## _wp12140

## Source details

**Canonical URL:** [_wp12140](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12140.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12140.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12140.pdf.json)

---

### I. INTRODUCTION
- Net interest margin (difference between lending and deposit rates) measures cost of bank intermediation and affects economic performance.
- Objective: examine influence of market concentration, bank regulations, institutional development, and macroeconomic environment on bank margins across LICs and EMs, controlling for bank-specific factors.
- Data: bank-level data on 359 commercial banks in 48 LICs, and 2535 commercial banks in 67 EMs for the period 1996-2010.
- Key empirical result summary:
  - Concentrated market structures and lack of competition in LICs remain key impediments to reducing financial intermediation costs.
  - Relaxing restrictions to bank entry could help reduce costs in LICs.
  - Low institutional capacity boosts margins.
  - Bank-specific characteristics explain substantial variation: margins increase with higher riskiness of credit portfolio, lower bank capitalization, and smaller bank size.

### II. INTEREST MARGIN DECOMPOSITION
- Analytical identity and decomposition (Equation (1)):
  - i_L - i_D = ρ * i * L/D + OC/D + Prov/D + ROA * A/D - NII/D + ε
- Interpretations of components:
  - Reserve requirement costs: ρ * i * L/D — high reserve requirements raise margins.
  - Operational costs: OC/D — higher operating costs imply higher margins (bank efficiency measure).
  - Loan loss provision costs: Prov/D — proxies credit risk and increases margin.
  - Profitability: ROA * A/D — banks may add profit margins if competition is weak.
  - Non-interest income: - NII/D — higher non-interest income allows lower interest margins.
- Decomposition results and empirical patterns:
  - Median spread in LICs is 11 percent, about 100 basis points higher than median for EMs.
  - Mann-Whitney test statistic of -6.36: difference in medians significant at the 1 percent confidence level.
  - Both 25th and 75th percentiles in LICs exceed those for EMs by approximately the same amount — distribution in LICs is shifted upward.
  - Key drivers of higher margins in LICs:
    - relatively higher profitability (suggesting weaker competition),
    - higher share of loan loss reserves (higher credit risk).
  - Median operating costs in LICs are slightly lower than in EMs.
  - LIC banks earn slightly higher non-interest revenues.
  - Cross-country median spreads:
    - EMs: range from about 5 percent (Tunisia) to 30 percent (Paraguay).
    - LICs: range from about 5 percent (Bangladesh) to more than 35 percent (Yemen).

### III. ECONOMETRIC FRAMEWORK
- Empirical specification (equation (2)):
  - Margin_it = α_i + Σ β_k B_{k, i, t-1} + Σ γ_m M_{m, j, t} + Σ δ_n I_{n, j, t} + Σ λ_l R_{l, j, t} + TE_t + ε_{i j t}
  - B = bank-specific determinants (lagged); M = macroeconomic; I = institutional; R = regulatory; α_i fixed effects; TE time dummies.
- Dependent variable:
  - Net interest margin = total interest revenues / interest bearing assets - total interest expenditures / interest bearing liabilities.
- Bank-specific determinants (B) included:
  - Market concentration: Herfindahl index (total assets).
  - Operating costs: operating expenses / total assets.
  - Risk aversion: equity-to-total assets ratio.
  - Credit risk: loan loss provisions / net loans.
  - Liquidity: liquid assets / total assets.
  - Size: log(total assets) or log(total loans) in robustness checks.
- Macroeconomic determinants (M): Real GDP growth rate; Inflation (percentage change of CPI).
- Institutional determinants (I): KKZ index; Rule of law; Control of corruption; Regulatory quality (Kaufmann et al., 2002).
- Regulatory determinants (R):
  - Fraction of entry applications denied (World Bank banking regulation surveys).
    - Average share of denied entries: LICs 71 percent, EMs 39 percent.
  - Activity restrictions: index (1 to 2.5) across securities, insurance, real estate, owning shares in non-financial firms.
  - Reserve requirements: dummy = 1 if reserve requirements exist, 0 otherwise.
    - Reserve requirements prevalence: LICs 57 percent, EMs 82 percent.

