## 6. Differences in Credit Growth in Banks with High Exposures to

## Source details

**Canonical URL:** [6. Differences in Credit Growth in Banks with High Exposures to](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08219.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08219.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08219.pdf.json)

---

### Introduction and research question
- Context: bank credit in emerging Europe has been growing rapidly—at average annual rates of 25–40 percent—during the last decade.
- Core question: Are credit expansions being led by weak banks (lower measures of bank soundness), after controlling for feedback from credit growth to bank soundness?
- Sample and period: bank-level balance sheet data for the Baltic countries (Estonia, Latvia, Lithuania) and other central and eastern European countries (the Czech Republic, Hungary, Poland, the Slovak Republic, and Slovenia) for the period 1995–2004.
- Data sources: Bankscope database (bank financial ratios), supplementary supervisory data for subsample analyses, and IMF’s International Financial Statistics for macro variables.

### Empirical model, estimation method, and key variables
- Simultaneous-equation framework:
  - BankCreditGrowth = annual percent change in real bank credit to the private sector.
  - Distance to default (DD) ≡ (k+μ)/σ, where k is equity capital as percent of assets, μ is return on average assets in percent, and σ is the standard deviation of return on average assets (DD calculated using annual balance sheet data; σ computed for the entire sample period in baseline).
  - Control variables: GDP per capita, real GDP growth, real interest rate (RIR), annual percent change in the real exchange rate (ΔRER), CostToIncome, InterestMargin (net interest margin), Liquidity ratio, Bank size, Public and Foreign ownership indicators.
  - All right-hand-side variables, except ownership measures, are lagged by one year; lagged dependent variables included for persistence.
- Estimation method:
  - Three-stage least squares (3SLS) with dynamic random effects; robustness checks include Arellano-Bond for the DD equation and alternative specifications.
- Sample coverage and descriptives:
  - Sample includes 217 commercial banks operating during 1995–2004.
  - Average number of observations per bank: around 7.
  - Bankscope covers around 80 percent of banks operating in central and eastern Europe and, on average, more than 80 percent of total assets of the respective banking systems.
  - Selected summary statistics (period/regional means and standard deviations reported in source tables).

### Main empirical findings (simultaneous estimates and interpretation)
- Persistence and dynamics:
  - Lagged bank credit growth coefficients:
    - 1995-2004: 0.096*** [5.83]
    - 1995-2000: 0.100*** [3.89]
    - 2001-2004: 0.095*** [4.71]
  - Lagged DD coefficients in DD equation:
    - 1995-2004: 0.229** [2.16]
    - 1995-2000: 0.350* [1.94]
    - 2001-2004: 0.147 [1.20]
- Impact of DD on credit growth:
  - Earlier period (1995–2000): coefficient on DD positive and statistically significant; a one standard deviation increase in distance to default added about 4 percentage points to annual bank credit growth (magnitude reported in text).
  - Later period (2001–04): coefficient on DD became insignificant — banks with lower DD started to expand at least as fast as banks with higher DD.
- Selected other coefficients (sign and significance from Table 3):
  - Real GDP growth:
    - 1995-2004: 2.646*** [5.53]
    - 1995-2000: 2.415*** [2.92]
    - 2001-2004: 2.475*** [4.38]
  - GDP per capita:
    - 1995-2004: -0.116** [1.99]
    - 1995-2000: 0.017*** [2.83]
    - 2001-2004: -0.301*** [3.19]
  - Net interest margin:
    - 1995-2004: 0.689 [1.47]
    - 1995-2000: 1.757** [2.25]
    - 2001-2004: 1.200** [2.00]
  - Public ownership:
    - 1995-2004: -0.178*** [3.73]
    - 1995-2000: -0.153** [2.39]
    - 2001-2004: -0.067 [0.89]
  - Liquidity ratio:
    - 1995-2004: 0.020*** [2.67]
    - 1995-2000: 0.013 [1.17]
    - 2001-2004: 0.027** [2.55]
  - Bank size:
    - 1995-2004: 0.311*** [3.33]
    - 1995-2000: 0.240** [2.07]
    - 2001-2004: 0.324** [2.22]
  - Foreign ownership:
    - 1995-2004: 0.008*** [2.80]
    - 1995-2000: 0.012*** [3.28]
    - 2001-2004: 0.003 [0.69]
- Robustness to alternative soundness measures:
  - Defining weak banks as bottom quintile of DD distribution (Table 4) — interaction results indicate weak banks grew faster in the later period:
    - Distance to default: 1995-2000 = 0.417** [2.09]; 2001-2004 = 0.242* [1.86]
    - Distance to default of weak banks: 1995-2000 = -1.771 [1.33]; 2001-2004 = -3.859*** [2.80]
    - Dummy for weak banks: 1995-2000 = 8.784 [1.30]; 2001-2004 = 19.683*** [3.15]
    - F-test probability: 1995-2000 = 0.32; 2001-2004 = 0.01 (F-test statistics 2.25 and 10.62*** respectively)
  - Using NPL ratio as soundness measure: NPL coefficient positive and statistically significant during 2001–04 and negative and statistically insignificant during 1995–2000.
  - Results robust to alternative DD volatility calculation, controlling for time- and country-fixed effects, adding financial/institutional/regulatory measures, alternative ownership dummies, and Arellano-Bond estimation for DD.

