## _wp1334 - 7. Robustness Checks

## Source details

**Canonical URL:** [_wp1334 - 7. Robustness Checks](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1334.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1334.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1334.pdf.json)

---

### Data and sample
- Sample: 456 banks in 41 sub-Saharan African countries in 1995–2008, resulting in an unbalanced panel of 2,582 observations.
- Bank data source: Bankscope database; macroeconomic data: IFS database (IMF).
- Scope: commercial banks only.
- Net interest margins average 7.47% in the sample; regional averages vary from 6.58 % in Southern Africa to 7.85 % in East Africa.
- Net interest margin definition: total interest income minus total interest expense divided by average interest-earning assets.

### Robustness checks performed
- Alternative definitions for foreign and public ownership:
  - Used number of public banks and number of foreign banks in the market; results did not change (Table 7, columns 6 and 7).
  - Used a dummy for banks at least 25 percent foreign-owned; used total market share of foreign-owned banks; analogous alternatives for public ownership (25 percent threshold and market share). These alternative variables do not alter the results.
- Alternative proxies for credit risk:
  - Replaced loans/deposits and short-term funding with “net charge-offs” (Table 7, column 1): impact on interest margins remains positive and significant; model fit improves, but sample size drops substantially.
  - Used “impaired loans/equity” (Table 7, column 2): coefficient insignificant and resulting sample loss is large.
- Alternative market-size measures:
  - Market share and concentration also computed from total deposits and total loans; estimation results were similar.
- Legal-origin and regional tests:
  - Tested CFA franc zone behavior versus others; found no difference (Table 7, column 5).
  - Regional dummy for west and east Africa significant at a low level in some specifications (Table 7, column 4) but becomes insignificant when country fixed effects are included (Table 7, column 3).

### Key empirical findings (robust to many checks)
- Market structure and concentration:
  - Market concentration (HHI) translates into higher interest margins in SSA after controlling for bank-specific and macroeconomic variables.
  - The impact of concentration depends on bank efficiency: interaction of concentration and inefficiency is negative and significant, indicating that more efficient banks increase margins more in concentrated markets than inefficient banks (Table 6, column 3).
  - Bank market share is insignificant for interest margins in the sample; interaction of market share and inefficiency is also insignificant (Table 6).
  - Foreign and public ownership dummies are insignificant in most specifications and alternative definitions did not change this result (Table 6, columns 4–7; Table 7, columns 6–7).
- Bank-specific factors:
  - Liquidity risk: liquidity ratio (liquid assets/deposits and short-term funding) negatively and significantly affects interest margins (i.e., lower liquidity ratio → higher margins).
  - Credit risk: positively and significantly associated with net interest margins (robust across proxies where sample allows).
  - Equity (equity/total assets): positive and significant coefficient, consistent with a premium associated with solvency regulation or signaling/regulatory effects.
  - Inefficiency (overhead/average assets): positively associated with higher interest margins for banks of any size.
- Macroeconomic factors:
  - Inflation: positive and significant coefficient across specifications (Tables 6 and 7).
  - GDP growth: generally insignificant once country fixed effects are included; positive and marginally significant only when country fixed effects are not accounted for (Table 6, column 1).

### Regional differences
- Raw regional means (Table 4):
  - West: Obs 105, Mean 8.77, Std. Dev. 4.19
  - Central: Obs 16, Mean 6.32, Std. Dev. 6.46
  - East: Obs 78, Mean 7.85, Std. Dev. 4.68
  - South: Obs 68, Mean 6.58, Std. Dev. 5.05
- Mean-difference tests (Table 5):
  - Central - West: Difference -1.422, t value -2.702, * (Reject H0 at 1% level)
  - Central - East: Difference -1.530, t value -2.849, * (Reject H0 at 1% level)
  - Central - South: Difference -0.263, t value -0.482
  - South - West: Difference -1.159, t value -4.994, * (Reject H0 at 1% level)
  - South - East: Difference -1.267, t value -4.961, * (Reject H0 at 1% level)
  - West - East: Difference -0.108, t value -0.510
- After controls: regional dummy for west and east Africa is significant at a low level in some specifications, but becomes insignificant when country fixed effects are included (Table 7, columns 3–4).
- Conclusion: some regional differences persist in levels, but country-specific factors explain much of the variation when fixed effects are used.

