## _wp0702

## Source details

**Canonical URL:** [_wp0702](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp0702.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp0702.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp0702.pdf.json)

---

### I. Motivation and literature overview — key findings and context
- Cooperative (mutual) banks:
  - Market share (total banking sector assets) in the paper’s sample increased from about 9 percent in the mid-1990s to about 14 percent in 2004.
  - Account on average for about 10 percent of banking system assets in advanced economies and emerging markets; up to 30 percent in some countries.
  - In retail business, market share is even more substantial (five EU member countries have more than a 40 percent market share of cooperative banks in terms of branch networks).
- Research gap and relevance:
  - Only about 0.1 percent of all banking-related entries in EconLit relate to cooperative banking (search performed June 15, 2006).
  - Cooperative banks receive disproportionately little attention given their market share.
- Potential implications for financial stability:
  - Cooperative banks’ objective is not to maximize profits but members’ consumer surplus; this can lead to lower profitability and potential balance sheet risks relative to capital.
  - Disclosure practices are substantially below commercial banks, reducing the effectiveness of the Basel II third pillar for cooperatives.
  - Cooperative banks rely less on shareholder pressure, interbank markets, and debt issuance; loyal and insured retail depositors may not provide timely market discipline.
- Conflicting prior evidence:
  - Some evidence of greater vulnerability (e.g., Swedish cooperative sector in early 1990s); other studies suggest mutual institutions adopt less risky strategies.
- Research questions:
  - Are cooperative banks relatively weaker in responding to stress because of their business model?
  - Does the presence of cooperative banks reduce the stability of other banks?

### II. Data — coverage, sample and limitations
- Data source and sample:
  - BankScope database (Bureau van Dijk).
  - Countries: 29 major advanced economies and emerging markets that are OECD members (see Appendix I in source for list).
  - Total observations: 16,577 banks (1994–2004), comprising 11,090 commercial banks, 3,072 cooperative banks, and 2,415 savings banks.
- Sample details:
  - Large banks defined as assets larger than US$1 billion; small banks defined as assets smaller than US$1 billion.
  - Coop5 countries (cooperatives’ market share exceeds 5 percent in the sample): Austria, France, Germany, Italy, Japan, Netherlands and the United Kingdom.
  - Slovak Republic excluded because BankScope contains no data on cooperative banks.
- Coverage and caveats:
  - BankScope coverage varies by country; for most countries in the sample, BankScope data cover 80 to 90 percent of total banking system assets.
  - Coverage of cooperative banks is lower than for commercial banks (notably few cooperative banks included in the United States).
  - Sample primarily based on unconsolidated bank statements (about 90 percent of BankScope observations for the selected countries and periods are unconsolidated). Consolidated-data calculations produced very similar results.
  - BankScope classifies specialization in the current year; historical status changes were adjusted where information was available.
- Governance index:
  - Single governance index constructed by averaging six governance measures from Kaufmann, Kraay, and Mastruzzi (2005) across years 2004, 2002, 2000, 1998, and 1996.

### III. Measuring bank stability — z-score definition and properties
- Primary dependent variable:
  - z-score defined as z ≡ (k + μ) / σ, where k is equity capital as percent of assets, μ is average after-tax return as percent on assets, and σ is standard deviation of the after-tax return on assets.
  - Interpretation: z-score measures the number of standard deviations a return must fall to deplete equity assuming normality; higher z-score ⇒ lower upper bound of insolvency risk ⇒ lower probability of insolvency.
- Applicability to cooperative banks:
  - z-score is an objective insolvency-risk measure across bank types; lower risk-adjusted returns can be offset by higher capitalization to produce similar or higher z-scores.
  - Market-price data often unavailable for cooperative banks; accounting-data-based z-score used.