### IV. DESCRIPTIVE STATISTICS AND CORRELATIONS
- Selected medians and moments (Table 2 entries preserved):
  - Net interest margin: 14.7 10.9 14.6 15.4 9.9 21.1
  - Market concentration: 30.9 21.0 24.0 17.1 13.3 13.5
  - Operating costs: 5.0 4.3 5.3 9.0 4.3 18.8
  - Risk aversion: 15.0 10.2 63.9 16.8 12.2 30.2
  - Credit risk: 2.2 1.1 5.3 2.2 1.0 8.7
  - Liquidity: 33.2 27.7 27.8 30.9 25.4 42.4
  - Size of operations: 12.3 12.2 1.5 13.0 13.0 2.3
  - Real GDP growth: 5.3 5.6 3.5 4.8 5.1 4.8
  - Inflation: 9.3 7.1 21.0 7.8 5.5 16.8
  - KKZ index: -0.7 -0.6 0.4 -0.3 -0.4 0.5
  - Fraction of entries denied: 71.3 100.0 41.5 39.2 21.1 37.7
  - Activity restrictions: 1.3 1.3 0.4 1.3 1.0 0.3
  - Reserve requirements: 0.6 1.0 0.5 0.8 1.0 0.4
- Selected correlations (Table 3):
  - LICs: Net interest margin correlations: Market concentration 0.15; Operating costs 0.14; Credit risk 0.21; Liquidity 0.21; Size of operations -0.13; Real GDP growth 0.10; Inflation 0.04; KKZ index -0.06; Rule of law -0.10; Control of corruption -0.09.
  - EMs: Net interest margin correlations: Market concentration -0.01; Operating costs 0.19; Credit risk 0.17; Liquidity 0.30; Size of operations -0.18; Real GDP growth -0.18; Inflation 0.12; KKZ index 0.00; Rule of law -0.09; Control of corruption 0.02.

### V. BANK-SPECIFIC DETERMINANTS — ESTIMATION HIGHLIGHTS
- Market concentration:
  - Enters positively and significantly in both LIC and EM specifications when estimating bank-specific determinants (Table 4).
  - When time fixed effects included, impact becomes insignificant in EMs but remains significant in LICs.
  - Economic magnitude in LICs: a standard deviation increase in market concentration in LICs (23.6) raises the margin by 1.6 percentage point.
  - Example coefficients (LICs, Table 4): Market concentration 0.0698***; 0.0852***.
- Operating costs:
  - Positive association with margins; examples: 0.1400; 0.1206*** (LICs) and 0.2536**; 0.1303*** (EMs).
- Risk aversion (equity/asset):
  - Consistently negative: examples -0.0253***; -0.0656*** (LICs); -0.0355***; -0.0663*** (EMs).
- Credit risk:
  - Positive and significant: examples 0.3091***; 0.1276*** (LICs); 0.2864***; 0.1166*** (EMs).
- Liquidity:
  - Positive and significant: examples 0.0480***; 0.0264*** (LICs); 0.0649***; 0.0322*** (EMs).
- Size:
  - Negative and significant: examples -1.3495***; -2.2041*** (LICs); -0.7934*; -0.8595*** (EMs).
- Goodness of fit:
  - LICs: Observations 2,187; Number of banks 341; R-squared 0.105.
  - EMs: Observations 11,446; Number of banks 2,069; R-squared 0.036.

### VI. MACROECONOMIC DETERMINANTS
- LICs:
  - Macroeconomic variables have no significant impact on interest margins, neither individually nor jointly (Table 5, columns (1)-(3)).
- EMs:
  - Higher inflation has significant positive impact; higher output growth tends to reduce margins (Table 5, columns (4)-(6)).
  - Economic magnitudes in EMs:
    - A standard deviation increase in output growth (4.8) reduces the margin by 1.9 percentage points.
    - A standard deviation increase in inflation (16.8) raises the margin by 2.5 percentage points.
- Market concentration remains significant in LICs with macro controls; in EMs market concentration becomes insignificant once macroeconomic variables are included.

### VII. INSTITUTIONAL DETERMINANTS
- Institutional quality indicators (KKZ index, Rule of law, Control of corruption, Regulatory quality) show strong negative associations with margins in both LICs and EMs (Table 6).
- Examples:
  - LICs: KKZ index -9.5314*** [1.701]; Rule of law -3.0738*** [1.162]; Control of corruption -4.1745*** [1.062].
- Economic magnitudes for the composite KKZ index:
  - A standard deviation increase in the KKZ index reduces the margin by 4.1 percentage points in LICs.
  - A standard deviation increase in the KKZ index reduces the margin by 6.9 percentage points in EMs.
- Inclusion of institutional variables renders market concentration insignificant in both LICs and EMs in many specifications.
- Institutional characteristics diminish the impact of operational costs and bank size in many specifications.