### Selected summary statistics (Table 2)
- Net interest margin: Mean 7.47, Std. Dev. 4.94, Min -46.98, Max 49.94
- Equity/assets: Mean 13.41, Std. Dev. 11.80, Min -44.31, Max 99.68
- Credit risk: Mean 63.53, Std. Dev. 57.94, Min 0.03, Max 986.36
- Liquidity risk: Mean 53.08, Std. Dev. 48.23, Min 0.00, Max 897.62
- Market share: Mean 0.16, Std. Dev. 0.21, Min 0.00, Max 1.00
- Concentration: Mean 0.27, Std. Dev. 0.19, Min 0.08, Max 1.00
- Op. Inefficiency: Mean 8.00, Std. Dev. 5.78, Min -27.79, Max 79.09
- GDP growth: Mean 0.05, Std. Dev. 0.04, Min -0.18, Max 0.19
- Inflation: Mean 0.10, Std. Dev. 0.11, Min -0.08, Max 0.98
- Foreign: Mean 0.50, Std. Dev. 0.50, Min 0.00, Max 1.00
- Public: Mean 0.04, Std. Dev. 0.20, Min 0.00, Max 1.00

### Selected regression results (Tables 6 and 7)
- General model features:
  - Models (1) through (7) in Table 6 include country fixed effects and time dummies; Observations: 2,582 in each model; R-squared ranges: 0.340, 0.349.
  - Table 7 presents alternative specifications; Observations vary (e.g., 984, 1,417, 2,582); R-squared ranges: 0.200 to 0.466.
- Representative significant coefficients (with standard errors and significance from Tables 6 and 7):
  - Equity/assets: 0.046 [0.009] ***, ranges up to 0.074 [0.014] ***
  - Credit risk: 0.007 [0.002] ***
  - Liquidity risk: -0.010 [0.002] ***, ranges around -0.006 to -0.009 [0.002] ***
  - Op. inefficiency: 14.993 [2.091] ***, with larger magnitudes in some robustness models (e.g., 72.365 [5.454] ***)
  - Inflation: 3.644 [1.465] ***, with robustness values (e.g., 6.028 [1.672] ***, 4.441 [2.378] *)
  - GDP growth: example 2.985 [1.837] ***
  - Interaction (Inefficiency) x (Concentration): -67.996 [12.092] ***, with robustness examples -47.438 [22.981] ***, -180.324 [15.377] ***
  - Market share and concentration coefficients vary across specifications and sometimes change sign and significance.
  - Foreign and Public dummies: generally insignificant (examples: Foreign -0.371 [0.437]; Public 0.097 [1.130]).
  - Additional robustness controls: NCO average assets 0.039 [0.012] ***, Impaired loans/equity -0.001 [0.001].
- Country fixed effects and time dummies: included in most specifications; one specification lists Country fixed effects = NO.

### Interpretations and implications drawn in robustness section
- The positive concentration effect on margins may reflect non-competitive pricing in concentrated markets (SCP hypothesis), but the effect is heterogeneous across banks depending on efficiency.
- The insignificant role of foreign and public ownership in determining margins is robust to alternative definitions; potential explanations include: legal/regulatory environment, limited competitive pressure on foreign banks in SSA, or dataset coverage limitations.
- Use of alternative credit-risk measures supports the positive role of credit risk in margins but highlights trade-offs between better proxies and sample size/coverage.

*Source: _wp1334 - 7.  Robustness Checks (chapter section extracted from the provided PDF content).*

### References --------------------------------------------------------------------------------------------------  14

### _wp1334 - References --------------------------------------------------------------------------------------------------  14

### Figures
- 1.  Net Interest Margins: International Comparisons ---------------------------------------------- 4
- 2.  Market Concentration and Size of the Banking Sector ---------------------------------------- 5