### IV. Methodology — empirical approach and model specification
- Decomposition and correlations:
  - Decomposition of observed z-score differences into components: capitalization (k), returns (μ), and return volatility (σ).
  - Calculation of correlation coefficients between z-scores and other variables.
- Main empirical approach:
  - Panel regressions of z-scores on bank-specific variables B_{i,j,t-1}, industry-specific variables I_{j,t-1}, bank-type dummies and interactions T_s, macroeconomic vector M_{j,t}, country dummies C_j, year dummies D_t, and residual ε_{j,i,t}.
  - Two dummies: commercial = 1 if commercial bank (0 otherwise); savings = 1 if savings bank (0 otherwise). Cooperative banks are the omitted category.
- Testing systemic impact:
  - Cooperative banks’ market share by assets (calculated annually and by country) is interacted with the commercial bank dummy to test whether cooperative presence affects commercial banks’ stability.
  - Example interpretation: negative sum of coefficients of cooperative market share and its interaction with commercial dummy ⇒ decrease in commercial banks’ stability.
- Control variables:
  - Bank-level: asset size in billions of U.S. dollars, loans over assets, cost-income ratio.
  - Income diversity: (Total operating income − Net interest income) / Total operating income − 1 (higher values → higher diversification away from net interest income); interacted with cooperative dummy.
  - Country-level: GDP growth rate, inflation, the real long-term interest rate, exchange rate appreciation, Herfindahl index (sum of squared market shares in assets).
- Timing and robustness:
  - Bank-specific and macroeconomic variables, Herfindahl index, cooperative market share and its interaction with commercial dummy are lagged; robustness checks use contemporaneous specifications.
  - Outliers: z-scores mostly in the 20–80 range; sample range from -81 to 14,811 with an average of 57. Main results exclude 1st and 99th percentiles of z-score distribution.

### V. Decomposition of z-scores — main descriptive findings (selected statistics)
- Aggregate (full sample):
  - Z-score (All banks): Commercial 46.51; Cooperative 56.96; Savings 55.47.
  - Equity to Assets (percent): Commercial 11.21; Cooperative 6.84; Savings 7.99.
  - ROA (percent): Commercial 0.90; Cooperative 0.37; Savings 0.53.
  - Standard deviation of ROA (percent points): Commercial 0.65; Cooperative 0.31; Savings 0.35.
- Large banks:
  - Z-score: Commercial 29.67; Cooperative 46.65; Savings 47.35.
  - Standard deviation of ROA (percent points): Commercial 0.71; Cooperative 0.37; Savings 0.35.
- Small banks:
  - Z-score: Commercial 50.01; Cooperative 60.87; Savings 60.19.
  - Standard deviation of ROA (percent points): Commercial 0.59; Cooperative 0.28; Savings 0.35.
- Country examples (Table 3):
  - Germany: Commercial 37.31; Cooperative 78.8.
  - Netherlands: Commercial 44.4; Cooperative 82.2.
  - United Kingdom: Commercial 28.3; Cooperative 70.9.
- Volatility decomposition (Table 4):
  - ROA downward volatility (percentage points): Commercial -0.46; Cooperative -0.19; Savings -0.21.
  - ROA upward volatility (percentage points): Commercial 0.38; Cooperative 0.20; Savings 0.21.
  - Equity to assets downward volatility (percentage points): Commercial -1.53; Cooperative -0.53; Savings -0.78.
- Distribution of z-scores by bank type:
  - Less than 0: Commercial 0.37; Cooperative 0.62; Savings 0.13.
  - 0 to 10: Commercial 13.65; Cooperative 59.20; Savings 6.38.
  - More than 30: Commercial 57.52; Cooperative 66.42; Savings 68.84.

### VI. Core empirical results from panel regressions
- Main cross-cutting result:
  - Cooperative banks have higher z-scores than commercial and savings banks across pooled OLS and FE specifications (cooperative bank dummy positive and significant in pooled OLS; commercial and savings dummies negative and significant in most models).
- Market share (system) effects:
  - Estimated slope on “share of cooperatives” is positive and significant in most specifications: a higher share of cooperative banks increases stability (z-score) of an average bank in the same banking system.
  - Interaction of share of cooperatives with commercial bank dummy is generally negative and in some specifications significant — higher cooperative share can reduce commercial banks’ stability, particularly for weaker commercial banks (evidence strongest in quantile regressions).
- Subsamples by bank size:
  - Large banks: a strong cooperative sector on average does not weaken commercial large banks; share of cooperatives positive and significant for large-bank regressions in many specifications (e.g., share of cooperatives (-1) = 0.519 in OECD OLS, p-value (0.000)*** reported in source).
  - Small banks: small commercial banks tend to be riskier than small cooperative banks; no consistent significant negative effect of higher cooperative share on small commercial banks’ individual risk.
- Quantile regressions:
  - The cooperative z-score advantage widens at higher quantiles: cooperatives’ advantage larger among strong banks than among weak banks.
  - Increased presence of cooperative banks has a negative effect on the weakest commercial banks (25th percentile).
- Other explanatory variables:
  - Bank size: larger banks tend to have lower z-scores (engage in riskier activities).
  - Loans/Assets: higher loan-to-asset ratios associated with higher risk in several specifications (negative and significant coefficients reported).
  - Cost-Income Ratio: higher cost-to-income ratio associated with lower z-scores (negative and significant coefficients).
  - Income diversity: increased income diversity tends to increase banks’ risk overall; however, cooperative banks tend to become more stable with diversification (sum of income diversity coefficient and its interaction with cooperative dummy is positive for cooperatives).
  - Concentration (Herfindahl Index): mixed and unclear effects across specifications.
  - Governance: higher governance index associated with higher z-scores (Governance = 21.577, p-value (0.000)*** in OECD OLS reported in source).

