021020-print-pdf - References .............................................................................................................*

### Appendix Table A.2 includes the source of the variables

### wpiea2021020-print-pdf - Appendix Table A.2 includes the source of the variables

### Definitions of external stress episodes
- Two types of external stress episodes are considered: external crisis and sudden stops with high output declines.
- External crisis:
  - An external crisis is an episode of either a sovereign debt default or restructuring, or an IMF-supported program (excluding precautionary and nondisbursing arrangements).
  - Sovereign debt defaults and restructuring episodes are identified based on an updated version of the data sets in Das, Papaioannou, and Trebesch (2011) and Asonuma and Trebesch (2016); and recent Paris Club reports. (http://www.clubdeparis.org/en/traitements).
  - As shown in Table A.3, 128 cases of external crisis are identified.
  - The distribution of crisis episodes is evenly spread over the time horizon (Figure 2).
- Sudden stops with high adverse growth impact:
  - Defined as episodes where net private capital inflows tighten financial constraints, resulting in unusually large recessions or recourse to IMF financial support.
  - Sudden stop episodes are defined as those where net private capital inflows are either:
    - (i) 1½ standard deviation below their mean and the annual decline is ¾ standard deviations from a year earlier (defined in real U.S. dollars); or
    - (ii) have declined by at least 3 percentage points of GDP relative to the previous year and 2 percentage points from two years earlier.
  - A large output decline episode is one where real GDP growth, relative to the average of the past 5-years, is in the bottom 5th percentile of the distribution (across time and across countries).
  - As shown in Table A.4, 32 episodes involving a combination of sudden stops and large output declines were identified.
  - Sudden stop only episodes are evenly spread over the time horizon, while those with large growth impact are concentrated (peaks during the GFC and the Asian Financial Crisis) and have been less visible in recent years (Figure 3).
- Notes:
  - The underlying capital flow data comes from the IMF’s Financial Flows Analytics database, the World Bank’s World Development Indicators database, Haver Analytics, CEIC Asia database, and CEIC China database.
  - The currency denomination of external debt assets and liabilities dataset is available starting in 1991; this restriction determines the initial year of the sample.
  - The analysis considers episodes with consecutive years; results are generally robust to an alternative episode definition (Section 4.B).

### Event analysis: pre- and post-crisis IIP dynamics
- Sample and method:
  - Sample comprises 73 advanced and emerging market economies during 1991–2018.
  - Event-study analysis controls for country and time fixed effects.
  - Specification: y_{c,t} = α_c + β_t + Σ_θ β_θ D_{t+θ} + ε_{c,t}, where α_c and β_t are country and time fixed effects; D_{t+θ} are dummy variables; y_{c,t} is the IIP component.
- Main empirical observations (Figure 4):
  - Deterioration of net IIP as a share of GDP ahead of an external crisis is driven predominantly by a sharp increase in foreign currency-denominated external debt liabilities.
    - This reflects valuation effects from currency depreciation, and lower real GDP growth in the years preceding the crisis.
  - Foreign-currency external debt assets also rise ahead of crisis, reflecting private capital flight and valuation effects.
  - Foreign exchange reserves decline sharply just ahead of the crisis.
  - Equities play a mitigating role: the value of equity liabilities falls ahead of crisis.
  - The rise in domestic-currency external indebtedness is much less pronounced.
  - On a net basis, foreign-currency-denominated assets (assets minus liabilities) tend to decrease before the crisis, implying that the rise in foreign-currency-denominated debt liabilities outstrips the rise in private capital outflows.
  - Following the crisis, net IIP typically rises, driven primarily by a significant drop in foreign-currency-denominated external debt liabilities (deleverage, restructuring, and some positive valuation effects from currency appreciation).
  - The decline in foreign official reserves precedes the crisis but continues in its aftermath.
  - Other components exhibit smaller fluctuations or remain essentially flat.
- Large external crises (IMF financial assistance exceeding 200 percent of quota):
  - Drop in net IIP is far more pronounced.
  - Driven even more by rise in foreign-currency-denominated debt liabilities and decline in official reserve assets.
  - IIP components rebound more sharply following the crisis, but tend to end well below pre-crisis peaks.