### Tables
- 1.  Variables Definitions ---------------------------------------------------------------------------- 17
- 2.  Summary Statistics -------------------------------------------------------------------------------  17
- 3.  Net Interest Margins: Country Classifications ----------------------------------------------- 18
- 4.  Net Interest Margins by Region  --------------------------------------------------------------- 19
- 5.  Mean Difference Test ---------------------------------------------------------------------------- 19
- 6.  Net Interest Margins: Bank Characteristics and Macroeconomic Variables ------------- 20

*Source PDF filename: _wp1334 - References --------------------------------------------------------------------------------------------------  14; Canonical source URL: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1334.pdf*

### 7.  Robustness Checks ------------------------------------------------------------------------------- 21

### 7.  Robustness Checks

### Data and sample
- Sample: 456 banks in 41 sub-Saharan African countries in 1995–2008, resulting in an unbalanced panel of 2582 observations.
- Bank data source: Bankscope database; macroeconomic data: IFS database (IMF).
- Scope: commercial banks only.
- Net interest margins average 7.47% in the sample; regional averages vary from 6.58 % in Southern Africa to 7.85 % in East Africa.
- Net interest margin definition: total interest income minus total interest expense divided by average interest-earning assets.

### Robustness checks performed
- Alternative definitions for foreign and public ownership:
  - Used number of public banks and number of foreign banks in the market; results did not change (Table 7, columns 6 and 7).
  - Used a dummy for banks at least 25 percent foreign-owned; used total market share of foreign-owned banks; analogous alternatives for public ownership (25 percent threshold and market share). These alternative variables do not alter the results.
- Alternative proxies for credit risk:
  - Replaced loans/deposits and short-term funding with “net charge-offs” (Table 7, column 1): impact on interest margins remains positive and significant; model fit improves, but sample size drops substantially.
  - Used “impaired loans/equity” (Table 7, column 2): coefficient insignificant and resulting sample loss is large.
- Alternative market-size measures:
  - Market share and concentration also computed from total deposits and total loans; estimation results were similar.
- Legal-origin and regional tests:
  - Tested CFA franc zone behavior versus others; found no difference (Table 7, column 5).
  - Regional dummy for west and east Africa significant at a low level in some specifications (Table 7, column 4) but becomes insignificant when country fixed effects are included (Table 7, column 3).

### Key empirical findings (robust to many checks)
- Market structure and concentration:
  - Market concentration (HHI) translates into higher interest margins in SSA after controlling for bank-specific and macroeconomic variables.
  - The impact of concentration depends on bank efficiency: interaction of concentration and inefficiency is negative and significant, indicating that more efficient banks increase margins more in concentrated markets than inefficient banks (Table 6, column 3).
  - Bank market share is insignificant for interest margins in the sample; interaction of market share and inefficiency is also insignificant (Table 6).
  - Foreign and public ownership dummies are insignificant in most specifications and alternative definitions did not change this result (Table 6, columns 4–7; Table 7, columns 6–7).
- Bank-specific factors:
  - Liquidity risk: liquidity ratio (liquid assets/deposits and short-term funding) negatively and significantly affects interest margins (i.e., lower liquidity ratio → higher margins).
  - Credit risk: positively and significantly associated with net interest margins (robust across proxies where sample allows).
  - Equity (equity/total assets): positive and significant coefficient, consistent with a premium associated with solvency regulation or signaling/regulatory effects.
  - Inefficiency (overhead/average assets): positively associated with higher interest margins for banks of any size.
- Macroeconomic factors:
  - Inflation: positive and significant coefficient across specifications (Tables 6 and 7).
  - GDP growth: generally insignificant once country fixed effects are included; positive and marginally significant only when country fixed effects are not accounted for (Table 6, column 1).

### Regional differences
- Raw regional means: interest margins lower in central and southern Africa compared to west and east Africa (Tables 4 and 5).
- After controls: regional dummy for west and east Africa is significant at a low level in some specifications, but becomes insignificant when country fixed effects are included (Table 7, columns 3–4).
- Conclusion: some regional differences persist in levels, but country-specific factors explain much of the variation when fixed effects are used.

### Interpretations and implications drawn in robustness section
- The positive concentration effect on margins may reflect non-competitive pricing in concentrated markets (SCP hypothesis), but the effect is heterogenous across banks depending on efficiency.
- The insignificant role of foreign and public ownership in determining margins is robust to alternative definitions; potential explanations include: legal/regulatory environment, limited competitive pressure on foreign banks in SSA, or dataset coverage limitations.
- Use of alternative credit-risk measures supports the positive role of credit risk in margins but highlights trade-offs between better proxies and sample size/coverage.