### VII. Robustness checks and alternative measures
- Robust estimation (Table 11): down-weighting outliers supports main conclusions about cooperative banks’ higher z-scores and the sign pattern for share effects across full sample, EU12, and Coop5.
- Alternative z-score definitions: tested (including modified z-score using absolute value of downward volatility; ln(1+(z/100)) transformation) — main conclusions unchanged.
- Ratings (Table 5): Fitch long-term ratings show about 90 percent of institutions have investment grade (BBB- or higher), but ratings distribution dominated by German cooperative banks (all 637 German cooperative banks rated A+), limiting cross-country usefulness.
- Correlation analysis (Table 6): cooperative bank dummy positively correlated with z-score; Share Coop * Commercial interaction negatively correlated with z-score and significant at 1 percent in reported correlations.
- Estimation variants (pooled contemporaneous-only variables, weighted regressions by total assets, exclusion of 1st and 99th percentiles): findings remain robust.

### VIII. Interpretations, policy-relevant implications and caveats
- Drivers of cooperative banks’ higher z-scores:
  - Higher z-scores of cooperative banks are driven mainly by much lower volatility of returns rather than higher capitalization or profitability. Cooperative banks have lower average ROA and lower equity-to-assets ratios but much lower ROA volatility which elevates z-scores.
  - Mechanism: cooperatives pass on returns to customers in normal times (consumer surplus) and can recoup some surplus in weak times; mutual support networks may contribute somewhat.
  - Replenishing capital via consumer surplus is indirect and requires time compared with undistributed profits.
- System-wide effects:
  - A higher share of cooperative banks increases the stability (z-score) of the average bank in a system.
  - However, a stronger cooperative sector can weaken commercial banks, particularly those already weak (consistent with Goodhart’s hypothesis on crowding out retail business and inducing risk-shifting by commercial banks).
- Caveats:
  - z-score uses accounting data and emphasizes capital and profits; liquidity and asset-quality dimensions are not directly captured.
  - Ratings data limited by country concentration (Germany).
  - Analysis focuses on individual-bank z-scores (average impacts per bank); systemic assessment requires losses given default and correlation of losses across defaults (beyond scope).
  - Corporate governance in cooperatives and network effects require further study.
  - Share of cooperatives treated as exogenous; endogeneity could be tested with longer time series.

### IX. Topics for further research (identified in the paper)
- Study corporate governance at the institution level for cooperatives: owner-less endowment issues, collective action problems, potential for empire-building by management.
- Analyze the impact of cooperative bank networks on stability, including adverse side-effects of integration and lateral contracts.
- Move beyond individual-bank z-scores to systemic measures: losses given default and correlation of losses across defaults.
- Test whether the share of cooperative banks is endogenous with respect to banking system stability with longer time series.

*Content based solely on the supplied PDF content unit.*

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

### References

### Tables and Figures listed in the content unit
- Tables:
  - 1. Summary Statistics of Bank-Specific Variables in the Sample, 1994–2004
  - 2. Decomposition of Z-Scores for the Full Sample, 1994–2004
  - 3. Decomposition of Z-Scores for Selected Countries, 1994–2004
  - 4. Sensitivity of the Z-score Decomposition
  - 5. Fitch: Long-Term Ratings: Distribution of the Banks in Sample
  - 6. Correlation Coefficients between the Z-Score and Selected Key Variables, 1994–2004
  - 7. Regression Results (Full Sample)
  - 8. OECD Regressions with Governance Variable
  - 9. Regression Results (Large Banks)
  - 10. Regression Results (Small Banks)
  - 11. Robust Regressions
  - 12. Quantile Regressions (Full Sample)
- Figure:
  - 1. Cooperative Banks: Retail Market Shares in Selected Countries

