## 0.5 probability of exit from the crisis.

## Source details

**Canonical URL:** [0.5 probability of exit from the crisis.](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09103.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09103.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09103.pdf.json)

---

### I. Introduction — scope and key conceptual points
- Capital account crises: episodes of financial distress characterized by abrupt capital outflows with deep macroeconomic and social consequences.
- Distinction from earlier currency crises: capital account crises involve balance sheet interlinkages that enable rapid transmission of distress and can produce "twin" or "triple" crises.
- Crisis duration is a substantive state of the economy that can persist; duration correlates with complexity and depth: more complex crises tend to last longer and be more damaging.
- Dataset: 18 crisis episodes across 12 emerging market economies (Argentina, Brazil, Ecuador, Indonesia, Korea, Malaysia, Mexico, Philippines, Russia, Thailand, Turkey, Uruguay) covering episodes starting 1994 (Turkey, Mexico) through Uruguay’s crisis in 2002.

### II. Measurement approach and index (IKAC)
- Index construction inputs:
  - foreign exchange reserves (FX),
  - nominal effective exchange rate (NEER),
  - secondary market spreads on sovereign bonds (S),
  - net private capital flows scaled to GDP (K).
- Index rules and transformations:
  - Each term standardized (mean zero, standard deviation one); positive index values denote financial pressure.
  - NEER enters as percentage changes; spreads and capital flows compared to country-specific sample means; reserves compared to long-run trend.
  - Start of crisis: first of two consecutive quarters with IKAC > 0.
  - End of crisis: first of two consecutive quarters with IKAC < 0.
- Sample outcome: application produced 18 crisis episodes; 16 out of 18 crises lasted less than 12 quarters.

### III. Stylized findings on duration, complexity, and macroeconomic costs
- Average duration (quarters) by crisis type:
  - Single crises: 7.1
  - Twin crises: 9.9
  - Triple crises: 4.7
- Crisis composition:
  - 15 out of 18 crises involved twin or triple crises.
  - Virtually all triple and half of twin crises had longer-than-median duration; three single crises were all shorter than the median.
- Macroeconomic impact (averages by crisis type):
  - Real GDP loss (percent):
    - Single crises: -3.9
    - Twin crises: -7.1
    - Triple crises: 0.4
  - Change in inflation (percent):
    - Single crises: -3.7
    - Twin crises: 13.5
    - Triple crises: 41.1
  - Change in net private capital flows (percent of GDP):
    - Single crises: -1.1
    - Twin crises: -2.0
    - Triple crises: -2.3
- Post-crisis external debt increase (percent of GDP; difference of four-quarter averages post- vs pre-crisis):
  - Single crises: 8.2
  - Twin crises: 21.4
  - Triple crises: 2.3

### IV. Determinants conceptual framework
- Four determinant categories:
  - Initial conditions: policy record, external imbalances, solvency/liquidity risks, balance sheet vulnerabilities.
  - External conditions: investors’ appetite, global liquidity, terms of trade, export market demand.
  - Policy response: fiscal adjustment, short-term interest rate hikes, exchange rate flexibility, structural policies (difficult to quantify).
  - IMF financial support: availability, extent, timing of official financing and Fund-supported programs.
- Empirical note: IMF-supported programs were in place in 15 out of 18 crises; no programs (at the time) in Malaysia (1997), Philippines (2002), Turkey (1998).

### V. Econometric methodology and model specifications
- Model: grouped duration (survival) model with discrete hazard formulation and time-varying covariates.
- Time-dependence functional forms tested:
  - Logistic time-dependent baseline hazard: λ0 = α0 + α1 ln(t)
  - Linear time-dependent baseline hazard: λ0 = α0 + α1 t
  - Time-invariant baseline hazard: λ0 = α0
- Dataset structure: 18 crisis episodes; 153 quarterly observations.
- General-to-specific search: initial wide variable set reduced by sequential elimination using Bayesian and Akaike information criteria.
- Key explanatory variables (initial specification):
  - Initial conditions: pre-crisis external debt (% of GDP) and lag; pre-crisis short-term debt (% of reserves) and lag; pre-crisis primary fiscal balance; public debt-to-GDP ratio.
  - External conditions: net private capital flows to emerging markets (ratio to GDP); three-month LIBOR; changes in terms of trade; trade-weighted partner demand.
  - Policy response channels: change in primary fiscal balance (percent of GDP) contemporaneous and up to two lags; policy interest rate differentials vis-à-vis LIBOR adjusted for inflation differential (contemporaneous and up to six lags); exchange rate regime (AREAER classification).
  - IMF support: cumulative disbursements (percent of GDP) interacting with IMF program dummy (fitted cumulative sum with Tobit-generated fitted values).