### Heterogeneity and pockets of vulnerabilities (subsample findings)
- Regional split — Baltics vs CEECs (Table 5):
  - Role of weaker banks in credit expansion increased over time in the Baltics.
  - In the CEECs, sounder banks expanded more rapidly during 2001–04.
  - Selected coefficients:
    - Bank credit growth (lagged):
      - 1995-2000: 0.095*** [3.64]
      - 2001-2004: 0.094*** [4.70]
    - Distance to default:
      - 1995-2000: 0.241 [1.23]
      - 2001-2004: 0.433*** [3.01]
    - Distance to default of Baltic banks:
      - 1995-2000: 0.684 [1.46]
      - 2001-2004: -0.961*** [3.72]
    - Baltic banks dummy:
      - 1995-2000: -6.839 [0.81]
      - 2001-2004: 18.209*** [2.77]
    - F-test probabilities: 1995-2000 = 0.34; 2001-2004 = 0.00 (F-test statistics 2.14 and 15.40*** respectively)
  - Interpretation: In the Baltics, weak banks became important drivers of credit expansions in 2001–04; in the CEECs, the opposite pattern.
- Ownership and parent-bank effects:
  - Weaker foreign-owned banks appear to lend more aggressively than domestically owned banks.
  - Controlling for parent-bank DD: rapid credit growth of foreign affiliates is positively correlated with DD of parent banks — suggesting access to parent funding matters.
  - Foreign-owned affiliate lending becomes uncorrelated with affiliate DD in recent years but remains linked to parent-bank DD.
  - Nordic foreign-owned affiliates: lending least related to DD among foreign banks.
- Exposure by loan currency and borrower type (Table 6):
  - Banks with high exposure to foreign-currency lending:
    - Credit growth through these banks is negatively correlated with DD in the later period — weaker banks expanding faster in foreign-currency lending segments.
    - Selected coefficients (Foreign-Currency Lending panel):
      - Bank credit growth (lagged):
        - 1995-2000: 0.157*** [4.25]
        - 2001-2004: 0.073** [2.51]
      - Distance to default:
        - 1995-2000: 0.422** [2.44]
        - 2001-2004: 0.279* [1.68]
      - Distance to default of exposed banks:
        - 1995-2000: 0.006 [0.01]
        - 2001-2004: -0.794* [1.74]
      - Exposed bank dummy:
        - 1995-2000: 23.238 [1.48]
        - 2001-2004: 29.541*** [2.80]
      - F-test probabilities: 1995-2000 = 0.01; 2001-2004 = 0.02 (F-test statistics 10.33*** and 8.11*** respectively)
  - Banks with high exposure to household lending:
    - Weaker banks expanded faster in household lending in the later period.
    - Selected coefficients (Household Lending panel):
      - Bank credit growth (lagged):
        - 1995-2000: 0.180*** [5.45]
        - 2001-2004: 0.086*** [3.80]
      - Distance to default:
        - 1995-2000: 0.613*** [3.33]
        - 2001-2004: 0.355** [2.38]
      - Distance to default of exposed banks:
        - 1995-2000: -0.886** [2.16]
        - 2001-2004: -1.889*** [2.86]
      - Exposed bank dummy:
        - 1995-2000: 50.436*** [4.35]
        - 2001-2004: 28.312*** [2.68]
      - F-test probabilities: 1995-2000 = 0.00; 2001-2004 = 0.01 (F-test statistics 23.79*** and 9.13*** respectively)
  - Data coverage note for subsamples: sample covers Czech, Estonian, Lithuanian, Polish, Slovak, and Slovenian banks (Hungary and Latvia excluded due to data limitations); exposures defined by higher-than-average proportion and higher-than-average growth in the share of loans in respective categories.