### I. Motivation and literature overview — key findings and context
- Cooperative (mutual) banks:
  - Market share (total banking sector assets) in the paper’s sample increased from about 9 percent in the mid-1990s to about 14 percent in 2004.
  - Account on average for about 10 percent of banking system assets in advanced economies and emerging markets, reaching as much as 30 percent in some countries in terms of assets.
  - In retail business, market share is even more substantial (e.g., five EU member countries have more than a 40 percent market share of cooperative banks in terms of branch networks).
- Sample and research gap:
  - Only about 0.1 percent of all banking-related entries in EconLit relate to cooperative banking (search performed June 15, 2006).
  - Cooperative banks receive disproportionately little attention in the empirical literature relative to their market share.
- Potential implications for financial stability:
  - Cooperative banks’ objective is not to maximize profits but members’ consumer surplus; this can result in lower profitability that may cause balance sheet risks to grow faster than capital, potentially deteriorating solvency.
  - Cooperative banks’ disclosure practices and requirements are substantially below those of commercial banks; the third pillar of Basel II (market discipline via disclosure) has significantly reduced effectiveness in the case of cooperative banks.
  - Cooperative banks rely less on shareholder pressure, interbank markets, and debt issuance; loyal and insured retail depositors are unlikely to exert effective market disciplining pressure early enough.
- Conflicting prior evidence:
  - Some evidence suggests cooperative banks may be more vulnerable to shocks (e.g., Swedish cooperative sector did not survive the early 1990s crisis in cooperative form).
  - Other studies suggest mutual financial institutions tend to adopt less risky strategies than demutualized ones.
- Research questions addressed:
  - Are cooperative banks relatively weaker in responding to stress because of their business model?
  - Does the presence of cooperative banks reduce the stability of other banks?

### II. Data — coverage and limitations
- Data source and sample:
  - BankScope database (Bureau van Dijk).
  - Countries: 29 major advanced economies and emerging markets that are members of the OECD (see Appendix I for country list).
  - Total observations: 16,577 banks from 1994 to 2004, comprising:
    - 11,090 commercial banks
    - 3,072 cooperative banks
    - 2,415 savings banks
- Coverage and caveats:
  - BankScope coverage varies by country; for most countries in the sample, BankScope data cover 80 to 90 percent of total banking system assets.
  - Coverage of cooperative banks is lower than for commercial banks (in particular, only a small number of cooperative banks is included in the United States).
  - Sample is primarily based on unconsolidated bank statements (about 90 percent of BankScope observations for the selected countries and periods are unconsolidated). Consolidated-data calculations produced very similar results (available upon request).
  - BankScope classifies specialization (status) in the current year; some banks may have changed status historically (examples given: France reform June 1999; Alliance & Leicester demutualized 1997; First National (Ireland) demutualized 1998). Where information was available, status adjustments were made.
- Additional data:
  - Macroeconomic and system-wide indicators used (described in Appendix I).

### II. Measuring bank stability — z-score definition and properties
- Primary dependent variable:
  - z-score defined as z ≡ (k + μ) / σ, where:
    - k is equity capital as percent of assets
    - μ is average after-tax return as percent on assets
    - σ is standard deviation of the after-tax return on assets (proxy for return volatility)
  - Interpretation: z-score measures the number of standard deviations a return must fall to deplete equity under the assumption of normality of bank returns. Higher z-score ⇒ lower upper bound of insolvency risk ⇒ lower probability of insolvency.
- Considerations for cooperative banks:
  - Despite lower focus on returns, z-score is an objective measure of insolvency risk applicable across bank types; if an institution accepts lower risk-adjusted returns, it can still have the same or higher z-score with higher capitalization.
  - For most cooperative banks, market price data are not available; hence the accounting-data-based z-score specification is used.