### VI. Estimation highlights and coefficient patterns
- Sample and fit metrics:
  - Number of observations: 153
  - Reported log likelihoods: -30.822, -34.467, -35.858, -35.693, -36.955, -40.026
  - Reported AIC values: 111.564, 92.934, 91.715, 95.386, 93.909, 102.051
  - Reported BIC values: 187.405, 129.299, 122.020, 131.751, 124.214, 135.386
- Preferred specification: logistic time specification (best fit and lowest information criteria).
- Common variables retained in parsimonious specifications:
  - External debt (lag 1)
  - Pre-crisis current account balance
  - Capital flows to emerging markets
  - Three-month LIBOR
  - Change in trade-weighted partner country demand
  - Change in primary balance (lag 1)
  - Real interest rate differential (lags 2 and 3)
- Selected coefficient signs and significance patterns (reported examples):
  - Quarters in crisis (log): positive and significant (e.g., 2.719**, 2.680***, 2.501***).
  - External debt (lag 1): negative coefficients (e.g., -0.058; -0.051*; -0.039**; -0.049*; -0.036**; -0.042***).
  - Current account balance (pre-crisis): positive and significant (e.g., 0.214***; 0.280***; 0.239***; 0.311***; 0.261***; 0.162**).
  - Three-month LIBOR: negative and often significant (e.g., -0.373*; -0.636**; -0.577***; -0.594**; -0.543***; -0.510***).
  - Trade-weighted partner demand (change): positive and often significant (e.g., 1.318*; 0.985**; 1.165***; 1.282***; 1.354***; 1.159***).
  - Change in primary balance (lag 1): positive and significant in parsimonious specs (e.g., 0.174**; 0.092**; 0.083**; 0.087*; 0.082*; 0.096***).
  - Real interest rate differential (lag 2): negative and sometimes significant (e.g., -0.050; -0.035**; -0.031; -0.036**; -0.032; -0.040**).
  - Real interest rate differential (lag 3): positive and sometimes significant (e.g., 0.029; 0.038***; 0.032**; 0.039***; 0.035**; 0.044**).
  - Exchange rate regime: negative sign in several specs (e.g., -0.370; -0.298; -0.277; -0.386**) with p-values around 0.13 and 0.17 in some time-dependent specifications.
  - IMF financing * program dummy: positive across models (e.g., 0.080; 0.052; 0.075; 0.272**) but statistically significant only in the time-invariant baseline hazard specification.

### VII. Diagnostics and robustness
- Baseline model hazard predictions align with Nelson-Aalen estimator based on 18 observed crisis durations, indicating reasonable explanatory power.
- Time-varying explanatory variables are critical: setting them to zero causes predicted probability curve to shift dramatically away from Nelson-Aalen hazard.
- Logistic time specification predicts exit probabilities better for durations less than 12 quarters and is preferred for 16 out of 18 crises; linear specification fits longer crises better.
- Time-invariant model: comparable likelihood values indicate explanatory power though diagnostics are technically harder to apply.