*Source: _wp1334 - 7.  Robustness Checks (chapter section extracted from the provided PDF content).*

### REFERENCES

### _wp1334 - REFERENCES

### References
- Angbazo, L., 1997, “Commercial Bank Net Interest Margins, Default Risk, Interest Rate Risk, and Off-Balance Sheet Banking,” Journal of Banking & Finance, 21, pp. 55–87.
- Beck, T., and H. Hess, 2009, “Why Are Interest Spreads So High in Uganda?” Journal of Development Economics, 88, pp. 192–204.
- Berger, A., 1995, “The Relationship Between Capital and Earnings in Banking,” Journal of Money, Credit and Banking, 27, pp. 432–456.
- Berlin, M., and L.J. Mester, 1999, “Deposits and Relationship Lending,” The Review of Financial Studies, 12, pp. 579–607.
- Boot, A.W.A, 2000, “Relationship Banking: What Do We Know?” Journal of Financial Intermediation, 9, pp. 7–25.
- Brock, P., and L. Rojas Suarez, 2000, “Understanding the Behavior of Bank Spread in Latin America,” Journal of Development Economics, 63, pp. 113–134.
- Carbó Valverde, S., and F. Rodríguez Fernández, 2007, “The Determinants of Bank Margins in European Banking,” Journal of Banking and Finance, 31,  2043–2063.
- Cetorelli, N., and M. Gambera, 2001, “Banking Market Structure, Financial Dependence and Growth: International Evidence from Industry Data,” The Journal of Finance, Vol. LVI, No. 2, April.
- Claeys, S., and R. Vander Vennet, 2008, “Determinants of Bank Interest Margins in Central and Eastern Europe: A Comparison with the West,” Economic Systems, 32, pp. 197–216.
- Dabla-Norris, E., and H. Floerkemeier, 2007, “Bank Efficiency and Market Structure: What Determines Banking Spreads in Armenia?” IMF Working Paper No. 07/134 (Washington: International Monetary Fund).
- Demirgüç-Kunt, A., L. Leaven, and R. Levine, 2003, “Regulations, Market Structure, Institutions, and the Cost of Financial Intermediation,” NBER Working Paper No. 9890 (Washington: National Bureau of Economic Research).
- Demirgüç-Kunt, A., and H. Huizinga, 1999, “Determinants of Commercial Bank Interest Margins and Profitability: Some International Evidence,” World Bank Economic Review, Vol. 13, No. 2, pp. 379–408.
- Flamini, V., C. McDonald, and L. Schumacher, 2009, “The Determinants of Commercial Bank Profitability in Sub-Saharan Africa,” IMF Working Paper No. 09/15 (Washington: International Monetary Fund).
- Gelos, R.G., 2006, “Banking Spreads in Latin America,” IMF Working Paper No. 06/44 (Washington: International Monetary Fund).
- Ho, T., and A. Saunders, 1981, “The Determinants of Bank Interest Margins: Theory and Empirical Evidence,” Journal of Financial and Quantitative Analysis, 16, pp. 581–600.
- Levine, R., and S. Zervos, 1998, “Stock Markets, Banks, and Economic Growth,” American Economic Review, 88, pp. 537–558.
- Mlachila, M., and E.W. Chirwa, 2002, “Financial Reforms and Interest Rate Spreads in the Commercial Banking System in Malawi,” IMF Working Paper No. 02/6 (Washington: International Monetary Fund).
- Peterson, M.A., and R. Rajan, 1995, “The effect of Credit Market Competition on Lending Relationships,” The Quarterly Journal of Economics, 42, pp. 407–444.
- Saunders, A., and L. Schumacher, 2000, “The Determinants of Bank Interest Rate Margins: An International Study,” Journal of International Money and Finance, 19, pp. 813–832.

