## 5. Regulatory Environment, Contestability, and the Timing and Probability of Crises

## Source details

**Canonical URL:** [5. Regulatory Environment, Contestability, and the Timing and Probability of Crises](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2006/_wp06143.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2006/_wp06143.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2006/_wp06143.pdf.json)

---

### Key findings and policy-relevant conclusions
- Sample and scope
  - Cross-country analysis of 38 countries over the period 1980–2003.
  - Competitiveness measured by the Panzar and Rosse (1987) H-Statistic (data for 1994–2001 assumed constant over 1980–2003).
  - Crisis dating follows Demirgüç-Kunt and Detragiache (2005); up to 28 systemic crises recorded for the logit models and 22 crisis spans used in the duration models (depending on specification).

- Core empirical findings
  - Competitive behavior of financial institutions (H-Statistic) is associated with greater banking system stability:
    - Exponential duration models with time-varying covariates: H-Statistic enters positively and significantly (Specification (2): coefficient = 5.0854 with standard error = (1.7922)***; Specification (3): coefficient = 4.2562 with standard error = (1.9919)**).
    - Logit models: H-Statistic enters negatively and significantly for the probability of a crisis (Specification (6): coefficient = -5.8513 with standard error = (1.7523)***; Specification (7): coefficient = -5.2130 with standard error = (1.8819)***).
    - Positive duration-model coefficients interpreted as increased survival time; negative logit coefficients interpreted as reduced probability of experiencing a systemic crisis.
  - Evidence supports the “competition–stability” view; no empirical support found for the “competition–fragility” view.
  - Relationship between the H-Statistic and increased survival time holds when controlling for concentration (Specification (3)) and is robust to alternative samples, sampling periods, first differences for macro controls, and a market-based competition measure.
  - Concentration (average proportion of assets held by the three largest institutions) sometimes enters with a positive coefficient but is often insignificant once competition is controlled for; correlation between concentration and the H-Statistic is 0.1503.
  - Interaction H-Statistic * Concentration: when included, competition and concentration may lose individual significance in some specifications (Specification (4)); supplementary unreported regressions show the interaction term positive and significant in the duration model (5 percent level) and negative and significant in the logit model (1 percent level).

- Additional significant covariates (examples from Table 3)
  - Credit growth (real) repeatedly significant:
    - Duration models: coefficient = -0.0014 with standard error = (0.0004)*** in Specification (1); similar significance across Specifications (2)–(4).
    - Logit models: coefficients positive and significant in Specifications (6)–(8) (e.g., Specification (6): coefficient = 0.0016 with standard error = (0.0008)**).
  - Depreciation significant in some duration specifications (Specification (1): coefficient = -0.0039 with standard error = (0.0015)***).
  - Legal origin dummies (French, Scandinavian) significant in multiple specifications (examples: French legal origin duration coeff = -1.1209 with (0.6180)*; logit coeff = 1.6562 with (0.6230)***).

### Data, variables, and summary statistics (selected exact values)
- Sample: 38 countries, 1980–2003.
- Key descriptive statistics (Table 1 excerpt):
  - GDP growth (real) — Mean 3.20, Std.Dev. 3.13, Minimum -13.13, Maximum 14.82
  - Real interest rate — Mean 1.59, Std.Dev. 25.49, Minimum -558.91, Maximum 48.86
  - Inflation — Mean 16.67, Std.Dev. 83.40, Minimum -16.33, Maximum 2076.79
  - Terms of Trade — Mean -0.72, Std.Dev. 33.98, Minimum -607.00, Maximum 622.00
  - Depreciation — Mean 19.14, Std.Dev. 116.21, Minimum -320.37, Maximum 2421.59
  - M2/international reserves — Mean 10.06, Std.Dev. 9.37, Minimum 0.78, Maximum 59.48
  - Credit growth (real) — Mean 73.76, Std.Dev. 171.60, Minimum -256.35, Maximum 1421.95
  - Moral hazard index — Mean 1.41, Std.Dev. 0.72, Minimum 0.00, Maximum 2.03
  - Concentration (three-bank share averaged) — Mean 0.47, Std.Dev. 0.13, Minimum 0.16, Maximum 0.69
  - H-Statistic — Mean 0.67, Std.Dev. 0.12, Minimum 0.41, Maximum 0.92
  - Activity restrictions — Mean 9.36, Std.Dev. 2.59, Minimum 5, Maximum 15
  - Capital regulatory index — Mean 5.94, Std.Dev. 1.45, Minimum 3, Maximum 9
  - Government ownership — Mean 0.47, Std.Dev. 0.34, Minimum 0, Maximum 1
  - Foreign ownership — Mean 0.18, Std.Dev. 0.24, Minimum 0, Maximum 0.95

- Crisis classification criteria (Demirgüç-Kunt and Detragiache (2005))
  - One of the following must be met for a systemic crisis classification:
    - emergency measures such as deposit freezes or bank holidays are implemented;
    - large-scale bank nationalizations take place;
    - non-performing assets reach at least 10 percent of total assets;
    - fiscal cost of rescue operations reach 2 percent of GDP.
  - Depending on model specification, up to 28 systemic crises recorded between 1980 and 2003 for the logit model; duration analysis counts slightly lower due to span-based setup.