### II. Methodology — empirical approach and model specification
- Preliminary steps:
  - Decomposition of observed differences in z-scores into underlying factors: capitalization, returns, and volatility of returns.
  - Calculation of correlation coefficients between z-scores and other variables of interest.
- Main empirical approach:
  - Panel regressions of z-scores on bank-specific variables, industry-specific variables, macroeconomic variables, country dummies, and year dummies.
  - General model structure (variables described in text):
    - Dependent variable: z_{j,i,t}
    - Bank-specific vector: B_{i,j,t-1}
    - Industry-specific vector: I_{j,t-1}
    - Bank type dummies and interactions: T_s, I_{j,t-1} × T_s, B_{i,j,t-1} × T_s
    - Macroeconomic vector: M_{j,t}
    - Country dummies: C_j
    - Year dummies: D_t
    - Residual: ε_{j,i,t}
  - Two dummy variables included to distinguish bank type:
    - First dummy = 1 if commercial bank, 0 otherwise
    - Second dummy = 1 if savings bank, 0 otherwise
    - Cooperative banks implicitly the omitted category; a positive coefficient on a dummy indicates that type is relatively stronger (higher z-score) than cooperative banks.
- Testing systemic impact:
  - Cooperative banks’ market share by assets (calculated for each year and country) is interacted with the commercial bank dummy to test whether cooperative banks’ presence affects commercial banks’ stability.
  - Interpretation example: negative sign of sum of coefficients of cooperative market share and its interaction with the commercial dummy indicates a decrease in commercial banks’ stability (z-scores).
- Control variables (bank-level and country-level):
  - Bank-level controls: asset size in billions of U.S. dollars, loans over assets, cost-income ratio.
  - Income diversity measure (following Laeven and Levine): (Total operating income − Net interest income) / Total operating income − 1 (as defined in the source); higher values → higher diversification away from net interest income.
  - Income diversity is interacted with the cooperative bank dummy to capture business orientation differences.
  - Country-level controls: GDP growth rate, inflation, the real long-term interest rate, exchange rate appreciation, Herfindahl index (sum of squared market shares in terms of total assets).
  - Governance quality: single index constructed by averaging six governance measures (voice and accountability, political stability, government effectiveness, regulatory quality, rule of law, control of corruption) from Kaufmann, Kraay, and Mastruzzi (2005) across years 2004, 2002, 2000, 1998, and 1996.
- Timing and robustness:
  - All bank-specific and macroeconomic variables, the Herfindahl index, the cooperatives’ market share and its interaction with the commercial bank dummy are lagged to capture past effects; robustness checks restrict explanatory variables to contemporaneous effects.
  - Outliers in z-score distribution:
    - Across the whole sample, z-scores mostly in the 20–80 range; sample range from -81 to 14,811 with an average of 57.
    - Main presented results use a sample that eliminates the 1st and 99th percentile from the z-score distribution to assess robustness to outliers; full-sample results (including extreme outliers) are available from the authors.

### Key numeric values and sample statistics preserved from source
- Sample: 16,577 banks (1994–2004)
  - 11,090 commercial banks
  - 3,072 cooperative banks
  - 2,415 savings banks
- Market share of cooperative banks in sample:
  - Increased from about 9 percent in the mid-1990s to about 14 percent in 2004
  - Average about 10 percent of banking system assets in advanced economies and emerging markets; up to 30 percent in some countries
- Branch market shares: five EU member countries have more than a 40 percent market share of cooperative banks in terms of branch networks
- EconLit coverage: about 0.1 percent of all banking-related entries relate to cooperative banking (search date June 15, 2006)
- BankScope coverage: for most countries in sample, covers 80 to 90 percent of total banking system assets
- Unconsolidated data used: about 90 percent of BankScope observations for selected countries and periods are unconsolidated
- Z-score distribution in sample: range from -81 to 14,811; average 57
- Governance index years averaged: 2004, 2002, 2000, 1998, and 1996