### Probit estimation results: determinants of external crisis likelihood
- Model and variables:
  - Pooled probit model; dependent variable is occurrence of an external crisis (1 indicates crisis in a given country-year).
  - Explanatory variables include IIP components and standard macroeconomic variables: current account balance, real effective exchange rate gap (deviation from average of previous five years), relative income per capita, global risk indicator (VXO), credit gap, and financial development index (market depth, access, efficiency) interacted with relative income per capita.
  - All variables, except global risk aversion, are lagged one year.
  - Regression output reports log pseudolikelihood and AUROC.
- Key empirical findings (Table 1 and robustness checks):
  - A negative net IIP position is associated with a higher likelihood of an external crisis.
  - Disaggregating net IIP:
    - Net external debt and net foreign official reserves affect the probability of external crises.
    - Net equity position does not affect crisis probability.
  - Disaggregating net external debt by currency:
    - Only net foreign-currency-denominated debt component becomes significant.
  - Focusing on gross IIP components:
    - Foreign-currency denominated debt liabilities and foreign official reserves become the most robust and significant determinants.
  - Model fit:
    - The most disaggregated model with gross positions and additional controls fits best (based on log-likelihood and AUROC).
  - Among standard macro variables:
    - The current account balance (included as the lagged two year average) and global risk aversion (VXO) are the most robust determinants and are always significant when included.
    - Financial development index is significant when net or gross IIP components are considered, but requires interaction with relative income per capita.
    - The real exchange rate and credit gaps have the correct sign but are not significant in any specification.
  - EMDE subsample (Table 2, column 2):
    - Impact of foreign-currency-denominated debt liabilities increases.
    - Foreign official reserves play a more important mitigating role.
    - Crisis risks rise more with higher current account deficits and higher global risk aversion.
    - Financial development and income per capita are not significant in the EMDE sample.
    - External assets held by the private sector play a limited mitigating role (reflecting resident capital flight).
  - Robustness:
    - Foreign currency-denominated debt liabilities are always statistically significant with remarkably stable magnitude across specifications.
    - Robustness checks include changes in country sample (exclude oil exporters and financial centers; consider Spain’s ESM program), alternative crisis definitions (large crises >200 percent of quota; two-year window; excluding consecutive crisis years), and additional controls (exchange rate regime, capital account openness, commodity price changes, fiscal balance).
    - Only commodity prices are significant among additional controls; inclusion does not alter key results.
    - Windsorizing data at 1 percent tails and estimating logit regression deliver similar results.
  - Additional notes:
    - External equity assets become significant when controls are excluded.
    - Equity liabilities and external debt denominated in local currency are significant when other controls are included.
    - Domestic currency denominated debt in countries without their own central bank may pose similar risks to foreign-currency debt.

### Probit margins: nonlinear effects and economic significance
- Method:
  - Predicted probabilities computed by holding variables at sample means and changing variable of interest in specified increments.
- Findings (Figure 5):
  - In the EMDE sample:
    - An increase in foreign-currency-denominated debt liabilities from 40 percent of GDP (near the EMDE median) to 60 percent of GDP is associated with a 5-percentage point increase in the predicted probability of an external crisis.
  - In the full sample:
    - The same rise from 40 percent of GDP to 60 percent of GDP increases the crisis probability by 0.2 percentage points.
  - A decline in the current account balance from a surplus of 5 percent of GDP to a deficit of 5 percent of GDP is associated with an increase in the predicted probability of an external crisis by

*Source: wpiea2021020-print-pdf - Appendix Table A.2 includes the source of the variables*

