## wp18279

## Source details

**Canonical URL:** [wp18279](https://www.imf.org/-/media/files/publications/wp/2018/wp18279.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2018/wp18279.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2018/wp18279.pdf.json)

---

### Introduction
- Purpose: Examine whether financial intermediation costs of banks (net interest margins: NIM) are influenced by the quality of governance across countries and over time.
- Sample: 100 countries spanning the period 1996 to 2015; the dataset is highly unbalanced across countries and time.
- Primary measure: Net interest margin (net interest income in percent of interest-earning assets: NIM) used as a proxy for financial intermediation costs and efficiency.

### Conceptual framework and hypotheses
- Theoretical basis: Bank-dealership model (Ho and Saunders, 1981); banks are risk averse and set spreads to compensate for risks and uncertainty.
- Model predictions:
  - Greater market power → greater spread.
  - Higher risk aversion / higher capital adequacy → affects funding costs and margins (ambiguous direction).
  - Larger average transaction size → lower margins via returns to scale.
  - More volatile interest rates → higher risk premium and higher margins.
- Governance concept: Legislation, formal and informal rules governing behavior of society, institutions, organizations, firms, individuals, and markets; improves welfare by lowering costs, risks, and uncertainty.
- Expected governance effects:
  - Better property rights, rule of law, regulatory quality, and judicial efficiency → lower NIM.
  - Corruption generally increases intermediation costs, constrains efficient lending, and raises credit risk; short-run effects may be ambiguous in some theoretical models.

### Data, empirical approach, and robustness
- Data sources:
  - NIM and aggregated banking variables mainly from Beck et al. (2017) based on Bankscope.
  - Macroeconomic data primarily from International Financial Statistics, IMF.
  - Governance indicators from Doing Business (World Bank), Worldwide Governance Indicators (Kaufmann, Kraay and Mastruzzi, 2010), and ethics of private firms from the Global Competitiveness Index (World Economic Forum).
- Methodology:
  - Single-stage regression approach combining theoretical determinants and non-theoretical control variables.
  - Regression forms and diagnostics used: pooled OLS with time effects; random effects; instrumental variables for endogeneity; cross-sectional dependence tests (Pesaran 2006); Driscoll and Kraay (1998) standard errors; robustness checks for non-linearity, lagged dependent variables, and outliers.

### Key empirical findings
- Determinants of higher NIM:
  - Higher operating costs → higher NIM.
  - Greater risk aversion and higher credit risk → higher NIM.
- Determinants of lower NIM:
  - Larger transaction size / greater financial deepening (bank credit to private sector) → lower NIM.
  - Greater financial openness → tends to reduce NIM, likely via increased competition.
- Governance effects:
  - Various aspects of good governance significantly reduce NIM; the governance–NIM relationship remained intact after the Global Financial Crisis (GFC).
- GFC effects:
  - The impact of the GFC on NIM operated mainly through increased credit risk.
  - The GFC did not lead to a structural break in the governance–NIM relationship.
- Capital markets:
  - Size of capital market financing not found to be significant in reducing NIM in main specifications.

### Net Interest Margin (NIM) trends and drivers
- NIM series examined 1996–2015.
- Observation: "there seems to be a modest downward trend from 2013."
- Finding: "The GFC did not bring any significant change in the NIM trend."
- European context: net interest income in the Euro Area declined only marginally (Burke and Garcia, 2017); NIM remained broadly stable over the financial cycle (Detragiache et al., 2018).
- Policy environment: monetary policy loosening and nonstandard actions (including quantitative easing) affected deposit and lending rates; funding costs limited by the zero lower bound; competition for lending to viable projects may have further driven down lending rates causing some compression of the NIM, especially in European countries from 2013.

### Econometric approach and baseline equation
- Estimated specification:
  - NIM_it = α0 + Σ_k α1,k BSV_k,it + Σ_k α2,k MAC_k,it + Σ_k α3,k GOV_k,it + Σ_k α4,k OTH_k,it + ε_it
  - BSV: competition, operating cost, transaction size, risk aversion, credit risk.
  - MAC: inflation, real economic growth.
  - GOV: rule of law, regulatory quality, recovery of assets, perceived corruption, ethics of private firms.
- Four main estimation dimensions:
  - pooled OLS (including time effects);
  - random effects model with Breusch-Pagan (1979) test;
  - instrument variable estimations;
  - adjustments for cross-sectional dependence (Pesaran 2006) and Driscoll and Kraay (1998) standard errors.

### Basic specification: key coefficients (Table 1)
- Sample: Number of observations = 1187; R squared - overall = 0.72 (OLS) and 0.71 (GLS).
- Selected coefficients (OLS; GLS in parentheses) with reported standard errors:
  - Competition / concentration: -0.0087 *** (−0.0080 *) (0.0046)
  - Operating costs: 0.3960 *** (0.2610 **) (0.0702) / (−0.1050)
  - Transaction size / financial deepening: -1.6167 *** (-1.3850 ***) (0.1560) / (0.2320)
  - Risk-aversion (CAR): 0.0371 ** (0.0274 *) (0.0112) / (0.0119)
  - Credit risk (Provisions): 0.0034 *** (0.0027 ***) (0.0009) / (0.0006)
  - Inflation (CPI): 3.1134 ** (0.5900) (1.0739) / (1.1120)
- Interpretation:
  - Concentration measure is negatively correlated with NIM in the basic specification; significance varies in extended models.
  - Overhead costs positively associated with NIM.
  - Transaction size negatively associated with NIM (returns to scale).
  - Risk aversion and credit risk positively associated with NIM.
  - Macroeconomic variables generally not robustly significant; pooled basic model shows positive association of inflation with higher NIM.

### Extended models: governance, interconnectedness, capital markets
- Governance (GLS reported coefficients and standard errors):
  - Rule of law: -0.7133 *** (0.1522)
  - Regulatory quality: -0.3782 *** (0.1455)
  - Insolvency framework: -0.0160 *** (0.0040)
  - Contract enforcement: -0.0165 * (0.0090)
  - Government effectiveness: -0.6984 *** (0.1518)
  - Control for perceived corruption: -0.5069 *** (0.1159)
  - Ethics of private firms: -0.2362 * (0.1243)
  - Number of observations varies by specification (e.g., 1102, 835, 748); R squared - overall ranges 0.71–0.73.
- Interconnectedness:
  - Greater international debt and cross-border bank loans tend to reduce NIM.
  - Examples of coefficients:
    - Insolvency framework: -0.0144 *** (-0.0142 ***) (0.0040 / 0.0041)
    - Contract enforcement: -0.0178 * (-0.0166 *) (0.0096 / 0.0084)
    - International debt: -0.0050 (-0.0057 *) (0.0031 / 0.0029)
    - Cross-border bank loans: -0.0072 ** (-0.0083 ***) (0.0029 / 0.0026)
  - R squared - overall 0.73–0.74.
- Capital market variables:
  - Deeper private capital markets do not appear to significantly reduce NIM in main specifications.
  - Capital market proxies: stock market capitalization, traded stocks in percent of GDP, private bond issuance.
  - Operating costs and transaction size remain significant; some governance proxies significant in some specs.
  - Number of observations examples: 588, 604, 573, 318; R squared - overall 0.74–0.80.