### Interpretation and policy implications
- Overall empirical conclusion:
  - In an environment of sustained credit growth, banks with lower indicators of soundness (weaker banks) began to expand at least as fast as, and in some cases faster than, sounder banks — particularly during 2001–04 and concentrated in:
    - the Baltics,
    - banks with high exposure to foreign-currency lending,
    - banks with high exposure to household lending,
    - certain foreign-owned affiliates (conditional on parent-bank strength).
- Potential pathways and caveats:
  - The absence of a statistically significant negative effect of credit growth on contemporaneous DD suggests prudential consequences of aggressive expansion by weak banks may materialize with lag.
  - NPL ratio is an imperfect soundness measure (subject to evergreening); DD computation choices (σ over full sample vs subperiods) affect absolute risk assessment but not qualitative conclusions.
  - Data limitations (country coverage and exposure measures) make some subsample results preliminary.
- Policy recommendations:
  - Maintain forward-looking and risk-based supervision during credit booms.
  - Supervisors should carefully monitor rapidly expanding banks, especially:
    - banks in high-growth markets (e.g., the Baltics),
    - banks aggressively expanding in foreign-currency lending and household lending,
    - foreign-owned affiliates with access to parent funding but lower local DD.
  - Be prepared to take measures to limit the expansion of weak banks to prevent prudential risks from undermining systemic stability and imposing costs on the broader economy and taxpayers.

### APPENDIX I. DATA SOURCES AND METHODOLOGY — key points
- Macroeconomic data: February 2006 version of the IMF’s International Financial Statistics.
- Bank-level data: February 2006 version of Bankscope, complemented with the 1995–2002 Bankscope data set provided by Ugo Panizza.
- Confidential supervisory data: bank-by-bank composition of loans from central banks of all central and eastern European countries except Latvia and Hungary.
- Bank ownership information: Bankscope (current-year shares), Micco, Panizza and Yañez (2004) historical ownership up to 2002, banks’ websites, and Bankscope data on parent banks to update ownership for 2003 and 2004.
- Data processing and cleaning highlights:
  - Used Bankscope unique identifiers; handled merged/acquired banks with reranking rules.
  - Dropped about 4 percent of observations on tails of distributions of bank-level credit growth and distance to default.
  - When both consolidated and unconsolidated balance sheets existed, unconsolidated data preserved; otherwise consolidated data used.
  - Loan composition supervisory data transformed into dummy indicators for rapidly growing household and foreign-currency portfolios to preserve confidentiality and reduce measurement error.
- Variable definitions (selected):
  - Distance to default: Return on average assets plus equity (valued at market prices) as a percent of assets divided by the standard deviation of return on average assets.
  - Net interest margin: Interest income, on a taxable equivalent basis, earned on assets less interest expense paid on liabilities and capital divided by average earning assets.
  - Cost-to-income ratio: Total operating expenses divided by total operating income.
  - Liquidity ratio: Net liquid assets divided by total deposits.
  - Bank credit growth: Annual percentage change in total loans.
  - Bank size: Logarithm of total assets.
  - Foreign ownership: Share of capital held by foreign investors.
  - Public ownership: Share of capital held by the government.
  - GDP per capita: Real GDP per capita, in hundreds of USD.
  - Real interest rate: Money market rate minus inflation; where money market rate missing, deposit rate used instead.
  - Note: Data used for all calculations are in USD, unless noted otherwise.
- Summary statistics (Appendix Table 1, selected):
  - Observations: 1,087 for most variables.
  - Bank credit growth: Observations 1,087; Mean 25.31; Standard Deviation 40.80; Minimum -86.74; Maximum 198.24
  - Distance to default: Observations 1,087; Mean 13.55; Standard Deviation 12.89; Minimum -6.27; Maximum 75.48
  - Net interest margin: Observations 1,086; Mean 4.21; Standard Deviation 2.79; Minimum -4.50; Maximum 23.61
  - Cost-to-income ratio: Observations 1,081; Mean 71.99; Standard Deviation 76.90; Minimum -959.51; Maximum 1,946.87
  - Liquidity ratio: Observations 1,077; Mean 16.74; Standard Deviation 16.61; Minimum 0.00; Maximum 98.39
  - Bank size: Observations 1,087; Mean 6.40; Standard Deviation 1.44; Minimum 2.30; Maximum 10.30
  - Real GDP growth: Observations 1,087; Mean 3.83; Standard Deviation 2.78; Minimum -2.97; Maximum 12.05
  - GDP per capita: Observations 1,087; Mean 58.52; Standard Deviation 25.16; Minimum 24.60; Maximum 147.32
  - Real interest rate: Observations 1,087; Mean 2.29; Standard Deviation 3.74; Minimum -19.52; Maximum 10.73
  - Real depreciation: Observations 1,087; Mean -0.13; Standard Deviation 0.52; Minimum -2.94; Maximum 2.59
  - Foreign ownership: Observations 1,087; Mean 41.98; Standard Deviation 45.20; Minimum 0.00; Maximum 100.00
  - Public ownership: Observations 1,087; Mean 10.50; Standard Deviation 28.09; Minimum 0.00; Maximum 100.00

*Source: IMF staff estimates and analysis in the supplied chapter content.*

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

### _wp08219 - References..............................................................................................................