### Methodology summary
- Duration analysis
  - Parametric exponential duration model with time-varying covariates estimated by maximum likelihood.
  - Dependent variable: log of time to crisis (right-censored if no crisis during observation period).
  - Exponential specification assumed (constant baseline hazard λ0(t)); Weibull tests indicate α=1 cannot be rejected.
  - Coefficients reported in the accelerated failure time (AFT) metric (positive AFT coefficients imply increased survival time).
- Logistic probability analysis
  - Standard logit model estimating probability of observing a crisis in a given year; explanatory variables lagged by one period.
  - Provides complementary unconditional probability estimates to duration analysis’ conditional hazard framework.
- Competition measure
  - Panzar and Rosse (1987) H-Statistic: sum of elasticities of reduced-form revenues with respect to factor prices; interpretation:
    - H ≤ 0 indicates monopoly equilibrium
    - 0 < H < 1 indicates monopolistic competition
    - H = 1 indicates perfect competition

### Regression results — main specifications (selected exact coefficient values)
- Duration models (exponential AFT; dependent variable = log time to crisis; Observations = 619; Number of crises = 22)
  - Specification (1): canonical controls (no H-Statistic, no concentration)
    - Constant = 4.6274 (0.6236)***
    - Credit growth (real) = -0.0014 (0.0004)***
    - Depreciation = -0.0039 (0.0015)***
  - Specification (2): adds H-Statistic
    - H-Statistic = 5.0854 (1.7922)***
    - GDP growth (real) = -0.0214 (0.0904)
    - Credit growth (real) = -0.0019 (0.0005)***
  - Specification (3): H-Statistic + Concentration
    - H-Statistic = 4.2562 (1.9919)**
    - Concentration (three-bank average) = 2.3486 (2.4725) [insignificant]
  - Specification (4): adds H-Statistic * Concentration interaction
    - H-Statistic and Concentration lose individual significance when interaction included; interaction reported in text as significant in unreported variants.

- Logit models (dependent variable = crisis dummy; Observations = 567; Number of crises = 28)
  - Specification (5): canonical controls
    - Constant = -5.4307 (0.7062)***
    - Real interest rate = 0.0219 (0.0100)**
    - Inflation = 0.0053 (0.0031)*
  - Specification (6): adds H-Statistic
    - H-Statistic = -5.8513 (1.7523)***
    - Credit growth (real) = 0.0016 (0.0008)**
  - Specification (7): H-Statistic + Concentration
    - H-Statistic = -5.2130 (1.8819)***
    - Concentration = -11.8460 (11.2600) [insignificant]
  - Specification (8): includes interaction H-Statistic * Concentration
    - H-Statistic = -1.7627 (6.2430) [insignificant], interaction coefficient reported as -7.8430 (13.1666) (insignificant in Table); authors note unreported regressions where interaction is significant.

- Model diagnostics (logit panel)
  - Type I Error (%) for logit specifications: 25.00, 28.57, 25.00, 25.00 (Specifications (5)–(8)).
  - Type II Error (%) for logit specifications: 41.37, 33.95, 33.58, 33.21 (Specifications (5)–(8)).
  - Akaike Info. Criterion (duration models): 0.1610, 0.1555, 0.1571, 0.1578 (Specifications (1)–(4)); logit AIC: 0.400, 0.387, 0.390, 0.393 (Specifications (5)–(8)).
  - Pseudo R2 (logit): 0.091, 0.132, 0.135, 0.136 (Specifications (5)–(8)).

### Robustness, caveats, and interpretation limits
- Measurement and timing caveats
  - H-Statistic computed for 1994–2001; in some cases competitiveness measured after a crisis. Authors use alternative samples and sampling periods to mitigate concern but caution remains.
  - Regulatory variables collected toward end of sampling period; interpretation of regressions including regulatory variables should be cautious.
  - Crisis dating matters; authors use updated Demirgüç-Kunt and Detragiache (2005) scheme.
- Causality
  - Analysis does not establish causality; authors refrain from causal interpretation.
- Robustness checks and outcomes
  - Semiparametric Cox proportional hazards models with time-varying covariates produced very similar results on the H-Statistic contribution to system soundness.
  - Weibull specifications tested; Wald tests did not reject α=1, supporting exponential/constant hazard assumption.
  - Robustness exercises (examples reported):
    - Omitting the period 1994–2001 when the H-Statistic measured.
    - Excluding low-income economies (Bangladesh, India, Kenya, Nigeria, Pakistan).
    - Using first differences for macroeconomic control variables.
    - Constraining sampling horizon to 1985–2003.
    - Excluding major industrial countries.
  - Outcomes: significant relationship between competitiveness and banking system soundness persists across robustness exercises in both duration and logit frameworks.
  - H-Statistic sometimes rendered insignificant when foreign ownership included in one duration specification, but otherwise retains significance across tests.

### Quantitative magnitudes and interpretation (selected)
- Total number of crisis observations in the main sample: 28.
- Number of observations in the main sample referenced for error cutoff: 567.
- Type I / Type II Error cutoff point computed as 28 / 567 = 0.0494.
- H-Statistic: one standard deviation = 0.12.
  - Marginal effect used from the logit model (evaluated at the mean): -0.1498.
  - Computation reported: -0.1498 * 0.12 = -0.0180 (interpretation: a one standard deviation increase in competitiveness decreases the probability of observing a crisis by 1.8 percent).
- Concentration: one standard deviation = 0.13.
  - Marginal effect from logit reported: -0.0528.
  - Reported effect: decreases the probability of suffering a systemic crisis by 0.53 percent (-0.0528*0.13).