### VIII. REGULATORY DETERMINANTS
- Fraction of entries denied:
  - Strongly and positively associated with margins in both LICs and EMs.
  - A standard deviation increase in fraction of entries denied raises the margin by 1.6 percentage points in LICs and by 1.9 percentage points in EMs.
  - Example coefficients (Table 7): Fraction of entries denied 0.0381***; 0.0503*** in LIC specifications.
- Activity restrictions on non-traditional banking activities:
  - Do not have significant impact on margins in LICs.
  - In EMs, restrictions have significant positive impacts with magnitudes e.g., 1.7 percentage points in some specifications.
- Reserve requirements:
  - Presence does not seem to affect interest margins in LICs in main specifications.
  - In EMs, presence of reserve requirements has significant positive impacts in some specifications with magnitude example 0.9 percentage points.

### IX. ROBUSTNESS CHECKS
- Replacing Herfindahl concentration index with bank-specific market share:
  - Market share yields highly significant positive coefficients; macro variables remain insignificant in LICs; institutional indices and fraction of entries denied remain positive and highly significant.
- Using loan- or deposit-based concentration measures:
  - Positive and significant coefficients for concentration hold across assets- or liabilities-based measures.
- Regional interactions:
  - Impact of market concentration on margins significantly larger in CIS relative to Western Hemisphere; Asia and MENA show marginally significant differences in some specifications; Africa interaction terms generally insignificant.
- Country-level regressions:
  - Confirm positive impact of market concentration; some environmental variables (inflation, reserve requirement ratio) appear significant with unexpected signs in country-level averages.
- Annual-average specifications:
  - Market concentration coefficients remain positive and often significant; risk aversion shows larger negative magnitudes; liquidity and credit risk remain positive.

### X. CONCLUSIONS AND POLICY IMPLICATIONS
- Evidence of significantly higher interest margins in LICs compared to EMs.
- Decomposition: higher median margin in LICs mainly explained by greater profitability of banks (possibly due to less competitive environment) and higher credit risk.
- Econometric findings:
  - Margins in LICs are more responsive to market structure (competition) than in EMs.
  - Margins in LICs are less responsive to macroeconomic and regulatory environment compared to EMs.
  - Institutional setting has a dominant impact on margins in both groups, often overshadowing market concentration and bank-specific controls.
  - Bank characteristics explain substantial within-country variation: margins increase with higher credit portfolio risk, lower bank capitalization, and smaller bank size.
- Policy recommendations implied by findings:
  - Foster banking competition in LICs to reduce interest margins.
  - Relax restrictions to bank entry to help lower intermediation costs.
  - Improve informational, contractual, and enforcement frameworks (broad national institutions) to play a key role in lowering interest margins in LICs.

*Source: Excerpt from IMF working paper _wp12140 (selected tables, figures, and text as presented in the source PDF).*

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

### References

### I. INTRODUCTION
- Net interest margin (difference between lending and deposit rates) measures cost of bank intermediation and affects economic performance (Jayaratne and Strahan, 1996; Rajan and Zingales, 1998; Beck, et al., 2000).
- Importance for low-income countries (LICs): firms depend largely on bank financing in absence of developed stock markets.
- Potential causes of persistently high margins in LICs:
  - lack of competition,
  - perceived market and credit risks,
  - bank unsoundness,
  - scale diseconomies from small markets,
  - high operating costs due to low efficiency,
  - unfavorable institutional environment,
  - regulatory constraints distorting financial market activity.
- Objective: examine influence of market concentration, bank regulations, institutional development, and macroeconomic environment on bank margins across LICs and emerging economies (EMs), controlling for bank-specific factors.
- Data: bank-level data on 359 commercial banks in 48 LICs, and 2535 commercial banks in 67 EMs for the period 1996-2010.
- Key empirical result summary:
  - Concentrated market structures and lack of competition in LICs remain key impediments to reducing financial intermediation costs.
  - Relaxing restrictions to bank entry could help reduce costs in LICs.
  - Low institutional capacity boosts margins.
  - Bank-specific characteristics explain substantial variation: margins increase with higher riskiness of credit portfolio, lower bank capitalization, and smaller bank size.
- Paper structure:
  - Section II: decomposes interest margin into cost and profit components.
  - Section III: behavioral econometric specification and determinants (bank-specific, macroeconomic, institutional, regulatory).
  - Section IV: robustness checks.
  - Last section: conclusions and policy recommendations.