### References
- References................................................................................................................................21

### Tables
- 1.  Sample Coverage ..................................................................................................................7
- 2.  Summary Statistics by Period and Region............................................................................9
- 3.  Simultaneous Modeling of Bank Credit Growth and Distance to Default .........................10
- 4.  Credit Growth in Weak Banks ............................................................................................12
- 5.  Differences in Bank Credit Growth in the Baltics and Other Central and 
        Eastern European Countries.............................................................................................14

*Source: _wp08219 - References..............................................................................................................*

### 6.  Differences in Credit Growth in Banks with High Exposures to

### _wp08219 - 6.  Differences in Credit Growth in Banks with High Exposures to

### Introduction and research question
- Context: bank credit in emerging Europe has been growing rapidly—at average annual rates of 25–40 percent—during the last decade.
- Core question: Are credit expansions being led by weak banks (lower measures of bank soundness), after controlling for feedback from credit growth to bank soundness?
- Sample and period: bank-level balance sheet data for the Baltic countries (Estonia, Latvia, Lithuania) and other central and eastern European countries (the Czech Republic, Hungary, Poland, the Slovak Republic, and Slovenia) for the period 1995–2004.
- Data sources: Bankscope database (bank financial ratios), supplementary supervisory data for subsample analyses, and IMF’s International Financial Statistics for macro variables.

### Empirical model, estimation method, and key variables
- Simultaneous-equation framework modeling:
  - BankCreditGrowth = annual percent change in real bank credit to the private sector.
  - Distance to default (DD) ≡ (k+μ)/σ, where k is equity capital as percent of assets, μ is return on average assets in percent, and σ is the standard deviation of return on average assets (DD calculated using annual balance sheet data; σ computed for the entire sample period in baseline).
  - Control variables include: GDP per capita, real GDP growth, real interest rate (RIR), annual percent change in the real exchange rate (ΔRER), CostToIncome, InterestMargin (net interest margin), Liquidity ratio, Bank size, Public and Foreign ownership indicators.
  - All right-hand-side variables, except ownership measures, are lagged by one year; lagged dependent variables included for persistence.
- Estimation method:
  - Three-stage least squares (3SLS) used to estimate the simultaneous system with dynamic random effects; robustness checks include Arellano-Bond for the DD equation and alternative specifications.
- Sample coverage and descriptives:
  - Sample includes 217 commercial banks operating during 1995–2004.
  - Average number of observations per bank: around 7.
  - Bankscope covers around 80 percent of banks operating in central and eastern Europe and, on average, more than 80 percent of total assets of the respective banking systems.
  - Summary statistics (selected):
    - Bank credit growth mean and standard deviation:
      - CEECs 1995-2000: Mean 17.94, Standard Deviation 40.12
      - CEECs 2001-2004: Mean 27.33, Standard Deviation 28.75
      - Baltics 1995-2000: Mean 6.64, Standard Deviation 46.84
      - Baltics 2001-2004: Mean 3.8 (table aligns columns differently in source; full table retained in source)