### Global Financial Crisis (GFC) impact (Table 5)
- GFC dummy: defined as 1 in 2007 and afterwards.
- Evidence:
  - "There is only very limited evidence that the GFC tended to reduce the NIM, with the GFC dummy being significant in less than half the specifications."
  - Most interaction terms (GFC dummy times explanatory variables) generally not statistically significant.
  - Notable exception: credit risk interaction — "GFC dummy / credit risk 0.0045 ** 0.0043 ** (0.0019)(0.0019)" — implying credit risk became more strongly associated with NIM after the GFC.
- Coefficients across GFC-extended specifications:
  - Operating costs: consistently positive and significant (examples: 0.2363 *** ... 0.2386 ***; standard errors ~0.0749–0.0905).
  - Transaction size: consistently negative and significant (examples: -1.3100 *** ... -1.4159 ***; standard errors ~0.2449–0.3625).
  - Risk-aversion (CAR): consistently positive and significant (examples: 0.0404 *** ... 0.0302 ***).
  - Credit risk (Provisions): generally positive and significant (e.g., 0.0033 *** ... 0.0034 ***), with some isolated anomalous entries.

### Robustness checks and additional methods
- Additional governance indicators (property right protection, judicial independence, impartiality of courts) show similar results; coverage more limited.
- Instrumental variables (voice and accountability, political stability) used to address endogeneity; lagged explanatory variables included.
- Cross-sectional dependence accounted for (Pesaran 2006) and Driscoll and Kraay (1998) standard errors applied.
- Probit model: dependent variable cut-off at 75th percentile; results broadly similar to basic regressions.
- Panel estimations with lagged dependent variable: bias-corrected LSDV estimators and bootstrapping used; lagged dependent variable reduces coefficients somewhat but does not materially change results.
- Sensitivity to outliers: robust regressions (iterative weighted least squares) produce broadly similar results.
- Additional checks: governance indicators in differences; subsample analyses by income groups; alternative dependent variables (interest rate spreads, ROA before taxation) had weaker explanatory power.

### Robustness checks — selected estimates (Table 15)
- Selected coefficients and standard errors from Table 15 (columns (1) to (5)) — exact reported values:
  - concentration:
    - (1): -0.00227 (0.00159)
    - (2): -0.00536*** (0.00164)
    - (3): -0.00131 (0.00203)
    - (4): -0.00633*** (0.00199)
    - (5): -0.00206 (0.00168)
  - operating cost:
    - (1): 0.795*** (0.0140)
    - (2): 0.801*** (0.0147)
    - (3): 0.812*** (0.0175)
    - (4): 0.803*** (0.0178)
    - (5): 0.819*** (0.0143)
  - transaction size:
    - (1): -0.334*** (0.0657)
    - (2): -0.495*** (0.0679)
    - (3): -0.461*** (0.0742)
    - (4): -0.629*** (0.0693)
    - (5): -0.421*** (0.0654)
  - risk aversion - CAR:
    - (1): 0.0338*** (0.00677)
    - (2): 0.0356*** (0.00715)
    - (3): 0.0355*** (0.00894)
    - (4): 0.0338*** (0.00905)
    - (5): 0.0337*** (0.00697)
  - credit risk - provisions:
    - (1): 0.00214*** (0.000537)
    - (2): 0.00249*** (0.000566)
    - (3): 0.000897 (0.000634)
    - (4): 0.00122* (0.000642)
    - (5): 0.00242*** (0.000554)
  - GDP growth:
    - (1): 4.812*** (0.986)
    - (2): 5.275*** (1.042)
    - (3): 5.849*** (1.220)
    - (4): 5.147*** (1.244)
    - (5): 5.227*** (1.012)
  - inflation:
    - (1): -0.136 (0.573)
    - (2): 0.380 (0.615)
    - (3): 2.609*** (0.800)
    - (4): 3.853*** (0.799)
    - (5): 0.242 (0.586)
  - governance proxies (examples):
    - rule of law (1): -0.484*** (0.0455)
    - regulatory quality (2): -0.312*** (0.0525)
    - insolvency framework (3): -0.00995*** (0.00178)
    - contract enforcement (4): -0.00828*** (0.00257)
    - control of corruption (5): -0.336*** (0.0409)
  - Observations (N):
    - (1): 1,102
    - (2): 1,102
    - (3): 835
    - (4): 835
    - (5): 1,102

### Quantitative magnitudes and illustrative simulations
- Cross-country patterns:
  - High-income countries (Asia, Europe, North America) typically have NIM on average below 3 percent and do not exceed 10 percent.
  - Low-income countries (Sub-Saharan Africa, Latin America and the Caribbean) have NIM about twice as high as high-income countries; many low-income countries have NIM exceeding 15 percent.
- Policy simulations methodology:
  - Fit NIM using seven models in Table 2 (where data available).
  - Calculate average fitted NIM using actual values; recalculate assuming each governance variable is at the top 10 percent threshold.
  - Difference gives lower NIM; multiply difference by bank credit to private sector (percent of GDP) to yield estimated annual savings in percent of GDP.
  - Simple average of annual estimated savings calculated when data are available.
- Key simulation results (Table 6):
  - Whole sample estimated annual savings (Percent of GDP): 0.28; Lower net interest margin (Percentage points): 0.69; Number of countries: 95
  - High income countries: 0.23 (Percent of GDP); 0.34 (Percentage points); Number of countries: 35
  - Upper middle income countries: 0.35; 0.83; Number of countries: 28
  - Lower middle income countries: 0.31; 1.00; Number of countries: 24
  - Low income countries: 0.12; 1.04; Number of countries: 8
- Interpretation:
  - Low-income countries could reduce NIM by on average about 1 percentage point; high-income countries by about 0.3 percentage point (some high-income observations excluded as their governance already at top 10 percent).
  - Gains in percent of GDP are almost inverse across income groups due to greater financial deepening in higher income countries.
  - Annual savings in some countries could amount to about 1/3 percent of GDP per year; accumulated savings could be sizable.
- Relative impact of governance dimensions:
  - Improvements in general governance indicators (rule of law, government effectiveness, reducing corruption) have largest impact on NIM.
  - "Ethics of private firms" improvement may show even bigger impact (smaller sample).
- Caveats:
  - Simulations illustrative; governance indicators limited in coverage historically; some bank characteristics may correlate with governance; raising governance to top 10 percent is ambitious; simultaneous improvements and externalities not fully captured.