### Findings on regulatory, institutional, and macro controls
- Terms of trade change: weakly significant at the 10 percent level in one specification (Specification (2)) with inverse relationship to time to crisis; interpretation cautioned.
- Rate of depreciation: negative sign and significant in Specifications (1), (3), and (4); currency devaluations threaten bank profitability.
- Credit growth: strong evidence for the “boom and bust” hypothesis; strong credit growth shortens survival time (significant across multiple specifications).
- Legal origin dummies:
  - French legal origin dummy: negative and significant in duration regressions (shorter time to crisis); positive and significant in logit regressions (higher propensity to experience a crisis).
  - Scandinavian legal origin dummy: negative and significant in duration regressions.
- Regulatory and ownership environment (Table 5):
  - Activity restrictions become positive and significant at the 5 percent level in the logit model, indicating more restrictive environments are conducive to buildup of banking vulnerabilities.
  - Capital regulatory index, foreign ownership, and government ownership generally do not enter at meaningful significance levels in duration models; H-Statistic remains significant across most specifications in logit models.

### Comparative inference on competition vs. concentration
- Competitive conduct (H-Statistic) has a distinct and stronger relationship with banking-system fragility than industry concentration.
- Concentration does not significantly impact banking system soundness when competition is controlled for in most specifications.
- Results challenge findings that bank concentration boosts banking stability when competitive conduct is not accounted for.

### Policy-relevant implications and suggested further research
- Implication: Policies that enhance competitive behavior in banking systems may increase resilience to systemic crises; competition and soundness generally go hand in hand.
- Regulatory design: Excessive entry restrictions (activity restrictions) are associated with greater banking vulnerabilities; contestability may matter for stability.
- Suggested further research:
  - Test alternative measures of competitive behavior and identify potentially optimal levels of competition for stability.
  - Firm-level cross-country studies controlling for institutional and regulatory environment to explore transmission mechanisms.
  - Examine short-run versus long-run effects of competition on stability.
  - Use more continuous measures of fragility (e.g., distance to default), subject to data reliability constraints.

*Source: Excerpt from IMF working paper (_wp06143) — "Regulatory Environment, Contestability, and the Timing and Probability of Crises" (pages and tables as provided in the source content).*

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

### References

### Tables
- 1. Descriptive Statistics............................................................................................................14
- 2. Banking Sector Crises..........................................................................................................17
- 3. Competitiveness, Timing, and Probability of Systemic Banking Crises .............................21
- 4. Robustness Tests for Competitiveness and Timing and Probability of Crises ....................24

*Source: _wp06143 - References..............................................................................................................*

### 5. Regulatory Environment, Contestability, and the Timing and Probability of Crises ..........26

### 5. Regulatory Environment, Contestability, and the Timing and Probability of Crises

### Key findings and policy-relevant conclusions
- Sample and scope
  - Cross-country analysis of 38 countries over the period 1980–2003.
  - Competitiveness measured by the Panzar and Rosse (1987) H-Statistic (data for 1994–2001 assumed constant over 1980–2003).
  - Crisis dating follows Demirgüç-Kunt and Detragiache (2005); up to 28 systemic crises recorded for the logit models and 22 crisis spans used in the duration models (depending on specification).

- Core empirical findings
  - Competitive behavior of financial institutions (H-Statistic) is associated with greater banking system stability:
    - In the exponential duration models with time-varying covariates, the H-Statistic enters positively and significantly (Specification (2): coefficient = 5.0854 with standard error = (1.7922)***; Specification (3): coefficient = 4.2562 with standard error = (1.9919)**).
    - In the logit models, the H-Statistic enters negatively and significantly for the probability of a crisis (Specification (6): coefficient = -5.8513 with standard error = (1.7523)***; Specification (7): coefficient = -5.2130 with standard error = (1.8819)***).
    - The authors interpret positive duration-model coefficients as increasing time to crisis (i.e., increased survival time) and negative logit coefficients as reducing the probability of experiencing a systemic crisis.
  - The evidence supports the “competition–stability” view; no empirical support found for the “competition–fragility” view.
  - The positive relationship between the H-Statistic and increased survival time holds when controlling for concentration (Specification (3)), and is robust to alternative samples, different sampling periods, first differences for macro controls, and incorporating a measure of competition from financial markets.
  - Concentration (average proportion of assets held by the three largest institutions) enters with a positive coefficient in some specifications but is often insignificant once competition is controlled for; correlation between concentration and the H-Statistic is 0.1503.
  - Interaction between H-Statistic and concentration: when the interaction term is included, neither competition nor concentration is significant in some specifications (Specification (4)); supplementary unreported regressions show the interaction term positive and significant in the duration model (5 percent level) and negative and significant in the logit model (1 percent level) (results available on request).

- Additional significant covariates (examples from Table 3)
  - Credit growth (real) repeatedly significant:
    - Duration models: coefficient = -0.0014 with standard error = (0.0004)*** in Specification (1); similar significance across Specifications (2)–(4).
    - Logit models: coefficients positive and significant in Specifications (6)–(8) (e.g., Specification (6): coefficient = 0.0016 with standard error = (0.0008)**).
  - Depreciation shows significance in some duration specifications (Specification (1): coefficient = -0.0039 with standard error = (0.0015)***).
  - Legal origin dummies (French, Scandinavian) are significant in multiple specifications (examples: French legal origin duration coeff = -1.1209 with (0.6180)*; logit coeff = 1.6562 with (0.6230)***).

- Methodological contributions
  - First cross-country investigation using the Panzar and Rosse (1987) H-Statistic as a direct measure of competitive conduct (rather than concentration proxies).
  - Introduction of parametric duration models with time-varying covariates to analyze time to systemic banking crises; estimates produced in the accelerated failure time (AFT) metric and contrasted with standard logit probability models.
  - Use of time-varying covariates to account for panel structure and multiple observations per country.
  - Robustness checks reported, including semiparametric Cox models and Weibull specifications (α tests support exponential/constant hazard specification).