### Main empirical findings (simultaneous estimates and interpretation)
- Baseline simultaneous-model results (selected coefficients from Table 3; absolute value of z statistics in brackets; significance: * 10 percent; ** 5 percent; *** 1 percent):
  - Bank credit growth equation (coefficient on lagged bank credit growth):
    - 1995-2004: 0.096*** [5.83]
    - 1995-2000: 0.100*** [3.89]
    - 2001-2004: 0.095*** [4.71]
  - Distance to default equation (coefficient on lagged DD):
    - 1995-2004: 0.229** [2.16]
    - 1995-2000: 0.350* [1.94]
    - 2001-2004: 0.147 [1.20]
  - Impact of DD on credit growth:
    - Earlier period (1995–2000): coefficient on DD positive and statistically significant.
    - Later period (2001–04): coefficient on DD became insignificant — banks with lower DD started to expand at least as fast as banks with higher DD.
  - Magnitude interpretation reported in text: a one standard deviation increase in distance to default added about 4 percentage points to annual bank credit growth in the earlier period.
- Other notable coefficient signs and significance (Table 3, selected):
  - Real GDP growth:
    - 1995-2004: 2.646*** [5.53]
    - 1995-2000: 2.415*** [2.92]
    - 2001-2004: 2.475*** [4.38]
  - GDP per capita:
    - 1995-2004: -0.116** [1.99]
    - 1995-2000: 0.017*** [2.83]
    - 2001-2004: -0.301*** [3.19]
  - Net interest margin:
    - 1995-2004: 0.689 [1.47]
    - 1995-2000: 1.757** [2.25]
    - 2001-2004: 1.200** [2.00]
  - Public ownership (share of capital owned by the state):
    - 1995-2004: -0.178*** [3.73]
    - 1995-2000: -0.153** [2.39]
    - 2001-2004: -0.067 [0.89]
  - Liquidity ratio:
    - 1995-2004: 0.020*** [2.67]
    - 1995-2000: 0.013 [1.17]
    - 2001-2004: 0.027** [2.55]
  - Bank size:
    - 1995-2004: 0.311*** [3.33]
    - 1995-2000: 0.240** [2.07]
    - 2001-2004: 0.324** [2.22]
  - Foreign ownership:
    - 1995-2004: 0.008*** [2.80]
    - 1995-2000: 0.012*** [3.28]
    - 2001-2004: 0.003 [0.69]
- Robustness and alternative soundness measures:
  - Defining weak banks as the bottom quintile of DD distribution:
    - Interaction results (Table 4) indicate weak banks grew faster in the later period:
      - Distance to default coefficient: 1995-2000 = 0.417** [2.09]; 2001-2004 = 0.242* [1.86]
      - Distance to default of weak banks: 1995-2000 = -1.771 [1.33]; 2001-2004 = -3.859*** [2.80]
      - Dummy variable for weak banks: 1995-2000 = 8.784 [1.30]; 2001-2004 = 19.683*** [3.15]
      - F-test probability (1995-2000) 0.32; (2001-2004) 0.01 (F-test statistic reported as 2.25 and 10.62*** respectively)
  - Using the NPL ratio as a soundness measure yields consistent conclusions: the NPL ratio coefficient in the credit growth equation is positive and statistically significant during 2001–04 and negative and statistically insignificant during 1995–2000.
  - Results robust to: alternative DD volatility calculation (e.g., subperiod σ), controlling for time- and country-fixed effects, adding financial/institutional/regulatory measures, alternative ownership dummies, and Arellano-Bond estimation for DD.

### Heterogeneity and pockets of vulnerabilities (subsample findings)
- Regional split (Table 5):
  - Baltics vs CEECs:
    - Role of weaker banks in credit expansion increased over time in the Baltics.
    - In the CEECs, sounder banks expanded more rapidly during 2001–04.
    - Selected coefficients (from Table 5):
      - Bank credit growth (lagged):
        - 1995-2000: 0.095*** [3.64]
        - 2001-2004: 0.094*** [4.70]
      - Distance to default:
        - 1995-2000: 0.241 [1.23]
        - 2001-2004: 0.433*** [3.01]
      - Distance to default of Baltic banks:
        - 1995-2000: 0.684 [1.46]
        - 2001-2004: -0.961*** [3.72]
      - Baltic banks dummy:
        - 1995-2000: -6.839 [0.81]
        - 2001-2004: 18.209*** [2.77]
      - F-test probabilities: 1995-2000 = 0.34; 2001-2004 = 0.00 (F-test statistics 2.14 and 15.40*** respectively).
    - Interpretation: In the Baltics, weak banks became important drivers of credit expansions in 2001–04; in the CEECs, the opposite pattern.
- Ownership and parent-bank effects:
  - Weaker foreign-owned banks appear to lend more aggressively than domestically owned banks.
  - When controlling for the DD of parent banks, rapid credit growth of foreign affiliates is positively correlated with the DD of parent banks — suggesting access to parent funding matters.
  - Separate regressions:
    - Foreign-owned affiliate lending becomes uncorrelated with affiliate DD in recent years but remains linked to parent-bank DD.
    - Domestically owned banks: lending positively correlated with DD.
  - Nordic foreign-owned affiliates: lending least related to DD among foreign banks.
- Exposure by loan currency and borrower type (Table 6):
  - Banks with high exposure to foreign-currency lending:
    - Credit growth through these banks is negatively correlated with DD in the later period — weaker banks expanding faster in foreign-currency lending segments.
    - Selected coefficients (Foreign-Currency Lending panel, Table 6):
      - Bank credit growth (lagged):
        - 1995-2000: 0.157*** [4.25]
        - 2001-2004: 0.073** [2.51]
      - Distance to default:
        - 1995-2000: 0.422** [2.44]
        - 2001-2004: 0.279* [1.68]
      - Distance to default of exposed banks:
        - 1995-2000: 0.006 [0.01]
        - 2001-2004: -0.794* [1.74]
      - Exposed bank dummy:
        - 1995-2000: 23.238 [1.48]
        - 2001-2004: 29.541*** [2.80]
      - F-test probabilities: 1995-2000 = 0.01; 2001-2004 = 0.02 (F-test statistics 10.33*** and 8.11*** respectively).
  - Banks with high exposure to household lending:
    - Similar pattern: weaker banks expanded faster in household lending in the later period.
    - Selected coefficients (Household Lending panel, Table 6):
      - Bank credit growth (lagged):
        - 1995-2000: 0.180*** [5.45]
        - 2001-2004: 0.086*** [3.80]
      - Distance to default:
        - 1995-2000: 0.613*** [3.33]
        - 2001-2004: 0.355** [2.38]
      - Distance to default of exposed banks:
        - 1995-2000: -0.886** [2.16]
        - 2001-2004: -1.889*** [2.86]
      - Exposed bank dummy:
        - 1995-2000: 50.436*** [4.35]
        - 2001-2004: 28.312*** [2.68]
      - F-test probabilities: 1995-2000 = 0.00; 2001-2004 = 0.01 (F-test statistics 23.79*** and 9.13*** respectively).
  - Note on data coverage for these subsamples: sample covers Czech, Estonian, Lithuanian, Polish, Slovak, and Slovenian banks (Hungary and Latvia excluded due to data limitations); exposures defined by higher-than-average proportion and higher-than-average growth in the share of loans in the respective categories.