### VIII. Key empirical findings on determinants of duration
- Initial and external conditions:
  - Larger pre-crisis current account deficits are associated with longer crises.
  - Higher levels of external debt (and short-term external debt in the time-invariant model) strongly correlate with longer crises.
- External environment:
  - Benign global liquidity (lower three-month LIBOR) and favorable partner demand shorten crisis duration significantly.
  - Net private capital flows to emerging markets are weakly and non-robustly associated with exit probability.
- Policy responses:
  - Fiscal consolidation (change in primary balance, lag 1) shortens crisis duration across specifications, suggesting signaling/confidence effects dominate contractionary impact.
  - Monetary policy: raising real interest rates appears to shorten crisis duration with a lag of three quarters, but net effect ambiguous due to mixed coefficient signs across lags.
  - Exchange rate flexibility during a crisis tends to prolong the crisis (negative coefficient), likely reflecting balance-sheet and overshooting effects.
- IMF financial support:
  - Positive coefficient on cumulative IMF financing suggests larger financing packages raise exit probability, but statistical significance lacks robustness (significant only in time-invariant baseline hazard regression).

### IX. Counterfactual experiments and scenario analysis
- Framework: augmented logistic time specification (regression 2a); baseline probability constructed using estimated parameters and mean variable values.
- Counterfactual method: shift a variable by one standard deviation and compare predicted probabilities and expected durations.
- Main quantitative counterfactual insights:
  - Factors outside policy control at crisis onset (initial and external conditions) have the largest impact on expected duration.
  - Largest gains in shortening expected duration—by roughly 2–3 quarters, or about 30–40 percent of the median duration—are associated with either:
    - Stronger initial conditions (lower pre-crisis current account deficits and moderate external debt), or
    - A more benign external environment (better international liquidity conditions and buoyant trade partners’ demand).
- Specific modeled policy-scenario impacts (evaluated at median crisis duration of seven quarters or at 0.5 probability):
  - Stronger improvement in primary fiscal balance (by one standard deviation):
    - Increases probability of exit by about 10 percent; shortens duration by about one quarter.
  - Maintaining pre-crisis exchange rate regime (avoiding large devaluations):
    - Increases probability of exit by about 10 percent; shortens duration by about one quarter.
  - One standard deviation higher real interest rate differential:
    - Raises probability of exiting marginally by 3 percent; shortens duration by about 0.4 quarters.
  - One standard deviation higher IMF financing:
    - Increases probability of exit by about 12 percent; result based on a coefficient not statistically significant and should be interpreted with caution.
  - Other positive scenario effects at median duration include:
    - Capital flows to EM + 1SD
    - Trade-weighted partner country demand + 1SD
    - CAB-to-GDP (pre-crisis) + 1SD
    - LIBOR - 1SD
    - External debt-to-GDP (t-1) - 1SD
  - Descriptive statistics used in counterfactuals (N and Mean, Standard deviation):
    - Exchange rate regime (N=153): Mean = 6.67; Standard deviation = 2.02
    - Real interest rate differential (lag 2) (N=153): Mean = 6.27; Standard deviation = 43.41
    - Real interest rate differential (lag 3) (N=153): Mean = 6.12; Standard deviation = 40.34
    - Change in primary balance (lag 1) (N=153): Mean = 0.48; Standard deviation = 4.76
    - Capital flows to EM countries (N=153): Mean = 1.41; Standard deviation = 0.89
    - Three-month Libor rate (N=153): Mean = 4.54; Standard deviation = 1.84
    - Trade-weighted domestic demand (change) (N=153): Mean = 0.58; Standard deviation = 0.62
    - External debt (lag 1) (N=153): Mean = 70.04; Standard deviation = 28.92
    - Current account balance (pre-crisis) (N=18): Mean = -2.45; Standard deviation = 3.89
    - Cumulative IMF financing (N=151): Mean = 2.96; Standard deviation = 13.75

### X. Additional summary statistics and facts
- Median crisis duration: seven quarters.
- Average crisis duration: about 8½ quarters.
- Crisis duration range in sample: from three quarters (1998 Turkey) to 18 quarters (1997 Thailand).
- 0.5 probability threshold used as a reference for expected durations and scenario comparisons.
- 16 out of 18 crises in the sample lasted less than 12 quarters.