### Literature context and conceptual points
- Competing theoretical strands:
  - “Competition–fragility” literature: competition erodes charter values and incentivizes risk-taking (Smith, 1984; Keeley, 1990; Besanko and Thakor, 1993).
  - “Competition–stability” literature: monopoly or reduced contestability can encourage risk or poor asset quality (Caminal and Matutes, 2002; Perotti and Suarez, 2002; Beck, Demirgüc-Kunt, and Levine findings).
  - Allen and Gale (2004): relationship multifaceted; efficient levels of competition and stability depend on model features (deposit insurance, interbank market structure, inflation/monetary policy interactions).
- Regulatory environment and contestability:
  - Evidence that less contestable markets and higher entry/activity restrictions are associated with higher fragility in some studies (Barth, Caprio, and Levine; Beck, Demirgüc-Kunt, and Levine).
  - Debate persists; regulatory variables used here are collected toward the end of the sampling period and treated cautiously.

### Data, variables, and summary statistics (selected exact values preserved)
- Sample: 38 countries, 1980–2003.
- Key descriptive statistics (Table 1 excerpt):
  - GDP growth (real) — Mean 3.20, Std.Dev. 3.13, Minimum -13.13, Maximum 14.82
  - Real interest rate — Mean 1.59, Std.Dev. 25.49, Minimum -558.91, Maximum 48.86
  - Inflation — Mean 16.67, Std.Dev. 83.40, Minimum -16.33, Maximum 2076.79
  - Terms of Trade — Mean -0.72, Std.Dev. 33.98, Minimum -607.00, Maximum 622.00
  - Depreciation — Mean 19.14, Std.Dev. 116.21, Minimum -320.37, Maximum 2421.59
  - M2/international reserves — Mean 10.06, Std.Dev. 9.37, Minimum 0.78, Maximum 59.48
  - Credit growth (real) — Mean 73.76, Std.Dev. 171.60, Minimum -256.35, Maximum 1421.95
  - Moral hazard index — Mean 1.41, Std.Dev. 0.72, Minimum 0.00, Maximum 2.03
  - Concentration (three-bank share averaged) — Mean 0.47, Std.Dev. 0.13, Minimum 0.16, Maximum 0.69
  - H-Statistic — Mean 0.67, Std.Dev. 0.12, Minimum 0.41, Maximum 0.92
  - Activity restrictions — Mean 9.36, Std.Dev. 2.59, Minimum 5, Maximum 15
  - Capital regulatory index — Mean 5.94, Std.Dev. 1.45, Minimum 3, Maximum 9
  - Government ownership — Mean 0.47, Std.Dev. 0.34, Minimum 0, Maximum 1
  - Foreign ownership — Mean 0.18, Std.Dev. 0.24, Minimum 0, Maximum 0.95

- Crisis classification criteria (Demirgüç-Kunt and Detragiache (2005) root source):
  - One of the following must be met for a systemic crisis classification:
    - emergency measures such as deposit freezes or bank holidays are implemented;
    - large-scale bank nationalizations take place;
    - non-performing assets reach at least 10 percent of total assets;
    - fiscal cost of rescue operations reach 2 percent of GDP.
  - Depending on the model specification, up to 28 systemic crises recorded between 1980 and 2003 for the logit model; duration analysis counts are slightly lower due to span-based setup.

### Methodology summary
- Duration analysis
  - Parametric exponential duration model with time-varying covariates estimated by maximum likelihood.
  - Dependent variable: log of time to crisis (right-censored if no crisis during observation period).
  - Exponential specification assumed (constant baseline hazard λ0(t)); Weibull tests indicate α=1 cannot be rejected.
  - Coefficients reported in the accelerated failure time (AFT) metric (positive AFT coefficients imply increased survival time).
- Logistic probability analysis
  - Standard logit model estimate of probability of observing a crisis in a given year; explanatory variables lagged by one period.
  - Provides complementary unconditional probability estimates to duration analysis’ conditional hazard framework.
- Competition measure
  - Panzar and Rosse (1987) H-Statistic: sum of elasticities of reduced-form revenues with respect to factor prices; interpretation:
    - H ≤ 0 indicates monopoly equilibrium
    - 0 < H < 1 indicates monopolistic competition
    - H = 1 indicates perfect competition

### Regression results — main specifications (selected exact coefficient values from Table 3)
- Duration models (exponential AFT; dependent variable = log time to crisis; Observations = 619; Number of crises = 22)
  - Specification (1): canonical controls (no H-Statistic, no concentration)
    - Constant = 4.6274 (0.6236)***
    - Credit growth (real) = -0.0014 (0.0004)***
    - Depreciation = -0.0039 (0.0015)***
  - Specification (2): adds H-Statistic
    - H-Statistic = 5.0854 (1.7922)***
    - GDP growth (real) = -0.0214 (0.0904)
    - Credit growth (real) = -0.0019 (0.0005)***
  - Specification (3): H-Statistic + Concentration
    - H-Statistic = 4.2562 (1.9919)**
    - Concentration (three-bank average) = 2.3486 (2.4725) [insignificant]
  - Specification (4): adds H-Statistic * Concentration interaction
    - H-Statistic and Concentration lose individual significance when interaction included; interaction reported in text as significant in unreported variants.
- Logit models (dependent variable = crisis dummy; Observations = 567; Number of crises = 28)
  - Specification (5): canonical controls
    - Constant = -5.4307 (0.7062)***
    - Real interest rate = 0.0219 (0.0100)**
    - Inflation = 0.0053 (0.0031)*
  - Specification (6): adds H-Statistic
    - H-Statistic = -5.8513 (1.7523)***
    - Credit growth (real) = 0.0016 (0.0008)**
  - Specification (7): H-Statistic + Concentration
    - H-Statistic = -5.2130 (1.8819)***
    - Concentration = -11.8460 (11.2600) [insignificant]
  - Specification (8): includes interaction H-Statistic * Concentration
    - H-Statistic = -1.7627 (6.2430) [insignificant], interaction coefficient reported as -7.8430 (13.1666) (insignificant in Table); authors note interaction results in unreported regressions (duration positive sig., logit negative sig.)