### Interpretation and policy implications
- Overall empirical conclusion:
  - In an environment of sustained credit growth, banks with lower indicators of soundness (weaker banks) began to expand at least as fast as, and in some cases faster than, sounder banks — particularly during 2001–04 and concentrated in:
    - the Baltics,
    - banks with high exposure to foreign-currency lending,
    - banks with high exposure to household lending,
    - certain foreign-owned affiliates (conditional on parent-bank strength).
- Potential pathways and caveats:
  - The absence of a statistically significant negative effect of credit growth on contemporaneous DD suggests that the prudential consequences of aggressive expansion by weak banks may materialize with lag.
  - The NPL ratio is an imperfect soundness measure (subject to evergreening), and DD computation choices (σ over full sample vs subperiods) affect the absolute risk assessment but not qualitative conclusions.
  - Data limitations (country coverage and exposure measures) make some subsample results preliminary.
- Policy recommendations emphasized by the study:
  - Importance of forward-looking and risk-based supervision during credit booms.
  - Supervisors should carefully monitor rapidly expanding banks, especially:
    - banks in high-growth markets (e.g., the Baltics),
    - banks aggressively expanding in foreign-currency lending and household lending,
    - foreign-owned affiliates with access to parent funding but lower local DD.
  - Supervisors should be prepared to take measures to limit the expansion of weak banks to prevent prudential risks from undermining systemic stability and imposing costs on the broader economy and taxpayers.

*Source: IMF staff estimates and analysis in the supplied chapter content.*

### APPENDIX I. DATA SOURCES AND METHODOLOGY

### APPENDIX I. DATA SOURCES AND METHODOLOGY

### Data sources
- Macroeconomic data: February 2006 version of the IMF’s International Financial Statistics.
- Bank-level data: February 2006 version of Bankscope, complemented with the 1995–2002 Bankscope data set provided by Ugo Panizza.
- Confidential supervisory data: Bank-by-bank composition of loans from the central banks of all central and eastern European countries except Latvia and Hungary.
- Bank ownership information: Bankscope (current-year shares), Micco, Panizza and Yañez (2004) historical ownership up to 2002, banks’ websites, and Bankscope data on parent banks to update ownership for 2003 and 2004.

### Data processing and cleaning
- Matching bank identifiers:
  - Used Bankscope unique identifiers; merged data with Ugo Panizza’s historical data using these identifiers and cross-checked with 2002 data.
  - New Bankscope identifier assigned only if a merger/acquisition intrinsically changed the bank.
- Avoiding duplications:
  - Bankscope contains consolidated and unconsolidated balance sheets with different identifiers; nonranked banks were dropped using the “rank” variable to avoid duplication.
  - If a nonranked bank had assets greater than the country average, its merger and acquisition history was examined and premerger banks were reranked while postmerger banks were deranked for the premerger period.
  - When both consolidated and unconsolidated balance sheets existed, unconsolidated data were preserved; if unconsolidated data unavailable, consolidated data were used.
- Excluding outliers:
  - About 4 percent of observations on the tails of the distributions of the two main variables (bank-level credit growth and distance to default) were dropped.
- Coding ownership:
  - Historical ownership data up to 2002 from Micco, Panizza and Yañez (2004); Bankscope current-year ownership used to extend to 2004 and complemented with banks’ websites and parent-bank information.
- Merging loan breakdowns:
  - Supervisory data covered 1995–2005 (Czech Republic: 2000–2005) and broke down total loans into:
    - (i) loans to households in local currency
    - (ii) loans to corporates in local currency
    - (iii) loans to households in foreign currency
    - (iv) loans to corporates in foreign currency
  - For confidentiality, bank identities mostly undisclosed; matching between supervisory and Bankscope data used total loans and total assets.
  - To reduce measurement error, dummy variables identifying banks with rapidly growing household and foreign currency portfolios were used instead of actual loan-level amounts.