### XI. Policy-relevant implications
- Prevention matters: prudent macroeconomic policies strengthen initial conditions and substantially reduce expected crisis duration; once a crisis erupts it is often too late to fully correct weak initial conditions.
- Fiscal policy: appropriate fiscal adjustment during a crisis can shorten crisis duration significantly via confidence and signaling effects.
- Monetary policy: evidence is mixed; raising real interest rates may help shorten crises with lags, but trade-offs exist.
- Exchange rate policy: large exchange rate devaluations or moves toward flexibility during crises are associated with longer crises, likely due to adverse balance-sheet effects.
- IMF support: larger IMF financing packages are associated with higher exit probability in some specifications, but statistical robustness is limited; disentangling financing effects from program-related policy adjustments is difficult.
- External environment: favorable global liquidity and partner demand ('good luck') materially shorten crisis duration.

*Source: IMF staff calculations and analysis in _wp09103 - 0.5 probability of exit from the crisis.*

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

### _wp09103 - References

### I. Introduction
- Capital account crises are episodes of financial distress characterized by abrupt capital outflows and are highly disruptive with deep macroeconomic and social consequences.
- These crises differ from earlier currency crises that were typically linked to an inconsistency between an exchange rate peg and fiscal/monetary policies; earlier crises were usually short-lived and sometimes had limited or even positive real economic implications.
- Capital account crises are closely associated with balance sheet interlinkages between sectors, enabling rapid transmission of distress and producing "twin" or "triple" crises.
- Key conceptual point: crisis duration is a substantive feature—crisis becomes a state of the economy that can persist. Crisis duration is strongly correlated with complexity (nature of crisis episodes) and depth (impact on macroeconomic variables): the more complex a crisis, the longer its duration and the more damaging its effects.
- This study uses a data set covering 18 crisis episodes (starting with Turkey and Mexico in 1994 and ending with Uruguay’s crisis in 2002) across 12 emerging market economies to examine relationships among crisis duration, complexity, and depth, and to provide systematic evidence on factors influencing crisis persistence.

### II. Conceptual framework and methodology
- Measurement approach:
  - An index methodology (originally developed in Mecagni et al (2007)) associates crises with elevated pressures on the exchange rate and reserves, large capital outflows, and significant deviations from “normal” levels of spreads on sovereign bonds.
  - Appendix I provides a brief description of the methodology to measure crisis duration; Mecagni et al (2007) offers more details on the index design and performance.
- Sample:
  - Focus on 12 emerging market economies commonly cited in the literature: Argentina, Brazil, Ecuador, Indonesia, Korea, Malaysia, Mexico, Philippines, Russia, Thailand, Turkey, and Uruguay.
  - The sample yields 18 crisis episodes.

### III. Stylized findings on duration, complexity, and macroeconomic costs
- Relationship between complexity and duration:
  - Average duration (in quarters) by crisis type (Figure 1):
    - Single crises: 7.1
    - Twin crises: 9.9
    - Triple crises: 4.7
  - Majority of crises: 15 out of 18 crises involved twin or triple crises (Table 1).
  - Virtually all triple and half of the twin crises had longer-than-median duration; the three single crises were all shorter than the median duration.
- Link between duration/complexity and macroeconomic impact (Figure 2):
  - Average real GDP loss (percent) by crisis type:
    - Single crises: -3.9
    - Twin crises: -7.1
    - Triple crises: 0.4
  - Average change in inflation (percent) by crisis type:
    - Single crises: -3.7
    - Twin crises: 13.5
    - Triple crises: 41.1
  - Average change in net private capital flows (percent of GDP) by crisis type:
    - Single crises: -1.1
    - Twin crises: -2.0
    - Triple crises: -2.3
  - Scatter relationships (Figure 2) plotting duration (quarters) versus:
    - Real GDP loss (percent) over durations 0 to 20 quarters.
    - Change in inflation (percent) over durations 0 to 20 quarters.
    - Change in net private capital flows (percent of GDP) over durations 0 to 20 quarters.
- Post-crisis vulnerabilities (Figure 3):
  - Average increase in external debt (percent of GDP) measured as difference between the four-quarter averages after the crisis and the four-quarter averages preceding the crisis:
    - Single crises: 8.2
    - Twin crises: 21.4
    - Triple crises: 2.3