### 5.3 percentage points for EMDEs. For the full sample, the probability rises by only 1.1 percentage

### 5.3 percentage points for EMDEs. For the full sample, the probability rises by only 1.1 percentage

### Nonlinear relations between reserves, global risk aversion, and external crises
- Predicted external stress probability is near zero when reserves are above 30 percent of GDP.
- In EMDEs:
  - A decline in foreign exchange reserves from 20 percent to 10 percent of GDP is associated with a 6.5 percentage points increase in the likelihood of an external crisis.
  - A further decline from 10 percent to 0 percent of GDP increases the predicted probability by an additional 12.6 percentage points.
  - A doubling of global risk aversion from its historical mean increases predicted crisis probabilities by 6.5 percentage points.
  - A further doubling to levels witnessed in the early phase of the Global Financial Crisis and the Covid-19 pandemic increases predicted crisis probabilities by 23.5 percentage points.
- For the full sample:
  - Corresponding reserve-decline impacts are 0.7 and 2.1 percentage points, respectively.
  - Global risk aversion increases predicted probabilities by 0.7 and 4 percentage points for the two doublings described above.
- The relationship between foreign official reserves and external crises is markedly nonlinear, with much larger marginal effects at low reserve levels in EMDEs than in the full sample.

### Interaction of multiple vulnerabilities and predicted crisis probabilities
- Combination of two or more external vulnerabilities greatly increases the probability of external crisis in EMDEs.
- Example: When foreign-currency-denominated debt is 40 percent of GDP, the predicted probability ranges from 2–13 percent depending on reserves and current account balance (75th versus 25th percentiles).
- Example: For a current account deficit of 5 percent:
  - Predicted crisis probabilities range from 2.5 to 21 percent depending on whether the economy is less vulnerable (official reserves above the 75th percentile and foreign-currency-denominated debt below the 25th percentile) or more vulnerable (official reserves below the 25th percentile and foreign-currency-denominated debt above the 75th percentile).
- Low foreign official reserves carry greater risks when current account deficits and foreign-currency-denominated debt are also high.

### Implications of global risk shocks for EMDEs
- When global risk aversion reaches peak values seen in the initial phase of the Covid-19 pandemic (about 80, a value also reached during the Global Financial Crisis):
  - The external crisis probability for an EMDE with median levels of preexisting vulnerabilities rises to about 40 percent.
  - This is more than double the estimated probability for less vulnerable EMDEs (as defined previously).
- Results reaffirm the importance of external buffers to deal with global risk shocks.

### Output costs and macroeconomic dynamics after crises (EMDEs)
- Method: Local projections (Jordà, 2005) trace dynamic responses of output, exchange rates, and the current account after crises for the EMDE sample.
- Vulnerability classification: countries with high foreign-currency-denominated debt (above median), high current account deficit (below median), and low foreign official reserves (below median) are flagged as highly vulnerable.
- Post-crisis outcomes for vulnerable EMDEs:
  - Initial output loss is over 4 percent; output remains below pre-crisis levels after five years.
  - Initial exchange rate depreciation averages 10 percent.
  - Current account adjustment averages 2 percent of GDP.
- Post-crisis outcomes for less vulnerable EMDEs:
  - Initial losses are smaller (1 percent).
  - Output rises to above pre-crisis level after four years.
- Conclusion: Post-crisis adjustment is larger and recovery slower when countries have multiple pre-existing external vulnerabilities.