### Variable definitions and measurement
- Distance to default: Return on average assets plus equity (valued at market prices) as a percent of assets divided by the standard deviation of return on average assets. (Bank risk / Market risk)
- Net interest margin: Interest income, on a taxable equivalent basis, earned on assets less interest expense paid on liabilities and capital divided by average earning assets. (Profitability)
- Cost-to-income ratio: Total operating expenses divided by total operating income. (Efficiency)
- Liquidity ratio: Net liquid assets divided by total deposits. (Liquidity)
- Bank credit growth: Annual percentage change in total loans.
- Bank size: Logarithm of total assets.
- Foreign ownership: Share of capital held by foreign investors.
- Public ownership: Share of capital held by the government.
- GDP per capita: Real GDP per capita, in hundreds of USD.
- Real GDP growth: Annual growth rate of real GDP.
- Real interest rate: Money market rate minus inflation; where money market rate missing, deposit rate used instead.
- Real depreciation: Annual percentage change in real exchange rate expressed in domestic currency per USD.
- Note: Data used for all calculations are in USD, unless noted otherwise.

### Summary statistics (Appendix Table 1)
- Observations: 1,087 for most variables (see individual entries).
- Bank credit growth: Observations 1,087; Mean 25.31; Standard Deviation 40.80; Minimum -86.74; Maximum 198.24
- Distance to default: Observations 1,087; Mean 13.55; Standard Deviation 12.89; Minimum -6.27; Maximum 75.48
- Net interest margin: Observations 1,086; Mean 4.21; Standard Deviation 2.79; Minimum -4.50; Maximum 23.61
- Cost-to-income ratio: Observations 1,081; Mean 71.99; Standard Deviation 76.90; Minimum -959.51; Maximum 1,946.87
- Liquidity ratio: Observations 1,077; Mean 16.74; Standard Deviation 16.61; Minimum 0.00; Maximum 98.39
- Bank size: Observations 1,087; Mean 6.40; Standard Deviation 1.44; Minimum 2.30; Maximum 10.30
- Real GDP growth: Observations 1,087; Mean 3.83; Standard Deviation 2.78; Minimum -2.97; Maximum 12.05
- GDP per capita: Observations 1,087; Mean 58.52; Standard Deviation 25.16; Minimum 24.60; Maximum 147.32
- Real interest rate: Observations 1,087; Mean 2.29; Standard Deviation 3.74; Minimum -19.52; Maximum 10.73
- Real depreciation: Observations 1,087; Mean -0.13; Standard Deviation 0.52; Minimum -2.94; Maximum 2.59
- Foreign ownership: Observations 1,087; Mean 41.98; Standard Deviation 45.20; Minimum 0.00; Maximum 100.00
- Public ownership: Observations 1,087; Mean 10.50; Standard Deviation 28.09; Minimum 0.00; Maximum 100.00
- Sources: Bankscope; and IMF staff estimates.