- Model diagnostics and classification errors (logit panel)
  - Type I Error (%) reported for logit specifications: 25.00, 28.57, 25.00, 25.00 (Specifications (5)–(8) respectively).
  - Type II Error (%) reported for logit specifications: 41.37, 33.95, 33.58, 33.21 (Specifications (5)–(8) respectively).
  - Akaike Info. Criterion (duration models): 0.1610, 0.1555, 0.1571, 0.1578 (Specifications (1)–(4)); (logit AIC: 0.400, 0.387, 0.390, 0.393 for Specifications (5)–(8)).
  - Pseudo R2 (logit): 0.091, 0.132, 0.135, 0.136 (Specifications (5)–(8)).

### Robustness, caveats, and interpretation limits (as stated in the source)
- Measurement and timing caveats
  - H-Statistic computed for 1994–2001; in some cases competitiveness measured after a crisis. Authors use alternative samples and sampling periods to mitigate concern but caution remains.
  - Regulatory variables collected toward end of sampling period; interpretation of regressions including regulatory variables should be cautious.
  - Crisis dating matters; authors use updated Demirgüç-Kunt and Detragiache (2005) scheme.
- Causality
  - Analysis does not establish causality (i.e., cannot definitively state competition causes stability or vice versa); authors refrain from causal interpretation.
- Additional robustness checks (reported or mentioned)
  - Semiparametric Cox proportional hazards models with time-varying covariates produced very similar results on H-Statistic contribution to system soundness.
  - Weibull specifications tested; Wald tests did not reject α=1, supporting exponential/constant hazard assumption.
  - Regressions with alternative legal origin and institutional proxies leave H-Statistic significance virtually unchanged (results available on request).

*Source: _wp06143 - 5. Regulatory Environment, Contestability, and the Timing and Probability of Crises ..........26 (IMF working paper content provided).*

### Appendix provides detailed information on the explanatory variables. Specifications (1) and (5) are our baseline models 

### _wp06143 - Appendix provides detailed information on the explanatory variables. Specifications (1) and (5) are our baseline models

### Main empirical result
- Higher degrees of competition in banking systems (measured by the Panzar and Rosse (1987) H-Statistic) are associated with:
  - A decreased probability of suffering a systemic banking crisis.
  - Longer survival time (time to crisis) for banking systems.
- The key finding is characterized as support for the “competition–stability” view and does not conform to the “competition–fragility” literature.

### Key quantitative findings and statistics
- Total number of crisis observations in the main sample: 28.
- Number of observations in the main sample referenced for error cutoff: 567.
- Type I / Type II Error cutoff point computed as 28 / 567 = 0.0494.
- H-Statistic: one standard deviation = 0.12.
  - Marginal effect used from the logit model (evaluated at the mean): -0.1498.
  - Computation reported: -0.1498 * 0.12 = -0.0180 (stated interpretation: a one standard deviation increase in competitiveness decreases the probability of observing a crisis by 1.8 percent).
- Concentration: one standard deviation = 0.13.
  - Marginal effect from logit reported: -0.0528.
  - Reported effect: decreases the probability of suffering a systemic crisis by 0.53 percent (-0.0528*0.13).
- Table 4 robustness sample results (selected):
  - Duration-model H-Statistic coefficients (Specifications (1)–(4)): 6.0382, 5.7925, 4.9358, 5.3638 (standard errors shown in table).
  - Logit-model H-Statistic coefficients (Specifications (5)–(8)): -6.5980, -6.1835, -5.2564, -4.3798 (standard errors shown in table).
  - Observations reported across robustness regressions: 388, 534, 594, 504 (duration models) and 395, 481, 538, 452 (logit models) in Table 4.
  - Number of crises in corresponding regressions in Table 4: 13, 19, 21, 20 (duration) and 20, 25, 25, 21 (logit).
  - Type I Error (%) in Table 4 logit columns: 28.27%, 16.00, 32.00, 23.81.
  - Type II Error (%) in Table 4 logit columns: 25.00%, 28.73, 30.02, 32.48.
  - Akaike Info. Criterion entries in Table 4: 0.2368, 0.1453, 0.1484, 0.2002 (duration); 0.418, 0.395, 0.379, 0.397 (logit).
  - Pseudo R2 reported in Table 4 logit columns (where available): 0.133, 0.1761, 0.129, 0.108.
- Table 5 selected numbers (duration and logit with regulatory variables):
  - H-Statistic coefficients in duration models (Specs (1)–(4)): 4.3392, 4.4300, 3.8345, 3.6107 (standard errors reported).
  - H-Statistic coefficients in logit models (Specs (5)–(8)): -5.7529, -5.4366, -5.3887, -4.3336 (standard errors reported).
  - Observations across Table 5 regressions: 598, 598, 409, 598 (duration) and 547, 547, 371, 545 (logit).
  - Number of crises in Table 5 regressions: 22, 22, 18, 22 (duration) and 28, 28, 20, 28 (logit).
  - Type I Error (%) for Table 5 logit columns: 17.86, 25.00, 25.00, 21.43.
  - Type II Error (%) for Table 5 logit columns: 31.98, 29.29, 27.07, 31.53.
  - Akaike Info. Criterion entries in Table 5 (duration): 0.1609, 0.1611, 0.1829, 0.1631; (logit): 0.393, 0.401, 0.432, 0.404.
  - Pseudo R2 entries in Table 5 logit columns: 0.163, 0.144, 0.151, 0.1376.