### Sudden stops: magnitudes, drivers, and differences from external crises
- Event analysis qualitative findings:
  - Sudden stops are preceded by net IIP declines driven mostly by increases in external debt liabilities denominated in foreign currency, with currency depreciation and negative real growth also contributing.
  - Foreign official reserves decline and continue to fall in the aftermath, implying reserve sales buffer sudden stops.
  - External debt liabilities denominated in domestic currency are flat around sudden stop episodes; equities follow a path similar to crisis episodes.
- Quantitative magnitudes (sudden stops vs external crises):
  - Net IIP declines by more than 40 percentage points around a sudden stop episode.
  - Net IIP declines by about 3 percentage points for all external crises and by about 14 percentage points for large external crises (IMF financial assistance exceeding 200 percent of quota).
  - Foreign-currency-denominated external debt liabilities increase by almost 80 percentage points for sudden stops, compared with 12 percentage points for external crises and 22 percentage points for large external crises.
- Probit estimation results and margins:
  - Deterioration in net IIP increases the probability of sudden stops; effects driven by debt and its foreign-currency component.
  - Net equity position loses statistical significance once foreign and domestic currency components of net debt are considered.
  - Higher net foreign-currency-denominated assets reduce the likelihood of sudden stops; effect driven by lower gross foreign currency debt liabilities.
  - Domestic-currency-denominated debt liabilities appear to mitigate sudden stop risks, possibly reflecting stronger policy frameworks and deeper capital markets.
  - Overvalued exchange rates and excessive credit increase the likelihood of sudden stops (these variables were not significant for external crises).
  - Levels of global risk aversion play an even larger role for sudden stops.
  - Current account deficits and foreign official reserves are no longer statistically significant predictors of sudden stops in the probit specifications.
  - Fiscal balance becomes significant for sudden stops: higher fiscal deficits increase the probability of sudden stops.
  - Predicted probit margins for sudden stops are generally much lower than for external crises; sudden stops are rarer and harder to predict.
  - An increase of foreign-currency-denominated debt liabilities from 40 to 60 percent is associated with only a marginal increase in the probability of a sudden stop, though the impact is nonlinear and increases with the size of FX denominated debt and when combined with other vulnerabilities.
  - For global risk aversion, predicted probabilities can be quantitatively similar to the case of external crises.

### Synthesis and policy-relevant implications
- The size and currency composition of the IIP matter, especially for EMDEs:
  - Higher external debt increases the likelihood of external crises.
  - Higher foreign-currency-denominated debt increases the likelihood of sudden stops.
- Official reserves mitigate the probability of external crises, with diminishing returns.
- Macroeconomic variables:
  - Higher current account deficit increases the probability of external crises.
  - Overvalued exchange rates, high credit, and fiscal deficits increase the chances of a sudden stop with high growth impact.
  - Higher global risk aversion increases the likelihood of both external crises and sudden stops.
- Nonlinearities: Combinations of vulnerabilities in two or more variables amplify risks non-linearly.
- Policy priorities:
  - Guard against buildup of external debt, especially in foreign currency.
  - Maintain external buffers (official reserves) to manage global risk shocks.
  - Monitor flow and stock imbalances, including foreign-currency-denominated debt and current account positions.
  - Recognize that macroeconomic consequences (lost real GDP, sharp current account and REER adjustments) are larger for economies with greater preexisting vulnerabilities.
- Note on low-income countries (LICs): Vulnerabilities are relevant for LICs that have borrowed in foreign currency; data constraints prevented more LICs from being included and future research should study vulnerabilities for these countries.