*Content based solely on the supplied PDF content unit.*

### conclusions are the same for both approaches.

### _wp0702 - conclusions are the same for both approaches.

### Data, sample and estimation approach
- Sample covers banks in the following OECD member countries: Australia, Austria, Belgium, Canada, Czech Republic, Denmark, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Japan, Korea, Luxembourg, Mexico, the Netherlands, New Zealand, Norway, Poland, Portugal, Spain, Sweden, Switzerland, Turkey, United Kingdom, and the United States. The Slovak Republic is excluded because BankScope contains no data on cooperative banks.
- BankScope-based sample: 16,577 banks (1994–2004), comprising 11,090 commercial banks, 3,072 cooperative banks, and 2,415 savings banks.
- Large banks defined as assets larger than US$1 billion; small banks defined as assets smaller than US$1 billion.
- Coop5 countries (where cooperatives’ market share exceeds 5 percent in the sample): Austria, France, Germany, Italy, Japan, Netherlands and the United Kingdom.
- Main dependent variable: z-score defined as z ≡ (k + μ) / σ, where k is equity capital as percent of assets, μ is average return on assets (ROA), and σ is standard deviation of ROA.
- Estimation methods: pooled OLS, fixed effects (FE), robust estimation (down-weighting large residuals), and quantile regression (25th, 50th, 75th percentiles). Panel regressions include clustered standard errors (by bank), year and country dummies. The 1st and 99th percentiles of z-score distribution are excluded.

### Decomposition of z-scores — main descriptive findings
- Aggregate (full sample, Table 2):
  - Z-score (All banks): Commercial 46.51; Cooperative 56.96; Savings 55.47.
  - Equity to Assets (percent): Commercial 11.21; Cooperative 6.84; Savings 7.99.
  - ROA (percent): Commercial 0.90; Cooperative 0.37; Savings 0.53.
  - Standard deviation of ROA (percent points): Commercial 0.65; Cooperative 0.31; Savings 0.35.
- Large banks (Table 2):
  - Z-score: Commercial 29.67; Cooperative 46.65; Savings 47.35.
  - Standard deviation of ROA (percent points): Commercial 0.71; Cooperative 0.37; Savings 0.35.
- Small banks (Table 2):
  - Z-score: Commercial 50.01; Cooperative 60.87; Savings 60.19.
  - Standard deviation of ROA (percent points): Commercial 0.59; Cooperative 0.28; Savings 0.35.
- Country examples (Table 3): cooperative z-scores markedly higher in some countries (e.g., Germany: Commercial 37.31; Cooperative 78.8; Netherlands: Commercial 44.4; Cooperative 82.2; UK: Commercial 28.3; Cooperative 70.9).
- Downward vs upward volatility (Table 4):
  - Return on assets downward volatility (percentage points): Commercial -0.46; Cooperative -0.19; Savings -0.21.
  - Return on assets upward volatility (percentage points): Commercial 0.38; Cooperative 0.20; Savings 0.21.
  - Equity to assets downward volatility (percentage points): Commercial -1.53; Cooperative -0.53; Savings -0.78.
- Distribution of z-scores by bank type (Table 4, % of observations in same-type banks):
  - Less than 0: Commercial 0.37; Cooperative 0.62; Savings 0.13.
  - 0 to 10: Commercial 13.65; Cooperative 59.20; Savings 6.38.
  - More than 30: Commercial 57.52; Cooperative 66.42; Savings 68.84.

### Core empirical results from panel regressions
- Main cross-cutting result: Cooperative banks have higher z-scores than commercial and savings banks across pooled OLS and FE specifications (cooperative bank dummy positive and significant in pooled OLS; commercial and savings dummies negative and significant in most models).
- Market share effects:
  - The estimated slope on the “share of cooperatives” variable is positive and significant in most specifications: a higher share of cooperative banks increases stability (z-score) of an average bank in the same banking system (average effect across commercial, cooperative, and savings banks).
  - Interaction effect: the interaction of share of cooperatives with the commercial bank dummy is generally negative and in some specifications significant — implying that a higher cooperative share can reduce commercial banks’ stability. Evidence is strongest for weaker commercial banks (see quantile regressions).
- Subsamples by bank size:
  - Large banks (Table 9): results strongly support that a strong cooperative sector on average does not weaken commercial banks; pooled OLS OECD exception noted. Share of cooperatives positive and significant for large-bank regressions in many specifications (e.g., share of cooperatives (-1) = 0.519 in OECD OLS, p-value (0.000)***).
  - Small banks (Table 10): small commercial banks tend to be riskier than small cooperative banks; no consistent significant negative effect of higher cooperative share on small commercial banks’ individual risk.
- Quantile regressions (Table 12):
  - The gap between z-scores of commercial and cooperative banks widens at higher quantiles: cooperative z-score advantage is larger among strong banks than among weak banks.
  - An increased presence of cooperative banks has a negative effect on the weakest commercial banks (25th percentile) — weaker commercial banks suffer more from a stronger cooperative sector.
- Other explanatory variables:
  - Bank size: larger banks tend to have lower z-scores (engage in riskier activities).
  - Loans/Assets: higher loan-to-asset ratios associated with higher risk in several specifications (e.g., Loans/Assets (-1) coefficients negative and significant in many OECD regressions).
  - Cost-Income Ratio: higher cost-to-income ratio associated with lower z-scores (negative and significant coefficients in many regressions).
  - Income diversity: increased income diversity tends to increase banks’ risk overall; however, cooperative banks tend to become more stable with diversification (sum of income diversity coefficient and its interaction with cooperative dummy is positive for cooperatives).
  - Concentration (Herfindahl Index): mixed and unclear effects on individual bank risk across specifications.
  - Governance (Table 8): higher governance index (Kaufmann, Kraay, Mastruzzi) associated with higher z-scores (Governance coefficient positive and significant in OECD OLS: Governance = 21.577, p-value (0.000)***).