### Variable definitions (Table 1)
- Net interest margin: Difference between total interest income and expenses over average assets
- Equity: Equity/total assets
- Credit risk: Loans/deposits and short-term funding
- Liquidity risk: Liquid assets/short-term funding
- Operational inefficiency: Overheads/average assets
- Market share: Bank assets /total banks assets in economy, represents market power
- Concentration: Market concentration, measured by HHI index of total asset (sum of squared assets/total asset market shares of banks)
- Output growth: GDP growth rate
- Inflation: CPI growth rate
- Public: Dummy variable for publicly owned banks (at least 25% stake by the state)
- Foreign: Dummy variable for foreign-owned banks (at least 25% stake by foreign investors)
- Source: Bankscope, IFS and author's definitions

### Summary statistics (Table 2)
- Net interest margin: Mean 7.47, Std. Dev. 4.94, Min -46.98, Max 49.94
- Equity/assets: Mean 13.41, Std. Dev. 11.80, Min -44.31, Max 99.68
- Credit risk: Mean 63.53, Std. Dev. 57.94, Min 0.03, Max 986.36
- Liquidity risk: Mean 53.08, Std. Dev. 48.23, Min 0.00, Max 897.62
- Market share: Mean 0.16, Std. Dev. 0.21, Min 0.00, Max 1.00
- Concentration: Mean 0.27, Std. Dev. 0.19, Min 0.08, Max 1.00
- Op. Inefficiency: Mean 8.00, Std. Dev. 5.78, Min -27.79, Max 79.09
- GDP growth: Mean 0.05, Std. Dev. 0.04, Min -0.18, Max 0.19
- Inflation: Mean 0.10, Std. Dev. 0.11, Min -0.08, Max 0.98
- Foreign: Mean 0.50, Std. Dev. 0.50, Min 0.00, Max 1.00
- Public: Mean 0.04, Std. Dev. 0.20, Min 0.00, Max 1.00
- Source: Bankscope, IFS and author's calculations

### Country classifications (Table 3)
- Angola — Region: South — CFA Zone**: 0
- Benin — Region: West — CFA Zone**: 1
- Botswana — Region: South — CFA Zone**: 0
- Burkina Faso — Region: West — CFA Zone**: 1
- Burundi — Region: Central — CFA Zone**: 0
- Cameroon — Region: Central — CFA Zone**: 1
- Cape Verde — Region: West — CFA Zone**: 0
- Central African Republic — Region: Central — CFA Zone**: 1
- Chad — Region: Central — CFA Zone**: 1
- Congo — Region: Central — CFA Zone**: 1
- Congo, Democratic Rep. of — Region: Central — CFA Zone**: 0
- Equatorial Guinea, Rep. of — Region: Central — CFA Zone**: 1
- Eritrea — Region: East — CFA Zone**: 0
- Ethiopia — Region: East — CFA Zone**: 0
- Gabon — Region: Central — CFA Zone**: 1
- Gambia — Region: West — CFA Zone**: 0
- Ghana — Region: West — CFA Zone**: 0
- Guinea — Region: West — CFA Zone**: 0
- Cote d'Ivoire — Region: West — CFA Zone**: 1
- Kenya — Region: East — CFA Zone**: 0
- Lesotho — Region: South — CFA Zone**: 0
- Liberia — Region: West — CFA Zone**: 0
- Madagascar — Region: South — CFA Zone**: 0
- Malawi — Region: South — CFA Zone**: 0
- Mali — Region: West — CFA Zone**: 1
- Mauritius — Region: South — CFA Zone**: 0
- Mozambique — Region: South — CFA Zone**: 0
- Namibia — Region: South — CFA Zone**: 0
- Niger — Region: West — CFA Zone**: 1
- Nigeria — Region: West — CFA Zone**: 0
- Rwanda — Region: East — CFA Zone**: 0
- Sao Tome & Principe — Region: Central — CFA Zone**: 0
- Senegal — Region: West — CFA Zone**: 1
- Seychelles — Region: South — CFA Zone**: 0
- Sierra Leone — Region: West — CFA Zone**: 0
- South Africa — Region: South — CFA Zone**: 0
- Swaziland — Region: South — CFA Zone**: 0
- Tanzania — Region: East — CFA Zone**: 0
- Togo — Region: West — CFA Zone**: 1
- Uganda — Region: East — CFA Zone**: 0
- Zambia — Region: South — CFA Zone**: 0
- Source: Author's classifications.