*Source: wpiea2021020-print-pdf*

### REFERENCES

### REFERENCES

### Key cited works
- Ahmed, Shagil, and Andrei Zlate, 2014, “Capital Flows to Emerging Market Economies: A Brave New World?” Journal of International Money and Finance, Vol. 48, pp. 221–48.
- Arellano, Cristina, 2008, “Default Risk and Income Fluctuations in Emerging Economies,” American Economic Review, Vol. 98, No. 3, pp. 690–712.
- Asonuma, Tamon, and Christoph Tresbesch, 2016, “Sovereign Debt Restructurings: Preemptive or Post-Default.” Journal of the European Economic Association, Vol. 14, pp. 175–214.
- Bordo, Michael D., Øyvind Eitrheim, Marc Flandreau, and Jan F. Qvigstad, 2016, Central Banks at a Crossroads: What Can We Learn from History? (New York: Cambridge University Press).
- Calvo, Guillermo A., 1978, “On the Time Consistency of Optimal Policy in a Monetary Economy,” Econometrica, Vol. 46, pp. 1411–28.
- Catão, Luis A.V., and Gian Maria Milesi-Ferretti, 2014, “External Liabilities and Crises,” Journal of International Economics, Vol. 94, pp. 18–32.
- Dornbusch, Rudiger, Ilan Goldfajn, and Rodrigo O. Valdés. 1995. “Currency Crises and Collapses,” Brookings Papers on Economic Activity 2: 219–93.
- Eichengreen, Barry, Ricardo R. Hausmann, and Ugo Panizza, 2004, “The Pain of Original Sin,” in Other People’s Money: Debt Denomination and Financial Instability in Emerging Market Economies ed. by Barry Eichengreen and Ricardo Hausmann.
- Frankel, Jeffrey A., and George Saravelos, 2012, “Can Leading Indicators Assess Country Vulnerability? Evidence from the 2008–09 Global Financial Crisis,” Journal of International Economics, Vol. 87, No. 2, pp. 216–31.
- Lane, Philip R., and Gian Maria Milesi-Ferretti, 2007, “The External Wealth of Nations Mark II: Revised and Extended Estimates of Foreign Assets and Liabilities, 1970–2004,” Journal of International Economics, Vol. 73, pp. 223–50.
- Obstfeld, Maurice, Jay Shambaugh, and Alan Taylor, 2009, “Financial Instability, Reserves, and Central Bank Swap Lines in the Panic of 2008,” American Economic Review, Vol. 99, Vol. 2, pp. 480–86.
- Svirydzenka, Katsiaryna, 2016, “Introducing a New Broad-based Index of Financial Development,” IMF Working Paper No. 16/5 (Washington: International Monetary Fund).
- Turrini, Alessandro, and Stefan Zeugner, 2019, “Benchmarks for Net International Investment Positions,” Journal of International Money and Finance, Vol. 95 (C), pp. 149–64.
- International Monetary Fund, 2013, “The IMF-FSB Early Warning Exercise: Design and Methodological Toolkit,” Occasional Paper No. 274 (Washington: International Monetary Fund).
- International Monetary Fund, 2016, “Understanding the Slowdown in Capital Flows in Emerging Markets,” World Economic Outlook, Chapter 2, April (Washington).
- International Monetary Fund, 2020, External Sector Report, Chapters 1 and 2, August. (Washington).

### Empirical findings from Probit estimates for External Crisis Events (Table 1)
- Net IIP, share of GDP: -0.34* and -0.27* (standard errors (0.19)(0.15))
- Net debt, share of GDP: -0.55*** and -0.50*** (0.21)(0.16)
- Net foreign currency debt: -0.61** and -0.52*** (0.25)(0.17)
- Net domestic currency debt: 0.13 and -0.29 (0.40)(0.36)
- Foreign currency debt assets: 0.22 and 0.40 (0.33)(0.29)
- Foreign currency debt liabilities: 0.62** and 0.44** (0.28)(0.21)
- Domestic currency debt liabilities: 0.12 and 0.75* (0.39)(0.39)
- Foreign exchange reserves, share of GDP: -7.22***, -7.21***, -7.02***, -6.13***, -6.20***, -5.22*** (1.43)(1.44)(1.65)(1.65)(1.70)(1.81)
- Income per capita, relative to US: -3.92***, -2.90**, -2.88**, -1.28 (1.00)(1.18)(1.18)(1.34)
- Current account/GDP, 2yr MA: -5.32***, -5.71***, -5.73***, -6.89*** (1.52)(1.54)(1.53)(1.50)
- Financial development index: -4.96***, -2.87**, -2.72**, -0.42 (1.19)(1.22)(1.37)(1.34)
- Fin. develop. * income per capita relative to US: 5.76***, 1.80, 1.70, -5.03* (1.70)(2.16)(2.21)(3.00)
- VXO: 0.02** across multiple specifications (0.01)
- Constant terms: -1.95***, -1.41***, -1.35***, -1.24***, -0.08, -0.08, -0.10, -0.67** (various standard errors)
- Observations: 1,828 across the main specifications.
- Number of ifs_code: 73 across specifications.
- Log pseudolikelihood examples: -410.38, -385.51, -384.42, -377.31, -358.33, -342.39, -342.32, -333.29.
- AUROC estimates: 0.66, 0.67, 0.72, 0.82, 0.83, 0.84, 0.84, 0.87.
- Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. All country-specific variables are lagged by one year. Net refers to assets minus liabilities. Main regressions use stata command "xtprobit". Approximate AUROC estimates calculated using "probit" then "lroc".