### Robustness checks and alternative measures
- Robust estimation (Table 11): assigning lower weights to outliers supports main conclusions about cooperative banks’ higher z-scores and the sign pattern for share effects across full sample, EU12, and Coop5.
- Alternative z-score definitions tested (including modified z-score defined as capitalization plus ROA over absolute value of downward volatility of ROA; alternative transformations ln(1+(z/100))) — main conclusions unchanged.
- Ratings (Table 5): Fitch long-term ratings distribution shows no major difference between cooperative and commercial banks at first glance (about 90 percent of institutions have investment grade long-term credit rating (BBB- or higher)), but the rating distribution is dominated by German cooperative banks (all 637 German cooperative banks rated A+), limiting cross-country usefulness.
- Correlation analysis (Table 6): cooperative bank dummy positively correlated with z-score; interaction term Share Coop * Commercial negatively correlated with z-score across samples (Full sample, Large banks, Small banks) and significant at 1 percent in reported correlations.
- Estimation variants: pooled contemporaneous-only explanatory variables, weighted regressions (by total assets), alternative z-score specifications, exclusion of 1st and 99th percentiles — findings remain robust.

### Interpretations, policy-relevant implications and caveats
- Interpretation of cooperative banks’ higher z-scores:
  - Higher z-scores of cooperative banks are driven by much lower volatility of returns rather than higher capitalization or profitability. Cooperative banks exhibit lower average ROA and lower equity-to-assets ratios, but much lower ROA volatility which increases z-scores.
  - Proposed mechanism: cooperatives pass on returns to customers in normal times (consumer surplus) and can recoup some surplus in weak times, acting as a first line of defense; mutual support networks may contribute, though likely explain a small part of the difference.
  - Cooperative banks’ consumer surplus is hard to measure directly; replenishing capital via consumer surplus is a step removed and requires time compared with undistributed profits.
- System-wide effects:
  - A higher share of cooperative banks increases the stability (z-score) of the average bank in a system.
  - However, a stronger cooperative sector can weaken commercial banks, particularly those that are already weak (consistent with Goodhart’s hypothesis that non-profit-maximizing entities can pull down stability of other institutions by crowding out retail business or inducing risk-shifting by commercial banks).
- Caveats:
  - Z-score relies on accounting data and focuses on capital and profits rather than liquidity or asset quality.
  - Ratings data limited by country concentration (Germany) and hence not ideal for cross-country econometric analysis.
  - Analysis is based on individual-bank z-scores (average impacts per bank). A fuller systemic assessment requires looking at correlation of losses across defaults and losses given default (beyond scope of this paper).
  - Corporate governance in cooperatives and network effects are important topics not fully addressed here due to data limitations.
  - Share of cooperatives treated as exogenous; with longer time series one could test endogeneity between share of cooperatives and z-scores.

### Topics for further research (as identified in the paper)
- Study corporate governance at the institution level for cooperatives to assess owner-less endowment issues, collective action problems, and potential for empire-building by management.
- Analyze the impact of cooperative bank networks on stability, including potential adverse side-effects of integration and lateral contracts.
- Move beyond individual-bank z-scores to systemic measures: losses given default and correlation of losses across defaults (systemic loss metrics).
- Test whether the share of cooperative banks is endogenous with respect to banking system stability when longer time series become available.

*Source: _wp0702 - conclusions are the same for both approaches.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp0702.pdf_