### Summary statistics by country (Appendix Table 2) — selected entries (observations, mean, standard deviation)
- Czech Republic:
  - Bank credit growth: 159; 25.99; 46.19
  - Distance to default: 159; 13.32; 14.49
  - Net interest margin: 159; 2.38; 1.54
  - Cost-to-income ratio: 157; 73.58; 116.09
  - Liquidity ratio: 159; 26.32; 21.15
  - Bank size: 159; 7.15; 1.26
  - Real GDP growth: 159; 2.44; 2.33
  - GDP per capita: 159; 69.28; 13.13
  - Real interest rate: 159; 1.97; 2.70
  - Real depreciation: 159; -0.14; 0.33
  - Foreign ownership: 159; 46.65; 46.23
  - Public ownership: 159; 11.09; 29.88
- Slovenia:
  - Bank credit growth: 133; 15.93; 24.95
  - Distance to default: 133; 24.27; 14.36
  - Net interest margin: 133; 3.88; 1.90
  - Cost-to-income ratio: 133; 70.87; 37.23
  - Liquidity ratio: 133; 12.71; 7.26
  - Bank size: 133; 6.18; 1.05
  - Real GDP growth: 133; 3.66; 1.51
  - GDP per capita: 133; 112.30; 14.40
  - Real interest rate: 133; 0.13; 1.95
  - Real depreciation: 133; 0.07; 0.60
  - Foreign ownership: 133; 11.76; 29.35
  - Public ownership: 133; 8.29; 25.03
- Hungary:
  - Bank credit growth: 192; 20.73; 36.11
  - Distance to default: 192; 12.15; 9.15
  - Net interest margin: 191; 4.85; 3.35
  - Cost-to-income ratio: 191; 74.61; 59.54
  - Liquidity ratio: 187; 7.55; 6.26
  - Bank size: 192; 6.68; 1.23
  - Real GDP growth: 192; 3.77; 2.01
  - GDP per capita: 192; 58.40; 15.50
  - Real interest rate: 192; 2.68; 2.28
  - Real depreciation: 192; -0.09; 0.35
  - Foreign ownership: 192; 62.21; 45.10
  - Public ownership: 192; 5.94; 22.70
- Estonia:
  - Bank credit growth: 34; 40.67; 36.64
  - Distance to default: 34; 9.95; 9.51
  - Net interest margin: 34; 4.61; 1.84
  - Cost-to-income ratio: 34; 76.94; 31.46
  - Liquidity ratio: 31; 7.50; 11.96
  - Bank size: 34; 5.82; 1.63
  - Real GDP growth: 34; 6.93; 2.32
  - GDP per capita: 34; 47.25; 14.54
  - Real interest rate: 34; -1.20; 6.15
  - Real depreciation: 34; -0.66; 1.73
  - Foreign ownership: 34; 54.53; 40.33
  - Public ownership: 34; 0.00; 0.00
- Poland:
  - Bank credit growth: 262; 25.49; 36.95
  - Distance to default: 262; 12.51; 9.47
  - Net interest margin: 262; 5.18; 3.19
  - Cost-to-income ratio: 259; 62.45; 38.64
  - Liquidity ratio: 261; 11.85; 10.40
  - Bank size: 262; 6.76; 1.49
  - Real GDP growth: 262; 3.24; 2.04
  - GDP per capita: 262; 46.96; 4.60
  - Real interest rate: 262; 6.38; 2.88
  - Real depreciation: 262; -0.02; 0.16
  - Foreign ownership: 262; 39.46; 45.04
  - Public ownership: 262; 15.62; 33.08
- Latvia:
  - Bank credit growth: 137; 36.99; 54.27
  - Distance to default: 137; 8.86; 12.76
  - Net interest margin: 137; 4.66; 2.57
  - Cost-to-income ratio: 137; 82.63; 95.15
  - Liquidity ratio: 137; 12.22; 15.26
  - Bank size: 137; 5.09; 1.18
  - Real GDP growth: 137; 6.94; 2.18
  - GDP per capita: 137; 36.40; 8.86
  - Real interest rate: 137; -0.31; 2.47
  - Real depreciation: 137; -0.14; 0.26
  - Foreign ownership: 137; 23.60; 36.12
  - Public ownership: 137; 6.40; 18.93
- Slovak Republic:
  - Bank credit growth: 119; 17.65; 35.60
  - Distance to default: 119; 12.72; 14.47
  - Net interest margin: 119; 3.04; 1.88
  - Cost-to-income ratio: 119; 69.35; 120.21
  - Liquidity ratio: 118; 38.37; 18.44
  - Bank size: 119; 6.48; 1.13
  - Real GDP growth: 119; 1.89; 2.60
  - GDP per capita: 119; 46.50; 10.44
  - Real interest rate: 119; -0.27; 2.72
  - Real depreciation: 119; -0.22; 0.58
  - Foreign ownership: 119; 52.95; 45.21
  - Public ownership: 119; 13.14; 31.35
- Lithuania:
  - Bank credit growth: 51; 40.18; 50.61
  - Distance to default: 51; 13.94; 15.02
  - Net interest margin: 51; 4.67; 2.28
  - Cost-to-income ratio: 51; 82.93; 21.99
  - Liquidity ratio: 51; 23.91; 11.12
  - Bank size: 51; 5.57; 1.53
  - Real GDP growth: 51; 6.07; 4.48
  - GDP per capita: 51; 39.51; 10.88
  - Real interest rate: 51; 1.79; 2.26
  - Real depreciation: 51; -0.63; 0.72
  - Foreign ownership: 51; 58.56; 42.95
  - Public ownership: 51; 17.11; 35.68

### Methodological notes and supplements
- Bankscope limitations addressed: lack of historical ownership data, duplicated entries for consolidated/unconsolidated data, and possible measurement errors—handled through matching, reranking, choice rules for balance sheet type, and outlier exclusions.
- Loan composition variables from supervisory data were transformed into dummy indicators for rapidly growing household and foreign-currency portfolios to preserve confidentiality and reduce measurement error.
- Appendix Tables 1–3 present: summary statistics for the final data set and definitions/units of measurement for bank-level and macroeconomic variables.

*Source: APPENDIX I. DATA SOURCES AND METHODOLOGY*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08219.pdf_