### Correlations and patterns (selected)
- Selected pairwise correlations (exact reported values):
  - NIM — Interest spread: 0.65
  - NIM — ROA: 0.64
  - NIM — Operating costs: 0.74
  - NIM — Transaction size/financial depth: -0.80
  - NIM — Rule of law: -0.73
  - Transaction size/financial depth — Private sector bank credit: 0.92
  - Rule of law — Regulatory quality: 0.93
  - International debt — Cross-border bank loans: 1.00
- Note: full correlation matrix reported in Table 9.

### Country coverage and data notes
- Appendix I lists countries by income group (High income, Upper middle income, Lower middle income, Low income).
- Panel is unbalanced across time; some countries excluded due to data limitations.
- Appendix II provides variable definitions, summary statistics and sources for all variables used (examples include NIM, interest spread, ROA, competition/concentration, operating costs, transaction size, CAR, provisions, real GDP growth, CPI, rule of law, insolvency framework, international debt, stock market capitalization, private sector bank credit, proxy for corporate income tax).

### Policy implications and interpretation
- Improving public governance (property rights, rule of law, contract enforcement, judiciary efficiency) appears to lower bank intermediation costs and thus can support more efficient financial intermediation and sustainable economic growth.
- Financial openness can reduce NIM, suggesting a competition channel; policies that open banking sectors to competition may lower intermediation costs.
- Strengthening institutions that reduce operating costs, credit risk, and uncertainty (including anti-corruption measures and better information systems such as credit bureaus) is likely to yield tangible reductions in NIM, especially in low-income countries.
- Political economy caveats: collective action problems and concentrated interests benefiting from poor governance may hinder reforms.

### Conclusions (Section V)
- Main conclusions:
  - Good governance practices are associated with lower financial intermediation costs (NIM) across the 1996–2015 sample.
  - If governance improved to top 10 percent threshold:
    - NIM of low-income countries would decline on average by about 1 percentage point.
    - NIM of high-income countries would decline on average by about 0.3 percentage point.
  - Gains in percent of GDP are almost inverse due to greater financial deepening in higher income countries; could amount to about 1/3 percent of GDP per year in many countries.
- Reconfirmed findings:
  - Bank characteristics matter: higher operating costs, risk aversion, credit risk → higher margins; higher transaction size → lower margins.
  - GFC impact on NIM worked mainly through credit risk: credit risk became more strongly associated with NIM after the GFC.
- Further research:
  - Explore why governance improvements are not pursued more proactively and investigate additional channels and political economy constraints.

*Source: https://www.imf.org/-/media/files/publications/wp/2018/wp18279.pdf*

### REFERENCES ___________________________________________________________25

### wp18279 - REFERENCES ___________________________________________________________25

### Introduction
- Purpose: Examine whether financial intermediation costs of banks (net interest margins: NIM) are influenced by the quality of governance across countries and over time.
- Sample: 100 countries spanning the period 1996 to 2015; the dataset is highly unbalanced across countries and time.
- Primary measure: Net interest margin (net interest income in percent of interest-earning assets: NIM) used as a proxy for financial intermediation costs and efficiency.

### Conceptual framework and hypotheses
- Theoretical basis: Bank-dealership model (Ho and Saunders, 1981); banks are risk averse and set spreads to compensate for risks and uncertainty.
- Banking-sector predictions from the model:
  - Greater market power → greater spread.
  - Higher risk aversion / higher capital adequacy → affects funding costs and margins (ambiguous direction).
  - Larger average transaction size → lower margins via returns to scale.
  - More volatile interest rates → higher risk premium and higher margins.
- Governance concept used: Legislation, formal and informal rules governing behavior of society, institutions, organizations, firms, individuals, and markets; improves welfare by lowering costs, risks, and uncertainty.
- Expected governance effects:
  - Better property rights, rule of law, regulatory quality, and judicial efficiency → lower NIM through reduced costs, risks, and uncertainty.
  - Corruption generally increases intermediation costs, constrains efficient lending, and raises credit risk; however, some theoretical models (e.g., “grease-the-wheels”) can predict ambiguous short-run lending effects.

### Data, empirical approach, and robustness
- Data sources:
  - NIM and aggregated banking variables mainly from Beck et al. (2017) based on Bankscope.
  - Macroeconomic data primarily from International Financial Statistics, IMF.
  - Governance indicators from Doing Business (World Bank), Worldwide Governance Indicators (Kaufmann, Kraay and Mastruzzi, 2010), and ethics of private firms from the Global Competitiveness Index (World Economic Forum).
- Methodological choices:
  - Single-stage regression approach that combines theoretical determinants and non-theoretical control variables (consistent with McShane and Sharpe (1985), Angbazo (1997), Maudos and Fernandez de Guevara (2004), Demirgüç-Kunt et al. (2004), Poghosyan (2013)).
  - Robustness checks include alternative governance indicators, instrumental variables for endogeneity, cross-sectional dependence tests, non-linearity checks, lagged dependent variable specifications, and outlier analyses.

### Key empirical findings
- Determinants of higher NIM:
  - Higher operating costs → higher NIM.
  - Greater risk aversion and higher credit risk → higher NIM.
- Determinants of lower NIM:
  - Larger transaction size / greater financial deepening (bank credit to private sector) → lower NIM.
  - Greater financial openness → tends to reduce NIM, likely via increased competition.
- Governance effects:
  - Various aspects of good governance significantly reduce NIM; the relationship between governance and NIM remained intact after the global financial crisis (no structural break detected in that relationship).
- Global Financial Crisis (GFC) effects:
  - The impact of the GFC on NIM operated mainly through increased credit risk.
  - The GFC did not lead to a structural break in the governance–NIM relationship.
- Capital markets:
  - Size of capital market financing not found to be significant in reducing NIM in the main specifications.

### Quantitative magnitudes and illustrative simulations
- Cross-country patterns:
  - High-income countries (Asia, Europe, North America) typically have NIM on average below 3 percent and do not exceed 10 percent.
  - Low-income countries (Sub-Saharan Africa, Latin America and the Caribbean) have NIM about twice as high as high-income countries; many low-income countries have NIM exceeding 15 percent.
- Potential gains from improved governance:
  - If governance improves to the top 10 percent threshold:
    - NIM would decline on average by about 1 percentage point for low-income countries.
    - NIM would decline on average by about 0.3 percentage point for high-income countries.
  - The “gains” measured in percent of GDP are almost inverse between low- and high-income countries because of the greater size of bank intermediation (financial deepening) in high-income countries.

### Policy implications and interpretation
- Improving public governance (property rights, rule of law, contract enforcement, judiciary efficiency) appears to lower bank intermediation costs and thus can support more efficient financial intermediation and sustainable economic growth.
- Financial openness can reduce NIM, suggesting a competition channel; policies that open banking sectors to competition may lower intermediation costs.
- Strengthening institutions that reduce operating costs, credit risk, and uncertainty (including anti-corruption measures and better information systems such as credit bureaus) is likely to yield tangible reductions in NIM, especially in low-income countries.