### IV. Determinants of crisis duration
- Determinants conceptualized into four broad categories:
  - Initial conditions:
    - A country’s policy record and initial conditions (e.g., large external imbalances, pre-existing solvency or liquidity risks, balance sheet vulnerabilities) likely affect nature, duration, and depth of a crisis by influencing policy options and market reactions.
  - External conditions:
    - Restoration of market confidence and speed of economic recovery are influenced by investors’ appetite for emerging market assets, global liquidity conditions, terms of trade dynamics, and changes in export markets.
  - Policy response:
    - Authorities’ policy responses materially affect crisis duration, though interactions can be complex:
      - Fiscal adjustment can re-establish credibility in fiscally driven crises but can have adverse contractionary effects.
      - Hikes in short-term interest rates may stem capital outflows but can harm the real economy and domestic banks.
      - Moves toward greater exchange rate flexibility may be unavoidable when reserves run out but can adversely affect balance sheets in the presence of currency substitution and currency mismatches.
    - Structural policies are also likely important but difficult to quantify consistently; thus their inclusion in econometric analysis is limited.
  - IMF financial support:
    - Availability, extent, and timing of official financing, particularly from the Fund, may be key to recovery speed from sudden stops or private capital exits.
    - An IMF-supported program may signal a comprehensive policy adjustment package and provide financial resources that catalyze private or official flows.
    - IMF-supported programs were in place in 15 out of the 18 crises in the sample; there were no programs (at the time) in Malaysia (1997), Philippines (2002), and Turkey (1998).

### V. Econometric methodology
- The study estimates the relative impact of determinants using a grouped duration model, suitable for panel-like data.
- Dataset structure:
  - 18 crisis episodes with a binary indicator denoting exit from crisis.
- Model features:
  - The probability of exiting a crisis in each period (dependent variable) is modeled as a function of:
    - Time already spent in crisis (a time-dependent component).
    - Other covariates representing initial conditions, external conditions, policy responses, and IMF financial support.
- Notes:
  - The determinants relevant to the probability of remaining in or exiting a crisis differ from those relevant to the probability of entering a crisis.
  - Ramakrishnan and Zalduendo (2006) provide related work on duration for episodes of “market pressures.”

*Source: IMF staff calculations and assessments derived from the study text.*

### Appendix II provides details on our application of survival analysis to studying durational aspects of capital

### Appendix II — Survival Analysis of Durational Aspects of Capital Account Crises

### Methodology and Model Specifications
- Baseline formulation: hazard function with baseline probability, λ0, common to all crisis episodes, and time-varying and country-specific explanatory variables, X:
  - Equation form (as presented): ][[][))(exp(exp1,|)(,0uXtutXtp++−−=βλβ]
- Time-dependence functional forms tested:
  - Logistic time-dependent baseline hazard: λ0 = α0 + α1 ln(t)
  - Linear time-dependent baseline hazard: λ0 = α0 + α1 t
  - Time-invariant baseline hazard: λ0 = α0 (no time variable)
- General-to-specific specification search:
  - Initial model covers a wide range of variables capturing initial conditions, external conditions, policy response channels, and IMF financial support.
  - Reduced-form specifications obtained by sequential elimination of least significant variables using Bayesian and Akaike information criteria.

### Explanatory Variables Included (initial general specification)
- Initial conditions:
  - Pre-crisis external debt (in percent of GDP) and its one-period lag.
  - Pre-crisis short-term debt (in percent of reserves) and its one-period lag.
  - Pre-crisis primary fiscal balance and public debt-to-GDP ratio.
- External conditions:
  - Ratio to GDP of net private capital flows into emerging markets.
  - Three-month LIBOR rate.
  - Changes in terms of trade and trade-weighted partner-countries’ demand.
- Policy response channels:
  - Change in primary fiscal balance (in percent of GDP) over a four-quarter period (contemporaneous and up to two lags).
  - Policy interest rate differentials (vis-à-vis LIBOR) adjusted for inflation differential with the United States (contemporaneous and up to six lags).
  - Exchange rate regime (based on the Fund’s AREAER classification; higher values indicate a more flexible regime).
- IMF financial support:
  - Cumulative disbursements (in percent of GDP) interacting with an IMF program dummy (fitted cumulative sum starting with four quarters preceding the crisis; fitted values generated by a Tobit model).