### Robustness tests and methodological checks
- Two methodological approaches used: exponential duration models and logit models.
  - Specifications (1) and (5) are the baseline models.
  - White’s heteroskedasticity consistent standard errors are reported in parentheses.
- Robustness checks include:
  - Omitting the period 1994–2001 when the H-Statistic was measured (Claessens and Laeven (2004)).
  - Excluding low-income economies (Bangladesh, India, Kenya, Nigeria, Pakistan).
  - Using first differences for macroeconomic control variables instead of levels.
  - Constraining the sampling horizon to 1985–2003.
  - Excluding major industrial countries (results reported as available upon request).
  - Clustering error terms (not reported because findings on H-Statistic virtually unchanged).
- Outcomes of robustness tests:
  - The significant relationship between competitiveness and banking system soundness persists across robustness exercises in both duration and logit frameworks.
  - The H-Statistic typically enters negatively and significantly in logit models (significance levels often at 1 percent or 5 percent) and positively and significantly in duration models (consistent with longer time to crisis).
  - Concentration (three-bank concentration ratio) generally remains insignificant when competition is controlled for.
  - The H-Statistic is rendered insignificant in one duration specification when foreign ownership is included, but otherwise retains significance across tests.

### Findings on control variables and institutional/regulatory environment
- Terms of trade change: weakly significant at the 10 percent level in one specification (Specification (2)) with an inverse relationship to time to crisis; interpretation cautioned due to sample composition and commodity-export dependency.
- Rate of depreciation: negative sign and significant in Specifications (1), (3), and (4); currency devaluations threaten bank profitability.
- Credit growth: provides strong evidence for the “boom and bust” hypothesis; strong credit growth shortens survival time (significant across multiple specifications).
- Legal origin dummies:
  - French legal origin dummy: negative and significant in duration regressions (shorter time to crisis); positive and significant in logit regressions (higher propensity to experience a crisis).
  - Scandinavian legal origin dummy: negative and significant in duration regressions.
- Macroeconomic controls: some lack significance, potentially due to multicollinearity; retained to control macroeconomic environment while testing competition effect.
- Regulatory and ownership environment (Table 5):
  - Activity restrictions become positive and significant at the 5 percent level in the logit model, indicating more restrictive environments are conducive to buildup of banking vulnerabilities.
  - Capital regulatory index, foreign ownership, and government ownership generally do not enter at meaningful significance levels in duration models; logit models show the H-Statistic remains significant across most specifications.
  - Overall inference: regulatory and institutional environment do not eliminate the independent effect of the H-Statistic on fragility.

### Comparative inference on competition vs. concentration
- The competitive conduct of banks (H-Statistic) has a distinct and stronger relationship with banking-system fragility than industry concentration.
- Concentration does not significantly impact banking system soundness when competition is controlled for in most specifications.
- The results challenge findings that bank concentration boosts banking stability when competitive conduct is not accounted for.

### Policy-relevant implications and suggested further research
- Implication: Policies that enhance competitive behavior in banking systems may increase resilience to systemic crises; competition and soundness generally go hand in hand.
- Regulatory design: Excessive entry restrictions (activity restrictions) are associated with greater banking vulnerabilities; contestability may matter for stability.
- Open questions for further research suggested by the authors:
  - Test alternative measures of competitive behavior and identify potentially optimal levels of competition for stability.
  - Firm-level cross-country studies controlling for institutional and regulatory environment to explore transmission mechanisms.
  - Examine short-run versus long-run effects of competition on stability.
  - Use more continuous measures of fragility (e.g., distance to default), subject to data reliability constraints.

*Source: Excerpt from the Appendix and results discussion of the IMF working paper (_wp06143) covering methodology, robustness checks, regression tables, and variable definitions.*

### REFERENCES

### _wp06143 - REFERENCES

### Financial contagion, stability, and crises
- Allen, Franklin, and Douglas Gale, 2000,  “Financial Contagion,” Journal of Political Economy, Vol. 108, pp. 1–29.
- Allen, Franklin, and Douglas Gale, 2004, “Competition and Financial Stability,” Journal of Money, Credit and Banking, Vol. 36, No. 3, pp. 453–80.
- Demirgüç-Kunt, Asli and Enrica Detragiache, 1998,  “The Determinants of Banking Crises in Developing and Developed Countries,” IMF Staff Papers, Vol. 45, No. 1, pp. 81–109.
- Demirgüç-Kunt, Asli and Enrica Detragiache, 2002,  “Does Deposit Insurance Increase Banking System Stability? An Empirical Investigation,” Journal of Monetary Economics, Vol. 49, pp. 1373–1406.
- Demirgüç-Kunt, Asli, Baybars Karacaovali, and Luc Laeven, 2005,  “Deposit Insurance Around the World: A Comprehensive Database,” Policy Research Working Paper #3628 (Washington: World Bank), available at http://web.worldbank.org/servlets/ECR?entityID=000012009_20050608111717&collection=IMAGEBANK&sitePK=469233
- Detragiache, Enrica, and Antonio Spilimbergo, 2001,  “Crises and Liquidity–Evidence and Interpretation,” IMF Working Paper 01/2 (Washington: International Monetary Fund).
- Drees, Burkhard and Ceyla Pazarbasioglu (1998) The Nordic Banking Crises: Pitfalls in Financial Liberalization? IMF Occasional Paper 161/1998, Washington, D. C.: International Monetary Fund
- Eichengreen, Barry M., and Carlos Arteta, 2000,  “Banking Crises in Emerging Markets: Presumptions and Evidence,” Center for International and Development Economics Research Paper C00’115 (Berkeley: University of California).
- Honohan, Patrick, and Luc Laeven (eds.), 2005, Systemic Financial Crises: Containment and Resolution (Cambridge, Massachusetts: Cambridge University Press).