### Robustness checks for External Crisis Events (Table 2) — selected coefficients
- Foreign currency debt assets: example coefficient 0.56** (0.28) in one robustness column; other columns show variation including -1.06* (0.56).
- Foreign currency debt liabilities: consistently positive and significant; example 0.44** (0.21) and 1.78*** (0.33) in EMDEs-only column.
- Domestic currency debt liabilities: 0.75* (0.39) in baseline; some columns show higher significance, e.g., 1.10** (0.52).
- Foreign exchange reserves, share of GDP: -5.22*** (1.81) baseline; robustness range includes -10.57*** (3.68) in one column.
- Current account/GDP, 2yr MA: -6.89*** (1.50) baseline; robustness examples include -7.55*** (1.60) and -12.75*** (2.89).
- REER gap, Credit gap, Financial development index, Fin. develop. * income per capita, and VXO retain varied significance across robustness checks.
- Observations vary by sample: examples include 1,828 (Full Sample), 1,004 (EMDEs only), 1,706 (Exclude oil exporters), etc.
- AUROC estimates across robustness checks: examples 0.87, 0.81, 0.87, 0.88, 0.88, 0.86, 0.85, 0.88.

### Probit estimates for Sudden Stops (Table 3) — selected coefficients
- Net IIP, share of GDP: -0.26*** (0.07) and -0.50 (0.35) in selected columns.
- Net debt, share of GDP: -0.26*** (0.07) and -0.50* (0.27).
- Net foreign currency debt: -0.33*** (0.09) and -0.62*** (0.21).
- Net domestic currency debt: 1.05*** (0.35) and 1.86** (0.77).
- Foreign currency debt liabilities: 0.34*** (0.08) and 0.59*** (0.21).
- Domestic currency debt liabilities: -1.40*** (0.51) and -2.50*** (0.77).
- Income per capita, relative to US: 4.46***, 4.36***, 5.23***, 5.52*** (standard errors around 1.39–1.58).
- REER gap, relative to 5yr MA: 2.89***, 2.95***, 3.03***, 2.86*** (0.63–0.67).
- Credit gap, credit/GDP compared to 5yr MA: 4.31***, 4.45***, 4.68***, 4.35*** (1.23–1.33).
- Fin. develop. * income per capita relative to US: -6.79**, -6.70**, -9.60**, -10.92** (standard errors 3.27–4.49).
- VXO: 0.09*** across columns (0.01).
- Observations: examples 1,808 across many specifications.
- Log pseudolikelihood examples: -156.03, -156.02, -152.90, -149.78.
- AUROC estimates: 0.60, 0.61, 0.64, 0.73, 0.92, 0.92, 0.93, 0.93.

### Robustness checks for Sudden Stops (Table 4) — selected coefficients and diagnostics
- Foreign currency debt liabilities: 0.59*** (0.21) baseline and up to 1.60** (0.69) in EMDEs-only column.
- Domestic currency debt liabilities: -2.50*** (0.77) baseline; other columns show strong negative coefficients such as -3.78*** (1.24).
- Domestic currency debt assets: 1.88*** (0.65) baseline and other significant positive coefficients in robustness columns.
- REER gap, relative to 5yr MA: 2.86*** across many columns (standard errors around 0.63–1.44).
- Credit gap, credit/GDP compared to 5yr MA: 4.35*** baseline and up to 11.67*** in some robustness columns (standard errors vary).
- Fin. develop. * income per capita relative to US: -10.92** (4.49) baseline and multiple significant negative coefficients across specifications.
- VXO: 0.09*** to 0.13*** across columns (standard errors 0.01–0.02).
- Commodity prices: -0.34*** (0.09) in additional controls panel.
- Fiscal (example): -7.01** (3.38) in additional controls panel.
- Observations vary by sample: examples include 1,808 (Full Sample), 999 (EMDEs), 1,691 (Exclude oil exporters).
- Log pseudolikelihood examples: -100.55, -63.06, -91.19, -78.29.
- AUROC estimates: 0.93, 0.94, 0.94, 0.96, 0.93 across panels.
- Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. All country-specific variables are lagged by one year. Main regressions use stata command "xtprobit". Approximate AUROC estimates computed using "probit" then "lroc".