### Estimation Results (highlights from Table 2 and related text)
- Sample size and model fit metrics:
  - Number of observations: 153
  - Log likelihoods reported for models: -30.822, -34.467, -35.858, -35.693, -36.955, -40.026 (across regressions reported)
  - Akaike Information Criterion values reported: 111.564, 92.934, 91.715, 95.386, 93.909, 102.051
  - Bayesian Information Criterion values reported: 187.405, 129.299, 122.020, 131.751, 124.214, 135.386
- Preferred specification:
  - Logistic time specification provides the tightest fit (highest likelihood value) and lowest Bayesian and Akaike information criteria among alternatives.
- Common variables retained in parsimonious specifications (regressions 2b and 3b for logistic and linear time dependency respectively):
  - External debt (lag 1)
  - Pre-crisis current account balance
  - Capital flows to emerging market countries
  - Three-month LIBOR rate
  - Change in trade-weighted partner country demand
  - Change in primary balance (lag 1)
  - Real interest rate differential (with two- and three-quarter lags)
- Selected coefficient signs and significance patterns (as reported):
  - Quarters in crisis (log): positive and significant (e.g., 2.719**, 2.680***, 2.501*** in some regressions).
  - External debt (lag 1): negative coefficients in parsimonious regressions (e.g., -0.058, -0.051*, -0.039**, -0.049*, -0.036**, -0.042*** across specifications).
  - Current account balance (pre-crisis): positive and significant (e.g., 0.214***, 0.280***, 0.239***, 0.311***, 0.261***, 0.162**).
  - World interest rate (three-month LIBOR): negative and often significant (e.g., -0.373*; -0.636**; -0.577***; -0.594**; -0.543***; -0.510***).
  - Trade-weighted partner country demand (change): positive and often significant (e.g., 1.318*; 0.985**; 1.165***; 1.282***; 1.354***; 1.159***).
  - Change in primary balance (lag 1): positive and significant in parsimonious specifications (e.g., 0.174**, 0.092**, 0.083**, 0.087*, 0.082*, 0.096***).
  - Real interest rate differential (lag 2): negative and sometimes significant (e.g., -0.050; -0.035**; -0.031; -0.036**; -0.032; -0.040**).
  - Real interest rate differential (lag 3): positive and sometimes significant (e.g., 0.029; 0.038***; 0.032**; 0.039***; 0.035**; 0.044**).
  - Exchange rate regime: negative sign (e.g., -0.370; -0.298; -0.277; -0.386**) with p-values around 0.13 and 0.17 in some time-dependent specifications.
  - IMF financing*program dummy (cumulative IMF financing interacting with program dummy): positive sign across models (e.g., 0.080; 0.052; 0.075; 0.272**) but statistically significant only in the time-invariant baseline hazard specification.

### Diagnostic Results (Box 1)
- Model explanatory power:
  - Baseline model hazard predictions are similar to the Nelson-Aalen estimator of cumulative hazard (based on the 18 observed crisis durations), indicating reasonable explanatory power.
- Role of time-varying explanatory variables:
  - Setting explanatory variables to zero in the baseline model produces a dramatic shift in the predicted probability curve away from the Nelson-Aalen hazard, demonstrating the critical role of time-varying variables.
  - A simple duration model using time only produces a virtually flat probability curve, inconsistent with the data.
- Time dependence specification:
  - Logistic time specification predicts exit probabilities better for durations less than 12 quarters and is preferred for 16 out of 18 crises in the sample.
  - Linear specification provides better fit for longer crises.
- Note on diagnostics for time-invariant model:
  - Technically difficult to apply similar diagnostics for the model with no time variable without arbitrary assumptions on IMF disbursement profiles, but comparable likelihood value indicates significant explanatory power.