### Conclusion
- The study finds that governance quality is a significant determinant of bank net interest margins across a broad sample of 100 countries from 1996 to 2015.
- Improvements in governance can produce sizable reductions in financial intermediation costs, with larger margin declines in low-income countries but potentially larger GDP gains in high-income countries due to deeper financial intermediation.

*Source: https://www.imf.org/-/media/files/publications/wp/2018/wp18279.pdf*

### 2015. However, there seems to be a modest downward trend from 2013. Interestingly, the

### wp18279 - 2015. However, there seems to be a modest downward trend from 2013. Interestingly, the

### Net Interest Margin (NIM) trends and drivers
- NIM series examined 1996–2015 (Figures referenced: Net Interest Margins in Different Country Groups, Net Interest Margins over Time).  
- Observation: "there seems to be a modest downward trend from 2013."  
- Finding: "The GFC did not bring any significant change in the NIM trend."  
- European context: "net interest income in the Euro Area declined only marginally according to Burke and Garcia (2017) and the NIM remained broadly stable over the financial cycle according to Detragiache et al. (2018)."  
- Policy environment: monetary policy loosening and nonstandard actions (including quantitative easing) affected deposit and lending rates; funding costs limited by the zero lower bound; competition for lending to viable projects may have further driven down lending rates causing some compression of the NIM, especially in European countries from 2013.

### Econometric approach
- Four main dimensions:
  - pooled OLS (including time effects);
  - random effects model following Plumber et al. (2005, 2007) and Bell and Jones (2015), with Breusch-Pagan (1979) test for relevance;
  - instrument variable estimations to address endogeneity;
  - adjustments for cross-sectional dependence using Pesaran (2006) test and standard errors following Driscoll and Kraay (1998).
- Single-stage methodology estimating variants of:
  - NIM_it = α0 + Σ_k α1,k BSV_k,it + Σ_k α2,k MAC_k,it + Σ_k α3,k GOV_k,it + Σ_k α4,k OTH_k,it + ε_it
  - BSV: competition, operating cost, transaction size, risk aversion, credit risk.
  - MAC: inflation, real economic growth.
  - GOV: rule of law, regulatory quality, recovery of assets, perceived corruption, ethics of private firms, etc.

### Basic specification: key coefficients (Table 1)
- Sample: Number of observations = 1187; R squared - overall = 0.72 (OLS) and 0.71 (GLS).
- Coefficients (OLS; GLS in parentheses) with standard errors:
  - Competition / concentration: -0.0087 *** (−0.0080 *) (0.0046)
  - Operating costs: 0.3960 *** (0.2610 **) (0.0702) / (−0.1050)
  - Transaction size / financial deepening: -1.6167 *** (-1.3850 ***) (0.1560) / (0.2320)
  - Risk-aversion (CAR): 0.0371 ** (0.0274 *) (0.0112) / (0.0119)
  - Credit risk (Provisions): 0.0034 *** (0.0027 ***) (0.0009) / (0.0006)
  - Inflation (CPI): 3.1134 ** (0.5900) (1.0739) / (1.1120)
- Interpretation:
  - Concentration measure (share of assets of the three largest banks) is negatively correlated with NIM in basic specifications; extended models show negative coefficients but significance varies.
  - Overhead cost positively associated with NIM.
  - Transaction size negatively associated with NIM (increasing returns to scale).
  - Risk aversion and credit risk positively associated with NIM.
  - Macroeconomic variables generally not robustly significant, with pooled basic model showing positive association of inflation with higher NIM.

### Extended models: governance, interconnectedness, capital markets
- Governance (Table 2):
  - Governance indicators consistently associated with lower NIM.
  - Selected significant coefficients (GLS specifications):
    - Rule of law: -0.7133 *** (0.1522)
    - Regulatory quality: -0.3782 *** (0.1455)
    - Insolvency framework: -0.0160 *** (0.0040)
    - Contract enforcement: -0.0165 * (0.0090)
    - Government effectiveness: -0.6984 *** (0.1518)
    - Control for perceived corruption: -0.5069 *** (0.1159)
    - Ethics of private firms: -0.2362 * (0.1243)
  - Number of observations varies by specification (e.g., 1102, 835, 748); R squared - overall ranges 0.71–0.73.
- Interconnectedness (Table 3):
  - Greater international debt and cross-border bank loans tend to reduce NIM.
  - Examples of significant coefficients:
    - Insolvency framework: -0.0144 *** (-0.0142 ***) (0.0040 / 0.0041)
    - Contract enforcement: -0.0178 * (-0.0166 *) (0.0096 / 0.0084)
    - International debt: -0.0050 (-0.0057 *) (0.0031 / 0.0029)
    - Cross-border bank loans: -0.0072 ** (-0.0083 ***) (0.0029 / 0.0026)
  - R squared - overall reported as 0.73–0.74.
- Capital market variables (Table 4):
  - Deeper private capital markets do not appear to significantly reduce NIM.
  - Capital market proxies used: stock market capitalization, traded stocks in percent of GDP, private bond issuance.
  - Selected coefficients (GLS panels): operating costs and transaction size retain significance; some governance proxies (insolvency, contract enforcement) significant in some specifications.
  - Number of observations examples: 588, 604, 573, 318; R squared - overall 0.74–0.80.

### Global Financial Crisis (GFC) impact (Table 5)
- GFC dummy: defined as 1 in 2007 and afterwards.
- Evidence: "There is only very limited evidence that the GFC tended to reduce the NIM, with the GFC dummy being significant in less than half the specifications."
- Interaction results:
  - Most interaction terms (GFC dummy times explanatory variables) generally not statistically significant.
  - Notable exception: credit risk interaction — "GFC dummy / credit risk 0.0045 ** 0.0043 ** (0.0019)(0.0019)" — implying credit risk became more strongly associated with NIM after the GFC.
- Coefficients across many GFC-extended specifications:
  - Operating costs consistently positive and significant (examples: 0.2363 *** ... 0.2386 *** with standard errors ~0.0749–0.0905).
  - Transaction size consistently negative and significant (examples: -1.3100 *** ... -1.4159 *** with standard errors ~0.2449–0.3625).
  - Risk-aversion (CAR) consistently positive and significant across many specs (examples: 0.0404 *** ... 0.0302 ***).
  - Credit risk (Provisions) generally positive and significant (e.g., 0.0033 *** ... 0.0034 ***), with some specifications showing anomalous entries (e.g., -0.0005 in isolated columns).
- GFC dummy coefficients vary widely across columns, with several negative and some significant (e.g., -0.4544 * (0.2634), -0.7393 ** (0.2893), -1.0143 (1.3365) etc.), but overall limited robust evidence of a direct GFC effect on NIM.