### Bank competition, concentration, consolidation, and market structure
- Beck, Thorsten, Asli Demirgüç-Kunt and Ross Levine, 2005a,  “Bank Concentration, Competition, and Crises: First Results,” Journal of Banking and Finance, Vol. 30, No. 5, pp. 1581–603.
- Beck, Thorsten, Asli Demirgüç-Kunt and Ross Levine, 2005b,  “Bank Concentration and Fragility: Impact and Mechanics,” in Risk of Financial Institutions, ed. by Mark Carey and Rene Stulz (Cambridge, Massachusetts: National Bureau of Economic Research).
- Berger, Allen N., Asli Demirgüç-Kunt, Ross Levine, and Joseph G. Haubrich, 2004,  “Bank Concentration and Competition: An Evolution in the Making,” Journal of Money, Credit and Banking, Vol. 36, No. 3, pp. 434–450.
- Berger, Allen N., Robert DeYoung, Hesna Genay, and Gregory F. Udell, 2000,  “The Globalization of Financial Institutions: Evidence from Cross-Border Banking Performance,” Brookings –Wharton Papers on Financial Services, Vol. 3, pp. 23–158.
- Berger, Allen N., 1995,  “The Profit-Structure Relationship in Banking–Tests of Market Power and Efficient-Structure Hypotheses,” Journal of Money, Credit and Banking, Vol. 27, pp. 404–31.
- Berger, Allen N., and Timothy H. Hannan, 1989,  “The Price-Concentration Relationship in Banking,” Review of Economics and Statistics, Vol. 71, pp. 291–9.
- Boyd, John H., Gianni de Nicolo, and Bruce D. Smith, 2004,  “Crises in Competitive Versus Monopolistic Banking Systems,” Journal of Money, Credit and Banking, Vol. 36 (3), pp. 487–506.
- Boyd, John H. and Stanley L. Graham, 1991,  “Investigating the Banking Consolidation Trend,” Federal Reserve Bank of Minneapolis Quarterly Review, Spring, pp. 1–15.
- Boyd, John H. and Stanley L. Graham, 1996,  “Consolidation in U.S. Banking: Implications for Efficiency and Risk,” Federal Reserve Bank of Minneapolis Working Paper, No. 572 (Minneapolis).
- De Nicoló, Gianni, Phillip Bartholomew, Jahanara Zaman and Mary Zephirin, 2004,  “Bank Consolidation, Internationalization, and Conglomerization: Trends and Implications for Financial Risk,” Financial Markets, Institutions and Instruments, Vol. 13, No. 4, pp. 173–217.
- De Nicoló, Gianni, and Myron L. Kwast, 2002,  “Systemic Risk and Financial Consolidation: Are They Related?” Journal of Banking and Finance, Vol. 26, pp. 861–80.
- Mishkin, Frederic S., 1999,  “Financial Consolidation: Dangers and Opportunities,” Journal of Banking and Finance, Vol. 23, pp. 675–691.
- Paroush, Jacob, 1995,  “The Effects of Mergers and Acquisition Activity on the Safety and Soundness of a Banking System,” Review of Industrial Organization, Vol. 10, pp. 53–67.
- Benston, George W., William C. Hunter, and Larry D. Wall, 1995,  “Motivations for Bank Mergers and Aquisitions: Enhancing the Deposit Insurance Put Option versus Earnings Diversification,” Journal of Money, Credit, and Banking, Vol. 27 (3), pp. 777–788.
- Craig, Ben and João Cabral dos Santos, 1997,  “The Risk Effects of Bank Acquisitions,” Federal Reserve Bank of Cleveland Economic Review, Vol. 33, pp. 25–35
- Claessens, Stijn and Luc Laeven, 2004, “What Drives Bank Competition? Some International Evidence,” Journal of Money, Credit and Banking, Vol. 36, No. 3, pp. 563–83.
- Shaffer, Sherrill, 2004,  “Comment on ‘What Drives Bank Competition? Some International Evidence’ by Stijn Claessens and Luc Laeven,” Journal of Money, Credit, and Banking, Vol. 36, No. 3, pp. 585–92.

### Regulation, supervision, deposit insurance, moral hazard, and prudential policy
- Barth, James R., Gerard Caprio, Jr., and Ross Levine, 2001,  “The Regulation and Supervision of Banks Around the World - A New Database (Vol.1)” (Washington: World Bank), available at http://web.worldbank.org/servlets/ECR?entityID=000094946_01052204005474&collection=IMAGEBANK&sitePK=469233.
- Barth, James R., Gerard Caprio, Jr., and Ross Levine, 2005,  Rethinking Bank Supervision and Regulation: Until Angels Govern (Cambridge, Massachusetts: Cambridge University Press).
- Barth, James R., Gerard Caprio, Jr., and Ross Levine, 2004,  “Bank Regulation and Supervision: What Works Best?” Journal of Financial Intermediation, Vol. 13, pp. 205–48.
- Hellman, Thomas F., Kevin Murdoch, and Joseph E. Stiglitz, 2000,  “Liberalization, Moral Hazard in Banking and Prudential Regulation: Are Capital Requirements Enough?” American Economic Review, Vol. 90, No. 1, pp. 147–65.
- Keeley, Michael C., 1990,  “Deposit Insurance, Risk and Market Power in Banking,” American Economic Review, Vol. 80, pp. 1183–200.
- Cordella, Tito and Eduardo L. Yeyati, 2002,  “Financial Opening, Deposit Insurance and Risk in a Model of Banking Competition,” European Economic Review, Vol. 46, No. 3, pp. 471–85.
- Podpiera, Richard, 2004,  “Does Compliance with Basel Core Principles Bring Any Measurable Benefits?” IMF Working Paper 04/204 (Washington: International Monetary Fund).
- Drees, Burkhard and Ceyla Pazarbasioglu (1998) The Nordic Banking Crises: Pitfalls in Financial Liberalization? IMF Occasional Paper 161/1998, Washington, D. C.: International Monetary Fund