### Key Empirical Findings on Determinants of Crisis Duration
- Initial and external conditions are key determinants:
  - Larger pre-crisis current account deficits are associated with longer crises.
  - Higher levels of external debt, and in the time-invariant model short-term external debt, are strongly correlated with longer crises.
- External environment:
  - Benign global liquidity conditions (lower three-month LIBOR) and favorable partner countries’ demand shorten crisis duration significantly.
  - Net private capital flows to emerging markets are only weakly associated with the probability of exiting a crisis and lack robustness across specifications.
- Policy response effects:
  - Fiscal consolidation: change in primary balance (lag 1) shortens crisis duration across specifications, suggesting confidence and signaling effects dominate any contractionary impact.
  - Monetary policy: raising real interest rates (relative to global rates) appears to shorten crisis duration with a lag of three quarters; however, the net effect is not clear-cut because of mixed signs and significance for other lags.
  - Exchange rate regime: a shift toward a more flexible regime during a crisis tends to prolong the crisis (negative coefficient), possibly due to balance-sheet effects and overshooting.
- IMF financial support:
  - Positive coefficient on cumulative IMF financing implies higher probability of exiting crisis with larger financing packages, but statistical significance is not robust across specifications (significant only in time-invariant baseline hazard regression).

### Counterfactual Experiments and Scenario Analysis
- Experiments based on augmented model with logistic time specification (regression 2a):
  - Baseline predicted probability of exiting is constructed using estimated parameters and mean values of model variables.
  - Impact of changes in individual variables assessed by shifting the mean value of a variable by one standard deviation and comparing predicted probabilities and expected durations.
- Main counterfactual insights:
  - Factors outside policymakers’ control at crisis onset (initial and external conditions) have the largest impact on expected crisis duration.
  - Largest gain in shortening expected duration—by roughly 2–3 quarters, or about 30–40 percent of the median duration—is associated with either:
    - Stronger initial conditions (lower pre-crisis current account deficits and moderate external debt), or
    - A more benign external environment (more favorable international liquidity conditions and buoyant trade partners’ demand).
- Descriptive statistics for variables used in counterfactuals (Table 3):
  - Exchange rate regime (N=153): Mean = 6.67; Standard deviation = 2.02
  - Real interest rate differential (lag 2) (N=153): Mean = 6.27; Standard deviation = 43.41
  - Real interest rate differential (lag 3) (N=153): Mean = 6.12; Standard deviation = 40.34
  - Change in primary balance (lag 1) (N=153): Mean = 0.48; Standard deviation = 4.76
  - Capital flows to EM countries (N=153): Mean = 1.41; Standard deviation = 0.89
  - Three-month Libor rate (N=153): Mean = 4.54; Standard deviation = 1.84
  - Trade-weighted domestic demand (change) (N=153): Mean = 0.58; Standard deviation = 0.62
  - External debt (lag 1) (N=153): Mean = 70.04; Standard deviation = 28.92
  - Current account balance (pre-crisis) (N=18): Mean = -2.45; Standard deviation = 3.89
  - Cumulative IMF financing (N=151): Mean = 2.96; Standard deviation = 13.75

### Scenarios Illustrated (Figure 4 summary)
- Predicted probabilities of exit under different scenarios (quarters in crisis on x-axis):
  - Baseline probability (mean explanatory variables)
  - Strong initial conditions
  - Benign external conditions
  - Strong policy response
  - Greater IMF financing
- Evaluation measures:
  - Vertical comparison at median crisis duration of seven quarters shows differences in predicted hazard rates across scenarios.
  - Horizontal comparison at a chosen probability threshold (one-half) shows differences in expected crisis duration across scenarios.

*Source: IMF staff calculations, Appendix II.*

### 0.5 probability of exit from the crisis.

### _wp09103 - 0.5 probability of exit from the crisis.

### Key empirical findings on determinants of crisis exit and duration
- Median crisis duration: seven quarters.
- Average crisis duration: about 8½ quarters.
- Crisis duration range in sample: from three quarters (1998 Turkey) to 18 quarters (1997 Thailand).
- Initial conditions and the external environment are key explanatory variables for crisis duration; consistently prudent macroeconomic policies strengthen initial conditions (the “good policies” factor).
- Favorable external conditions matter as a “good luck” factor, influencing the probability of exit once a crisis occurs.
- Complexity of a capital account crisis is strongly linked to its duration.