### Robustness checks and additional methods
- Additional governance indicators (property right protection, judicial independence, impartiality of courts) show similar results (Table 10); coverage is more limited.
- Instrumental variables (Table 11):
  - Instruments: voice and accountability, political stability indicators; lagged explanatory variables included.
  - Purpose: address potential endogeneity of governance indicators.
- Cross-sectional dependence accounted for (Pesaran 2006 test) and standard errors adjusted via Driscoll and Kraay (1998) (Table 12).
- Probit model (Table 13):
  - Dependent variable cut-off at 75th percentile; results broadly similar to basic regressions.
- Panel estimations with lagged dependent variable (Table 14):
  - Bias-corrected LSDV estimators and bootstrapping used; lagged dependent variable reduces coefficients somewhat but does not materially change results.
- Sensitivity to outliers (Table 15):
  - Robust regressions (iterative weighted least squares) produce broadly similar results.
- Additional checks:
  - Specifications with governance indicators in differences confirm better governance associated with lower costs.
  - Subsample analyses by income groups performed (results available upon request).
  - Alternative dependent variables (interest rate spreads, ROA before taxation) had weaker explanatory power; taxation may be an important missing variable with limited comparable data.

### Policy simulations: estimated cost of poor governance (Tables 6 and methodology)
- Methodology:
  - Fit NIM using seven models in Table 2 (to extent data available).
  - Calculate simple average fitted NIM using actual values; recalculate assuming each governance variable is at the top 10 percent threshold.
  - Difference gives lower NIM; multiply difference by bank credit to private sector (percent of GDP) to yield estimated annual savings in percent of GDP.
  - Simple average of annual estimated savings calculated when data are available.
- Key simulation results (Table 6):
  - Whole sample estimated annual savings (Percent of GDP): 0.28; Lower net interest margin (Percentage points): 0.69; Number of countries: 95
  - High income countries: 0.23 (Percent of GDP); 0.34 (Percentage points); Number of countries: 35
  - Upper middle income countries: 0.35; 0.83; Number of countries: 28
  - Lower middle income countries: 0.31; 1.00; Number of countries: 24
  - Low income countries: 0.12; 1.04; Number of countries: 8
- Interpretation:
  - Low-income countries could reduce NIM by on average about 1 percentage point; high-income countries by about 0.3 percentage point (some high-income observations excluded as their governance already at top 10 percent).
  - Gains in percent of GDP are almost inverse across income groups due to greater financial deepening in higher income countries.
  - Annual savings in some countries could amount to about 1/3 percent of GDP per year; accumulated savings could be sizable.
- Relative impact of governance dimensions:
  - Improvements in general governance indicators (rule of law, government effectiveness, reducing corruption) have largest impact on NIM.
  - "Ethics of private firms" improvement may show even bigger impact (smaller sample).
- Caveats:
  - Simulations illustrative; governance indicators limited in coverage historically; some bank characteristics may correlate with governance; raising governance to top 10 percent is ambitious; simultaneous improvements across indicators and other externalities not fully captured.

### Conclusions (Section V)
- Main conclusions:
  - Good governance practices are associated with lower financial intermediation costs (NIM).
  - If governance improved to top 10 percent threshold:
    - NIM of low-income countries would decline on average by about 1 percentage point.
    - NIM of high-income countries would decline on average by about 0.3 percentage point.
  - Gains in percent of GDP are almost inverse due to greater financial deepening in higher income countries; could amount to about 1/3 percent of GDP per year in many countries.
- Reconfirmed findings:
  - Bank characteristics matter: higher operating costs, risk aversion, credit risk → higher margins; higher transaction size → lower margins (returns to scale).
  - GFC impact on NIM worked mainly through credit risk: credit risk more strongly associated with NIM after the GFC.
- Policy implication and further research:
  - Improvements in governance could yield substantial reductions in financial intermediation costs, but political economy obstacles (collective action problems, concentrated interests benefiting from poor governance) may hinder reforms.
  - Further research needed to explore why governance improvements are not pursued more proactively.

*Source: IMF staff working paper content (wp18279 - 2015: extracted sections provided).*

### REFERENCES

### REFERENCES

### Major cited works and authors
- Acemoğlu, Daron; and James A. Robinson, 2016, “Paths to Inclusive Political Institutions,” https://www.researchgate.net/publication/308340491_Paths_to_Inclusive_Political_Insti
tutions.
- Acemoğlu, Daron; and James A. Robinson, 2012, Why Nations Fail: The origins of Power, Prosperity, and Poverty, Crown Publishers, New York.
- Acemoğlu, Daron; Simon Johnson; and James A. Robinson, 2008, “The Role of Institutions in Growth and Development, Working Paper No. 10, Commission on Growth and Development, World Bank.
- A large set of empirical and theoretical works on institutions, governance, corruption, and banking intermediation, including (but not limited to): Arrow (1974); Coase (1960); North (1981, 1989, 1991); Shleifer and Vishny (1993); Mauro (1995); Levine (1997, 2005); Beck et al. (2017); Demirgüç‑Kunt and Huizinga (1999); Claessens and Laeven (2004); Barth et al. (2009, 2013); Pesaran (2006); Pesaran & Chudik (2015); and multiple IMF, World Bank, OECD, UN, and BIS publications cited throughout the reference list.
- Specific empirical studies on bank interest margins, profitability, concentration, and efficiency cited include: Ho and Saunders (1981); Angbazo (1997); Demirgüç‑Kunt, Laeven, and Levine (2004); Maudos and Fernández de Guevara (2004); Staikouras and Wood (2004); Saunders and Schumacher (2000); Claeys and Vander Vennet (2008); and many country- and region-focused studies (e.g., Kasman et al., Fungáčová and Poghosyan, Burke and Plata Garcia).

### Thematic emphases in the bibliography
- Institutions, governance, and economic growth: Acemoğlu & Robinson; North; Hall & Jones; Putterman; Bevir; Williamson.
- Corruption, rent capture, and public sector governance: Shleifer & Vishny; Mauro; Tanzi & Davoodi; World Bank and OECD diagnostics.
- Banking sector structure, competition, and efficiency: Beck et al. (2017); Demirgüç‑Kunt et al.; Claessens & Laeven; Kasman et al.; Barth et al.
- Determinants of bank interest margins, profitability, and financial intermediation costs: Ho & Saunders; Demirgüç‑Kunt & Huizinga; Maudos & Fernández de Guevara; Angbazo; Liebeg & Schwaiger; Weill.
- Measurement and econometrics for panel data and cross-sectional inference: Pesaran (2006); Driscoll & Kraay (1998); Breusch & Pagan (1979); Kiviet (1995, 1999); Bun & Kiviet (2003); Judson & Owen (1999); Plümper & Troeger (2005, 2007); Bell & Jones (2015).