### Figures and data sources
- Figure 1. Stock Imbalances, 1990–2018  
  - Data series shown: External Assets and Liabilities, Median Country (Percent of GDP) and External Assets and Liabilities (Percent of world GDP).  
  - Legend labels: AE (right scale) and EMDE.  
  - Sources: External Wealth of Nations database (Lane and Milesi-Ferretti 2007); and IMF, World Economic Outlook database.
  - Note: AE = advanced economies; EMDE = emerging market and developing economies.
- Figure 2. External Crisis Episodes, 1990–2018  
  - Sources: Das and others (2011); Asonuma and Trebesch (2016); Paris Club; and authors' calculations.
- Figure 3. Sudden Stop Episodes, 1991–2018  
  - Series include Sudden stops and Sudden stops with high growth impact.  
  - Sources: IMF, World Economic Outlook database and authors' calculations.
- Figure 4. Conditional Mean of Key IIP Components, the REER and Growth Around External Crisis Episodes  
  - Note: Solid blue line corresponds to the mean of all external crisis episodes, blue shaded area is the 90 percent confidence interval. Red line corresponds to the mean of large crisis episodes. An increase (decrease) in the Real Effective Exchange Rate denotes appreciation (depreciation).

*Content derived solely from the provided PDF section titled "REFERENCES" and associated tables and figure captions.*

### 1. Net International  Investment Position

### 1. Net International  Investment Position

### Model Predicted Probabilities of External Crises (Figure 5)
- All panels display the predicted probabilities of an external crisis episode, keeping all covariates except foreign currency debt, foreign exchange reserves, the current account, and global risk aversion at their sample mean.
- The solid blue line corresponds to the mean estimate, the dotted lines correspond to the 90 percent confidence interval.
- Predictors shown (as varied variables) include:
  - Gross FX Debt Liabilities
  - FX Reserves
  - Current Account Balance
  - Global Risk Aversion

### Selected Predictors in the EMDE Sample (Figure 6)
- Sources referenced for data and calculations: External Wealth of Nations database (Lane and Milesi-Ferretti 2007); Haver Analytics; IMF, World Economic Outlook database; and authors’ calculations.
- Panels display predicted probabilities of an external crisis episode, keeping all covariates except foreign currency debt, foreign exchange reserves, the current account, and global risk aversion at their sample mean.
- Definitions used for vulnerability groupings:
  - More vulnerable countries: foreign currency debt at the 75th percentile and foreign exchange reserves and current account balance at the 25th percentile of the sample.
  - Less vulnerable countries: foreign currency debt at the 25th percentile and foreign exchange reserves and current account balance at the 75th percentile.
  - Median countries: foreign currency debt, foreign exchange reserves, and current account balance at the median.

### Dynamic Evolution After External Crisis Episodes (Figure 7)
- The figure reports the dynamic evolution of output, real exchange rates, and current account balance following external crisis episodes.
- Shaded area corresponds to the 90 percent confidence interval.
- The horizontal axis denotes time in years, and 0 is the year of the external crisis episode.
- EMDEs = emerging market and developing economies; REER = real effective exchange rate.
- Vulnerable EMDEs are defined as those with foreign currency debt above the EMDE median, and current account balance and foreign exchange reserves below the EMDE median.

### Conditional Mean Around Sudden Stop Episodes (Figure 8)
- The solid blue line corresponds to the mean of all sudden stop episodes, while the blue shaded area corresponds to the 90 percent confidence interval.
- IIP components are expressed as percent of GDP.
- An increase (decrease) in the Real Effective Exchange Rate denotes appreciation (depreciation).
- Variables shown around sudden stop episodes include:
  - 1. Net International Investment Position
  - 2. Foreign Currency Debt Assets
  - 3. Domestic Currency Debt Assets
  - 4. Foreign Currency Debt Liabilities
  - 5. Domestic Currency Debt Liabilities
  - 6. Equity Assets
  - 7. Equity Liabilities
  - 8. Foreign Exchange Reserves
  - 9. Real Effective Exchange Rate
  - 10. Real GDP Growth
- Time axis for panels: from -5 to 5 years relative to the sudden stop (0 is year of episode).