### Bank lending, relationship banking, and community banking
- Berger, Allen N., and Gregory F. Udell, 2002,  “Small Business Credit Availability and Relationship Lending: The Importance of Bank Organizational Structure,” Economic Journal, Vol. 112, pp. F32–F53.
- Petersen, Mitchell A., and Raghuram G. Rajan, 1995,  “The Effect of Credit Market Competition on Lending Relationships,” Quarterly Journal of Economics, Vol. 110, pp. 407–443.
- DeYoung, Robert, William C. Hunter, and Gregory F. Udell, 2004,  “The Past, Present and Probable Future for Community Banks,” Journal of Financial Services Research, Vol. 25, pp. 85–133.
- Berger, Allen N., Iftekhar Hasan, and Leora F. Klapper, 2004,  “Further Evidence on the Link between Finance and Growth: An International Analysis of Community Banking and Economic Performance,” Journal of Financial Services Research, Vol. 25, pp. 169–202.
- Petersen, Torben, 1986,  “Fitting Parametric Survival Models with Time-Dependent Covariates,” Applied Statistics, Vol. 35, No. 3, pp. 281–8.

### Theoretical models of banking, risk-taking, and fragility
- Besanko, David and Anjan V. Thakor, 1993,  “Relationship Banking, Deposit Insurance, and Bank Portfolio,” in Capital Markets and Financial Intermediation, ed. by Colin Mayer and Xavier Vives  (Cambridge, Massachusetts: Cambridge University Press).
- Matutes, Carmen and Xavier Vives, 1996,  “Competition for Deposits, Fragility and Insurance,” Journal of Financial Intermediation, Vol. 5, No. 2, pp. 184–216.
- Matutes, Carmen and Xavier Vives, 2000,  “Imperfect Competition, Risk Taking and Regulation in Banking,” European Economic Review, Vol. 44, No. 1, pp. 1–34.
- Perotti, Enrico C., and Javier Suarez, 2002,  “Last Bank Standing: What Do I Gain If You Fail?” European Economic Review, Vol. 46, No. 9, pp. 1599–622, available at http://ideas.repec.org/s/eee/eecrev.html
- Boyd, John H. and Gianni de Nicolo, forthcoming,  “The Theory of Bank Risk-Taking and Competition Revisited,” Journal of Finance.
- Keeley, Michael C., 1990,  “Deposit Insurance, Risk and Market Power in Banking,” American Economic Review, Vol. 80, pp. 1183–200.

### Empirical methods, datasets, and measurement in banking research
- Beck, Thorsten, Asli Demirgüç-Kunt and Ross Levine, 2000, “A New Database on Financial Development and Structure,” World Bank Economic Review, Vol. 14, pp. 597–605.
- Beck, Thorsten, Asli Demirgüç-Kunt and Ross Levine, 2003,  “Law, Endowments and Finance,” Journal of Financial Economics, Vol. 70, pp. 137–81.
- Cleves, Mario A, William W. Gould, and Roberto A. Gutierrez, 2004,  An Introduction to Survival Analysis Using STATA, (College Station, Texas: Stata Press).
- Lane, William R., Stephen W. Looney, and James W. Wansley, 1986,  “An Application of the Cox Proportional Hazards Model to Bank Failure,” Journal of Banking and Finance, Vol. 10, pp. 511–531.
- Whalen, Gary, 1991,  “A Proportional Hazards Model of Bank Failure: An Examination of its Usefulness as an Early Warning Tool”. Federal Reserve Bank of Cleveland Economic Review, pp. 21–39.
- Petersen, Torben, 1986,  “Fitting Parametric Survival Models with Time-Dependent Covariates,” Applied Statistics, Vol. 35, No. 3, pp. 281–8.
- Podpiera, Richard, 2004,  “Does Compliance with Basel Core Principles Bring Any Measurable Benefits?” IMF Working Paper 04/204 (Washington: International Monetary Fund).

### Country and regional studies, comparative banking systems
- Bordo, Michael, Angela Redish, and Hugh Rockoff, 1995,  “A Comparison of the United States and Canadian Banking Systems in the Twentieth Century: Stability vs. Efficiency,” in Anglo-American Financial Systems: Institutions and Markets in the Twentieth Century, ed. by Michael Bordo and Richard Sylla (New York: Irvin).
- Capie, Forrest, 1995,  “Prudent and Stable (but inefficient?): Commercial Banks in Britain, 1890–1940,” in Anglo-American Financial Systems: Institutions and Markets in the Twentieth Century, ed. by Michael Bordo and Richard Sylla (New York: Irvin).
- Staikouras, Christos, and Geoffrey E. Wood, 2000,  ”Competition and Banking Stability in Greece and Spain,” Journal of International Banking Regulation, Vol. 2, No. 1.
- Hoggarth, Glenn A., Alistair Milne, and Geoffrey E. Wood, 1998,  “Alternative Routes to Banking Stability: A Comparison of UK and German Banking Systems,” Financial Stability Review, Vol. 5, pp. 55–68.

*References list as provided in the source PDF _wp06143 - REFERENCES.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2006/_wp06143.pdf_