---

### Appendix I. Countries Included in the Sample

- High income countries:
  - Australia, Austria, Belgium, Canada, Chile, Croatia, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Israel, Italy, Japan, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Norway, Poland, Portugal, Saudi Arabia, Seychelles, Singapore, Slovak Republic, Slovenia, Spain, Sweden, Switzerland, Trinidad and Tobago, United Arab Emirates, United Kingdom, United States, Uruguay
- Upper middle income countries:
  - Algeria, Argentina, Belarus, Bosnia and Herzegovina, Botswana, Brazil, Bulgaria, China, Colombia, Costa Rica, Dominican Republic, Ecuador, Georgia, Ghana, Kazakhstan, Lebanon, Macedonia, FYR, Malaysia, Mauritius, Mexico, Namibia, Panama, Paraguay, Peru, Romania, South Africa, Thailand, Turkey, Uruguay (listed previously under high income as well)
- Lower middle income countries:
  - Armenia, Bangladesh, Bhutan, Cameroon, El Salvador, Gabon, Guatemala, Honduras, India, Indonesia, Kenya, Kosovo, Lesotho, Moldova, Nigeria, Pakistan, Philippines, Sri Lanka, Swaziland, Tajikistan, Ukraine, Vietnam, West Bank and Gaza, Zambia
- Low income countries:
  - Afghanistan, Burundi, Cambodia, Guinea, Madagascar, Rwanda, Tanzania, Uganda

- Number of countries:    
  4028248

- Note: The panel is unbalanced across time. Other countries were not included primarily due to data limitations.

---

### Appendix II. Data Description and Sources (Table 8 — Variables, summary statistics, and sources)
- Panel coverage note: Countries in the Sample 1996 - 2015

- Financial intermediation costs
  - Net interest margin (NIM)
    - ObsMeanStd. Dev.Min.Max
    - 1,8974.73.10.121.2
    - Comment: "Accounting value of bank's net interest revenue as a share of its interest-bearing (total earning) assets."
    - Source(s): Beck et al. (2017)
  - Interest spread
    - ObsMeanStd. Dev.Min.Max
    - 1,4488.110.1-6.9216.4
    - Comment: Difference between lending rates and deposit rates. Negative spread from countries with interest controls and highest rate from countries with hyper inflation.
    - Source(s): International Financial Statistics, IMF
  - Return on Assets (ROA)
    - ObsMeanStd. Dev.Min.Max
    - 1,9011.21.6-9.99.3
    - Comment: "Average Return on Assets (net Income/total assets)."
    - Source(s): Beck et al. (2017)
  - Return on Assets (ROA_FSI)
    - ObsMeanStd. Dev.Min.Max
    - 1,4251.22.0-25.69.8
    - Comment: Return on Assets (net Income before extraordinary income and taxes/total assets). This variable is used in the regressions.
    - Source(s): Financial Soundness Indicators Database and Global Financial Stability Report (GFSR), IMF

- Bank characteristics
  - Competition/concentration
    - ObsMeanStd. Dev.Min.Max
    - 1,82766.419.417.3100.0
    - Comment: "Assets of three largest banks as a share of assets of all commercial banks."
    - Source(s): Beck et al. (2017)
  - Operating costs
    - ObsMeanStd. Dev.Min.Max
    - 1,9083.93.20.0457.8
    - Comment: "Accounting value of a bank's overhead costs as a share of its total assets."
    - Source(s): Beck et al. (2017)
  - Transaction size/ fin. Deep.
    - ObsMeanStd. Dev.Min.Max
    - 1,9593.80.9-0.15.6
    - Comment: Log to "claims on domestic real nonfinancial sector by deposit money banks as a share of GDP, calculated using the following deflation method:  {(0.5)*[Ft/P_et + Ft-1/P_et-1]}/[GDPt/P_at] where F is deposit money bank claims, P_e is end-of period CPI, and P_a is average annual CPI."
    - Source(s): Beck et al. (2017)
  - Risk-aversion (CAR)
    - ObsMeanStd. Dev.Min.Max
    - 1,42115.84.8-18.241.3
    - Comment: Regulatory capital to risk-weighted assets. Definition may differ across countries.
    - Source(s): Financial Soundness Indicators Database and Global Financial Stability Report (GFSR), IMF
  - Credit risk (Provisions)
    - ObsMeanStd. Dev.Min.Max
    - 1,23476.946.19.9806.8
    - Comment: Provision ratio, share of NPLs provisioned. The ratio can be above 100, in particular if general provisions are very high.
    - Source(s): Financial Soundness Indicators Database and GFSR, IMF
  - Non-performing loans (NPL)
    - ObsMeanStd. Dev.Min.Max
    - 1,4396.97.60.0274.1
    - Comment: NPLs in percent of gross loans. Definitions may vary across countries and over time.
    - Source(s): Financial Soundness Indicators Database and GFSR, IMF

- Business cycle
  - Real GDP growth
    - ObsMeanStd. Dev.Min.Max
    - 1,9880.040.04-0.150.62
    - Comment: Change of real GDP (rgdpg = rgdp/rgdp[_n-1]-1). Multiply by 100 to get it in percentage change.
    - Source(s): International Financial Statistics, IMF
  - Inflation (CPI)
    - ObsMeanStd. Dev.Min.Max
    - 1,9840.10.3-0.110.6
    - Comment: Change of CPI (inflg = infl/infl[_n-1]-1). Multiply by 100 to get it in percentage change.
    - Source(s): International Financial Statistics, IMF
  - Output gap
    - Source(s): World Economic Outlook, IMF
  - Policy rate
    - Source(s): International Financial Statistics, IMF
  - Real interest rate
    - Comment: Policy rate or T-bill rate deflated by CPI.
    - Source(s): International Financial Statistics, IMF

- Governance indicators
  - Rule of law
    - ObsMeanStd. Dev.Min.Max
    - 1,6960.21.0-2.02.1
    - Comment: "... captures perceptions of the extent to which agents have confidence in and abide by the rules of society, and in particular the quality of contract enforcement, property rights, the police, and the courts, as well as the likelihood of crime and violence." It is a composite index relying on various sources.
    - Source(s): Worldwide Governance Indicators http://info.worldbank.org/governance/wgi/
  - Regulatory quality
    - ObsMeanStd. Dev.Min.Max
    - 1,6920.30.9-2.22.3
    - Comment: "...  captures perceptions of the ability of the government to formulate and implement sound policies and regulations that permit and promote private sector development." It is a composite index relying on various sources.
    - Source(s): Worldwide Governance Indicators http://info.worldbank.org/governance/wgi/
  - Insolvency framework
    - ObsMeanStd. Dev.Min.Max
    - 1,06542.826.60.0100.0
    - Comment: Time, cost and outcome of recovery of debt. Distance to frontier.
    - Source(s): Doing Business, World Bank
  - Contract enforcement (two rows shown)
    - ObsMeanStd. Dev.Min.Max
    - 1,06568.215.60.0100.0
    - Comment: Enforcing contracts, distance to frontier, measuring time and cost of resolving a dispute at first-instance court.
    - Source(s): Doing Business, World Bank
    - ObsMeanStd. Dev.Min.Max
    - 1,06560.513.020.893.4
    - Comment: Enforcing contracts, distance to frontier
    - Source(s): Doing Business, World Bank
  - Government effectiveness
    - ObsMeanStd. Dev.Min.Max
    - 1,6930.31.0-2.32.4
    - Comment: "... captures perceptions of the quality of public services, the quality of the civil service and the degree of its independence from political pressures, the quality of policy formulation and implementation, and the credibility of the government's commitment to such policies." It is a composite index relying on various sources.
    - Source(s): Worldwide Governance Indicators http://info.worldbank.org/governance/wgi/
  - Control of corruption
    - ObsMeanStd. Dev.Min.Max
    - 1,6960.21.0-1.92.6
    - Comment: "... captures perceptions of the extent to which public power is exercised for private gain, including both petty and grand forms of corruption, as well as "capture" of the state by elites and private interests." It is a composite index relying on various sources.
    - Source(s): Worldwide Governance Indicators http://info.worldbank.org/governance/wgi/
  - Ethics of private firms
    - ObsMeanStd. Dev.Min.Max
    - 9134.31.02.66.8
    - Comment: Ethical behavior of private firms.
    - Source(s): Global Competitiveness Index, World Economic Forum