### Counterfactual experiment results (policy response and other covariates)
- Stronger policy response increases probability of exit and shortens duration:
  - A stronger improvement in the primary fiscal balance (by one standard deviation) is found to increase the probability of exiting from a crisis by about 10 percent, or equivalently shorten crisis duration by about one quarter.
  - Maintaining the pre-crisis exchange rate regime (avoiding large exchange rate devaluations during the crisis) is found to increase the probability of exiting from a crisis by about 10 percent, or equivalently shorten crisis duration by about one quarter.
  - A one standard deviation higher real interest rate differential is predicted to raise the probability of exiting from crisis marginally by 3 percent, and to shorten duration by about 0.4 quarters.
- A one standard deviation higher IMF financing would seem to increase the probability of exiting from a crisis by about 12 percent; this result is based on a regression coefficient which is not statistically significant and should be interpreted with caution because it is difficult to separate the impact of Fund financing from the effects of a Fund-supported program on a country’s policy response variables.
- Other modeled scenario impacts (evaluated at median crisis duration of seven quarters or at 0.5 probability of staying in crisis) include positive effects from:
  - Capital flows to EM + 1SD
  - Trade-weighted partner country demand + 1SD
  - CAB-to-GDP (pre-crisis) + 1SD
  - LIBOR - 1SD
  - External debt-to-GDP (t-1) - 1SD
  (Specific figure numerics are reported in Figures 5 and 6 of the source.)

### Policy-relevant implications and interpretation
- Good policies matter primarily through shaping initial conditions prior to crises; once a crisis erupts it is often too late to correct weak initial conditions.
- Appropriate fiscal adjustment during a crisis can shorten crisis duration significantly.
- Evidence for monetary policy is mixed, but an increase in (real) interest rates may help shorten a crisis.
- Changes in the exchange rate regime during a crisis—often associated with sharp depreciations—tend to be associated with longer crises, likely reflecting adverse balance sheet effects.
- Strong market pressures during crises severely limit actual policy options; this reinforces the importance of crisis prevention efforts.

### Measurement approach for crisis duration (IKAC index)
- Crisis duration is measured using an Index of Capital Account Crises (IKAC) constructed from quarterly data on:
  - foreign exchange reserves (FX),
  - the nominal effective exchange rate (NEER),
  - secondary market spreads on sovereign bonds (S),
  - net private capital flows scaled as a ratio to GDP (K).
- The IKAC associates:
  - Escalation of pressure indicators (FX, NEER, spreads) with crisis onset and stabilization with crisis end.
  - Capital outflows (sudden stops) with crisis onset and resumption of inflows with crisis end.
  - Deviation of spreads from “normal” levels with crisis onset and return to “normal” with crisis end.
- Index construction specifics:
  - Each term is standardized (mean zero, standard deviation one); positive index values denote financial pressure.
  - NEER enters in percentage changes; spreads and capital flows are compared to country-specific sample means; reserves are compared to long-run trend.
- Start of a crisis: first of two consecutive quarters in which IKAC > 0.
- End of a crisis: first of two consecutive quarters in which IKAC < 0.
- Application of IKAC to 12 emerging market economies produced 18 crisis episodes; 16 out of 18 crises in the sample lasted less than 12 quarters.

### Methodological note on duration modeling
- The paper uses survival analysis with a proportional hazard specification and time-varying covariates.
- The complementary log-log model is employed for discrete hazard rates; the baseline hazard may be specified as λ0(t)=log(t).
- Unobserved heterogeneity is modeled via an error term u, with potential distributions (Normal, Gamma, Inverse Gaussian) considered; heterogeneity is assumed independent of observed covariates.
- Parameters are estimated by maximizing the log-likelihood constructed from the discrete hazard probabilities.

*Source: IMF staff calculations and analysis in _wp09103 - 0.5 probability of exit from the crisis.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09103.pdf_