### Sample Economies (Appendix Table A.1)
- Listed sample economies include (selection as shown): Argentina, Australia, Austria, Bangladesh, Belarus, Belgium, Bosnia and Herzegovina, Brazil, Canada, Chile, China, Colombia, Croatia, Czech Republic, Denmark, Dominican Republic, Egypt, El Salvador, Estonia, FYR Macedonia, Finland, France, Georgia, Germany, Greece, Guatemala, Hong Kong SAR, Hungary, Iceland, India, Indonesia, Ireland, Israel, Italy, Jamaica, Japan, Jordan, Kazakhstan, Korea, Latvia, Lithuania, Malaysia, Mexico, Morocco, Netherlands, New Zealand, Nigeria, Norway, Oman, Pakistan, Peru, Philippines, Poland, Portugal, Romania, Russia, Singapore, Slovak Republic, Slovenia, South Africa, Spain, Sri Lanka, Sweden, Switzerland, Thailand, Tunisia, Turkey, Ukraine, United Kingdom, United States, Uruguay, Venezuela, Vietnam.

### Data Sources (Appendix Table A.2)
- Data sources are listed (table present); specific sources referenced elsewhere include External Wealth of Nations database, Haver Analytics, IMF World Economic Outlook database.

### External Crisis Episodes (Appendix Table A.3)
- The appendix tabulates external crisis episodes by country and year. Selected entries include:
  - Argentina: 1991‐92, 1996, 1998, 2000, 2002‐03, 2014, 2018
  - Bangladesh: 2003, 2012
  - Belarus: 2009
  - Bosnia and Herzegovina: 2002, 2009, 2012, 2016
  - Brazil: 1992, 1998, 2001‐02
  - Colombia: 1999
  - Dominican Republic: 1991, 1993, 2003‐05, 2009
  - Egypt: 1996, 2016
  - El Salvador: 1992‐93, 1995, 1997‐98
  - FYR Macedonia: 2000, 2003, 2005
  - Georgia: 2001, 2004, 2008, 2014, 2017
  - Guatemala: 1992‐93
  - Greece: 2010, 2012
  - Hungary: 1996, 2008
  - Iceland: 2008
  - India: 1991
  - Indonesia: 1994, 1997‐98, 2000, 2002, 2005
  - Ireland: 2010
  - Jamaica: 1991‐93, 2010, 2013, 2016
  - Jordan: 1992, 1994, 1996‐97, 1999, 2002, 2012, 2016
  - Kazakhstan: 1999
  - Korea: 1997
  - Latvia: 1999, 2008
  - Mexico: 1995, 1999
  - Morocco: (listed in sample; not listed in crisis table excerpt)
  - Nigeria: 1991, 2005
  - Pakistan: 1993‐95, 1997‐01, 2008, 2013
  - Peru: 1991, 1993, 1996, 1999
  - Philippines: 1991, 1994, 1998
  - Poland: (listed in sample; not listed in crisis table excerpt)
  - Portugal: 2011
  - Russia: 1999
  - Slovakia, Slovenia: (listed in sample; not listed in crisis table excerpt)
  - Sri Lanka: 1991, 2001, 2003, 2005, 2009, 2016
  - Thailand: 1997
  - Tunisia: 2013, 2016
  - Turkey: 1994, 1999, 2002, 2005
  - Ukraine: 1998‐01, 2004, 2008, 2010, 2014‐15, 2018
  - Uruguay: 1992, 1996‐97, 1999‐00, 2002, 2005
  - Venezuela: 1996
  - Vietnam: 2001

*Source: wpiea2021020-print-pdf - 1. Net International  Investment Position*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021020-print-pdf.pdf_