- Other variables
  - International debt
    - ObsMeanStd. Dev.Min.Max
    - 1,36425.432.70.04247.0
    - Comment: "International Debt Securities (Amt Outstanding) as a share of GDP."
    - Source(s): Beck et al. (2017)
  - Cross-border bank loans
    - ObsMeanStd. Dev.Min.Max
    - 1,92219.628.40.03248.6
    - Comment: "International Debt Securities (Net Issues)  as a share of GDP."
    - Source(s): Beck et al. (2017)
  - Stock market capitalization
    - ObsMeanStd. Dev.Min.Max
    - 1,43848.856.90.01857.6
    - Comment: "Value of listed shares to GDP, calculated using the following deflation  method:  {(0.5)*[Ft/P_et + Ft-1/P_et-1]}/[GDPt/P_at] where F is stock market capitalization, P_e is end-of period CPI, and P_a  is average annual CPI."
    - Source(s): Beck et al. (2017)
  - Stock traded value
    - ObsMeanStd. Dev.Min.Max
    - 1,47626.843.90.0004331.3
    - Comment: "Total shares traded on the stock market exchange to GDP."
    - Source(s): Beck et al. (2017)
  - Stock market turnover
    - ObsMeanStd. Dev.Min.Max
    - 1,43149.783.60.011,732.3
    - Comment: "Ratio of the value of total shares traded to average real market capitalization, the denominator is deflated using the following method:  Tt/P_at/{(0.5)*[Mt/P_et + Mt-1/P_et-1] where T is total value traded, M is stock market capitalization, P_e is end-of period CPI P_a is average annual CPI."
    - Source(s): Beck et al. (2017)
  - Private bond issuance
    - ObsMeanStd. Dev.Min.Max
    - 74027.430.80.002197.1
    - Comment: "Private domestic debt securities issued by financial institutions and  corporations as a share of GDP, calculated using the following deflation method:  {(0.5)*[Ft/P_et + Ft-1/P_et-1]}/[GDPt/P_at] where F is amount outstanding of private domestic debt securities, P_e is end-of period  CPI, and P_a  is average annual CPI."
    - Source(s): Beck et al. (2017)
  - Private sector bank credit
    - ObsMeanStd. Dev.Min.Max
    - 1,94651.242.90.19253.5
    - Comment: "Private credit by deposit money banks and other financial institutions to GDP, calculated using the following deflation method:  {(0.5)*[Ft/P_et + Ft-1/P_et-1]}/[GDPt/P_at] where F is credit to the private sector, P_e is end-of period CPI, and P_a is average annual CPI."
    - Source(s): Beck et al. (2017)
  - Proxy for corporate income tax
    - ObsMeanStd. Dev.Min.Max
    - 1,00642.325.00.4280.2
    - Comment: Total tax rate on commercial profits.
    - Source(s): Doing Business, World Bank

*References, country lists, and data descriptions are reproduced exactly as they appear in the source document.*

### Appendix III. Correlation Coefficients

### Appendix III. Correlation Coefficients

### Correlation matrix: Main variables (selected pairwise correlations)
- Net interest margin (NIM) with:
  - Interest spread: 0.65
  - Return on Assets (ROA): 0.64
  - Operating costs: 0.74
  - Transaction size/financial depth: -0.80
  - Rule of law: -0.73
  - Regulatory quality: -0.56
- Interest spread with:
  - ROA: 0.45
  - Transaction size/financial depth: -0.50
  - Rule of law: -0.45
  - Regulatory quality: -0.25
- Return on Assets (ROA) with:
  - ROA_FSI: 0.84
  - Transaction size/financial depth: -0.58
  - Operating costs: 0.36
  - Rule of law: -0.48
  - Regulatory quality: -0.39
- ROA_FSI with:
  - Transaction size/financial depth: -0.57
  - Operating costs: 0.42
  - Rule of law: -0.55
  - Regulatory quality: -0.48
- Competition/concentration with:
  - Operating costs: -0.39
  - Transaction size/financial depth: 0.35
  - Rule of law: 0.53
  - Regulatory quality: 0.53
- Operating costs with:
  - Transaction size/financial depth: -0.77
  - Credit risk (Provisions): 0.43
  - Inflation (CPI): 0.41
  - Rule of law: -0.59
  - Regulatory quality: -0.52
- Transaction size/financial depth with:
  - Credit risk (Provisions): -0.47
  - Stock market capitalization: 0.56
  - Private sector bank credit: 0.92
- Risk-aversion (CAR) with:
  - Credit risk (Provisions): -0.10
  - Non-performing loans (NPL): 0.30
- Credit risk (Provisions) with:
  - Non-performing loans (NPL): -0.20
  - Real GDP growth: 0.25
  - Inflation (CPI): 0.24
- Real GDP growth with:
  - ROA: 0.33
  - ROA_FSI: 0.33
  - Inflation (CPI): 0.02
- Inflation (CPI) with:
  - ROA: 0.32
  - ROA_FSI: 0.35
  - Non-performing loans (NPL): 0.24
- Rule of law with:
  - Regulatory quality: 0.93
  - Insolvency framework: 0.58
  - Contract enforcement: 0.96
  - Control of corruption: 0.96
  - Ethics of private firms: 0.85
- Regulatory quality with:
  - Insolvency framework: 0.57
  - Contract enforcement: 0.88
  - Control of corruption: 0.90
  - Ethics of private firms: 0.77
- Insolvency framework with:
  - Contract enforcement: 0.13
  - Government effectiveness: 0.60
  - Control of corruption: 0.62
  - Ethics of private firms: 0.62
  - Private sector bank credit: 0.51
- Private sector bank credit with:
  - Transaction size/financial depth: 0.92
  - Stock traded value: 0.76
  - Private bond issuance: 0.61

(Note: the full correlation matrix in the source lists pairwise correlations for all variables in Table 9; the bullets above highlight selected pairwise correlations as reported.)