### Net Interest Margin by Region (Table 4)
- Regions: Obs, Mean, Std. Dev.
  - West: 105, 8.77, 4.19
  - Central: 16, 6.32, 6.46
  - East: 78, 7.85, 4.68
  - South: 68, 6.58, 5.05

### Mean difference tests (Table 5)
- Hypotheses: H0: Mean Region 1 = Mean Region 2; H1: Mean Region 1 - Mean Region 2 < 0
- Central - West: Difference -1.422, t value -2.702, * (Reject H0 at 1% level)
- Central - East: Difference -1.530, t value -2.849, * (Reject H0 at 1% level)
- Central - South: Difference -0.263, t value -0.482
- South - West: Difference -1.159, t value -4.994, * (Reject H0 at 1% level)
- South - East: Difference -1.267, t value -4.961, * (Reject H0 at 1% level)
- West - East: Difference -0.108, t value -0.510

### Regression results — Net Interest Margin: Bank Characteristics and Macroeconomic Variables (Table 6)
- Models (1) through (7) include country fixed effects and time dummies; Observations: 2,582 in each model; R-squared ranges: 0.340, 0.349.
- Coefficients (selected, with standard errors in brackets and significance levels indicated by ***, **, *):
  - Equity/assets: 0.046 [0.009] ***
  - Credit risk: 0.007 [0.002] ***
  - Liquidity risk: -0.010 [0.002] ***
  - Market share: coefficients vary across models (e.g., 0.611 [0.760]; 0.040 [1.043]; 0.333 [0.757])
  - Concentration: coefficients vary (e.g., -1.272 [0.828]; 2.273 [1.037] ***)
  - Op. inefficiency: 14.993 [2.091] ***
  - GDP growth: 2.985 [1.837] ***
  - Inflation: 3.644 [1.465] ***
  - Foreign: -0.371 [0.437] (models where included)
  - Public: 0.097 [1.130] (models where included)
  - Interaction terms:
    - (Inefficiency) x (Market share): 12.333 [15.460]
    - (Inefficiency) x (Concentration): -67.996 [12.092] ***
    - Concentration x Public: 0.179 [2.389]
    - Concentration x Foreign: 0.319 [1.091]
  - Constant values vary across models (examples: -0.660 [3.294]; 3.732 [4.104])

### Robustness checks (Table 7)
- Models present alternative specifications; Observations vary (e.g., 984, 1,417, 2,582); R-squared ranges: 0.200 to 0.466.
- Selected coefficients (with standard errors and significance):
  - Equity/assets: ranges from 0.074 [0.014] *** to 0.045 [0.009] ***
  - Credit risk: 0.006 [0.002] ***
  - Liquidity risk: ranges around -0.006 to -0.009 [0.002] ***
  - Market share: coefficients vary widely (examples: 1.734 [1.101]; 1.904 [0.974]; 0.333 [0.757])
  - Concentration: mixed signs and significance across specifications
  - Op. inefficiency: large positive coefficients (examples: 72.365 [5.454] ***; 33.372 [3.871] ***)
  - (Inefficiency) x (Concentration): examples include -47.438 [22.981] ***, -180.324 [15.377] ***
  - GDP growth: examples include -0.720 [3.115], -3.056 [2.293], 2.867 [1.827]
  - Inflation: examples include 4.441 [2.378] *, 6.028 [1.672] ***, 3.725 [1.456] ***
  - Additional controls in some models:
    - NCO average assets: 0.039 [0.012] ***
    - Impaired loans/equity: -0.001 [0.001]
    - Dummy West and East Africa: 2.497 [4.257]; 0.774 [0.432] *
    - Dummy for CFAF zone: 3.635 [3.468]
    - Number of public banks: 0.103 [0.111]
    - Number of foreign banks: 0.003 [0.027]
- Country fixed effects and time dummies: included in most specifications; one specification lists Country fixed effects = NO.
- Observations reported by model: 984, 1,417, 2,582, 2,582, 2,582, 2,582.

*Source: _wp1334 - REFERENCES (PDF content).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1334.pdf_