### Correlation matrix: Governance, debt and market indicators (Table 9 continued)
- Pairwise correlations among governance and market indicators (exact values as reported):
  - Insolvency framework — Contract enforcement: 0.13
  - Insolvency framework — Government effectiveness: 0.60
  - Insolvency framework — Control of corruption: 0.62
  - Insolvency framework — Ethics of private firms: 0.62
  - Insolvency framework — International debt: 0.32
  - Insolvency framework — Cross-border bank loans: 0.32
  - Insolvency framework — Stock market capitalization: 0.31
  - Insolvency framework — Stock traded value: 0.50
  - Insolvency framework — Stock market turnover: 0.28
  - Insolvency framework — Private bond issuance: 0.42
  - Insolvency framework — Private sector bank credit: 0.51
  - Contract enforcement — Government effectiveness: 0.35
  - Contract enforcement — Control of corruption: 0.29
  - Contract enforcement — Ethics of private firms: 0.15
  - Contract enforcement — International debt: 0.19
  - Contract enforcement — Cross-border bank loans: 0.19
  - Contract enforcement — Stock market capitalization: 0.02
  - Contract enforcement — Stock traded value: 0.24
  - Contract enforcement — Stock market turnover: 0.31
  - Contract enforcement — Private bond issuance: 0.23
  - Contract enforcement — Private sector bank credit: 0.33
  - Government effectiveness — Control of corruption: 0.94
  - Government effectiveness — Ethics of private firms: 0.89
  - Government effectiveness — International debt: 0.51
  - Government effectiveness — Cross-border bank loans: 0.51
  - Government effectiveness — Stock market capitalization: 0.58
  - Government effectiveness — Stock traded value: 0.66
  - Government effectiveness — Stock market turnover: 0.29
  - Government effectiveness — Private bond issuance: 0.42
  - Government effectiveness — Private sector bank credit: 0.74
  - Control of corruption — Ethics of private firms: 0.91
  - Control of corruption — International debt: 0.57
  - Control of corruption — Cross-border bank loans: 0.57
  - Control of corruption — Stock market capitalization: 0.55
  - Control of corruption — Stock traded value: 0.63
  - Control of corruption — Stock market turnover: 0.25
  - Control of corruption — Private bond issuance: 0.37
  - Control of corruption — Private sector bank credit: 0.69
  - Ethics of private firms — International debt: 0.52
  - Ethics of private firms — Cross-border bank loans: 0.52
  - Ethics of private firms — Stock market capitalization: 0.70
  - Ethics of private firms — Stock traded value: 0.72
  - Ethics of private firms — Stock market turnover: 0.26
  - Ethics of private firms — Private bond issuance: 0.47
  - Ethics of private firms — Private sector bank credit: 0.76
  - International debt — Cross-border bank loans: 1.00
  - International debt — Stock market capitalization: 0.37
  - International debt — Stock traded value: 0.49
  - International debt — Stock market turnover: 0.19
  - International debt — Private bond issuance: 0.17
  - International debt — Private sector bank credit: 0.57
  - Cross-border bank loans — Stock market capitalization: 0.37
  - Cross-border bank loans — Stock traded value: 0.49
  - Cross-border bank loans — Stock market turnover: 0.19
  - Cross-border bank loans — Private bond issuance: 0.18
  - Cross-border bank loans — Private sector bank credit: 0.56
  - Stock market capitalization — Stock traded value: 0.69
  - Stock market capitalization — Stock market turnover: 0.12
  - Stock market capitalization — Private bond issuance: 0.29
  - Stock market capitalization — Private sector bank credit: 0.62
  - Stock traded value — Stock market turnover: 0.65
  - Stock traded value — Private bond issuance: 0.39
  - Stock traded value — Private sector bank credit: 0.76
  - Stock market turnover — Private bond issuance: 0.38
  - Stock market turnover — Private sector bank credit: 0.49
  - Private bond issuance — Private sector bank credit: 0.61

### Appendix IV: Robustness checks (selected tables and reported estimates)
- Appendix IV lists robustness checks including:
  - Table 10. Robustness Check of Other Governance Indicators
  - Table 11. Robustness Check for Endogeneity Using Instrument Variables (Voice, accountability, political stability indicators as instruments and lagged explanatory variables)
  - Table 12. Robustness Check for Cross-Sectional Dependence
  - Table 13. Robustness Check for Non-Linearity (Probit model with a cut-off point for the dependent variable set at the 75th percentile)
  - Table 14. Robustness Check with Lagged Dependent Variable
  - Table 15. Robustness Check for Outliers
- Selected coefficients and standard errors from Table 15 (columns (1) to (5)) — exact reported values:
  - concentration:
    - (1): -0.00227 (0.00159)
    - (2): -0.00536*** (0.00164)
    - (3): -0.00131 (0.00203)
    - (4): -0.00633*** (0.00199)
    - (5): -0.00206 (0.00168)
  - operating cost:
    - (1): 0.795*** (0.0140)
    - (2): 0.801*** (0.0147)
    - (3): 0.812*** (0.0175)
    - (4): 0.803*** (0.0178)
    - (5): 0.819*** (0.0143)
  - transaction size:
    - (1): -0.334*** (0.0657)
    - (2): -0.495*** (0.0679)
    - (3): -0.461*** (0.0742)
    - (4): -0.629*** (0.0693)
    - (5): -0.421*** (0.0654)
  - risk aversion - CAR:
    - (1): 0.0338*** (0.00677)
    - (2): 0.0356*** (0.00715)
    - (3): 0.0355*** (0.00894)
    - (4): 0.0338*** (0.00905)
    - (5): 0.0337*** (0.00697)
  - credit risk - provisions:
    - (1): 0.00214*** (0.000537)
    - (2): 0.00249*** (0.000566)
    - (3): 0.000897 (0.000634)
    - (4): 0.00122* (0.000642)
    - (5): 0.00242*** (0.000554)
  - GDP growth:
    - (1): 4.812*** (0.986)
    - (2): 5.275*** (1.042)
    - (3): 5.849*** (1.220)
    - (4): 5.147*** (1.244)
    - (5): 5.227*** (1.012)
  - inflation:
    - (1): -0.136 (0.573)
    - (2): 0.380 (0.615)
    - (3): 2.609*** (0.800)
    - (4): 3.853*** (0.799)
    - (5): 0.242 (0.586)
  - rule of law:
    - (1): -0.484*** (0.0455)
  - regulatory quality:
    - (2): -0.312*** (0.0525)
  - insolvency framework:
    - (3): -0.00995*** (0.00178)
  - contract enforcement:
    - (4): -0.00828*** (0.00257)
  - control of corruption:
    - (5): -0.336*** (0.0409)
  - Observations (N):
    - (1): 1,102
    - (2): 1,102
    - (3): 835
    - (4): 835
    - (5): 1,102

*Source: Appendix III and Appendix IV tables as presented in the supplied content.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2018/wp18279.pdf_