### II. INTEREST MARGIN DECOMPOSITION
A. Conceptual framework
- Income identity leads to decomposition: II - IP = OC + Prov + P - NII.
- Dividing by interest bearing liabilities (D) and using average interest bearing assets (L) and total assets (A) yields decomposition; with reserve requirement ratio (ρ) assumption:
  - Equation (1): i_L - i_D = ρ * i * L/D + OC/D + Prov/D + ROA * A/D - NII/D + ε
  - ROA = P/A; ε arises from combining flow and stock data and reserve assumption.
- Decomposed components and interpretations:
  - Reserve requirement costs: ρ * i * L/D — high reserve requirements raise margins as banks pass costs to customers.
  - Operational costs: OC/D — higher operating costs imply higher margins; measures bank efficiency.
  - Loan loss provision costs: Prov/D — proxies credit risk and increases margin.
  - Profitability: ROA * A/D — banks may add profit margins, particularly if competition is weak.
  - Non-interest income: - NII/D — higher non-interest income allows lower interest margins (enters with negative sign).

B. Decomposition results and empirical patterns
- Data sources and sample cleaning:
  - Main source: BankScope database (Bureau van Dijk).
  - Sample: unbalanced panel of 359 commercial banks in 48 LICs and 2535 commercial banks in 67 EMs for 1996-2010.
  - Excluded countries with less than 10 observations and cleaned obvious errors/extreme observations (e.g., negative incomes, equity exceeding assets).
- Median implicit interest margins and distributional findings:
  - Median spread in LICs is 11 percent, about 100 basis points higher than median for EMs.
  - Mann-Whitney test statistic of -6.36: difference in medians significant at the 1 percent confidence level.
  - Both 25th and 75th percentiles in LICs exceed those for EMs by approximately the same amount — distribution in LICs is shifted upward.
  - Conclusion: financial intermediation costs in LICs exceed those in EMs and may impede financial deepening.
- Component comparisons (percentile distributions, Figure 2 observations):
  - Key driver of higher margins in LICs: relatively higher profitability (suggesting weaker competition).
  - Higher share of loan loss reserves in LICs indicates higher credit risk.
  - Median operating costs in LICs are slightly lower than in EMs — possible interpretations:
    - median LIC banks are no less efficient than EM counterparts; or
    - LIC banks possess higher market power and reduce operating costs (e.g., employee salaries) to secure higher profits.
  - LIC banks earn slightly higher non-interest revenues — per decomposition, this would not exert upward pressure on margins.
- Cross-country variation (Figure 3 observations):
  - EMs: median spread ranges from about 5 percent (Tunisia) to 30 percent (Paraguay).
  - LICs: median spread ranges from about 5 percent (Bangladesh) to more than 35 percent (Yemen).
  - Wide variation suggests important roles for country-specific factors (banking regulation, institutional and macroeconomic environment).

### III. ECONOMETRIC ANALYSIS OF BANK- AND COUNTRY-SPECIFIC DETERMINANTS OF INTEREST MARGINS
A. Model specification
- Augmented dealership model (theoretical base: Ho and Saunders (1981); extensions include Allen (1988), Angbazo (1997), Maudos and Fernandez de Guevara (2004)).
- Dealership model: bank as risk-averse dealer; determinants include market structure, operating costs, managerial risk aversion, credit risk, liquidity, size.
- Cross-country augmentation: include macroeconomic, institutional, and regulatory variables.
- Empirical specification (equation (2)):
  - Margin_it = α_i + Σ β_k B_{k, i, t-1} + Σ γ_m M_{m, j, t} + Σ δ_n I_{n, j, t} + Σ λ_l R_{l, j, t} + TE_t + ε_{i j t}
  - i = bank, j = country, t = time; B = bank-specific determinants (lagged), M = macroeconomic, I = institutional, R = regulatory; α_i fixed effects; TE time dummies; ε_ijt i.i.d.

B. Variables
- Dependent variable:
  - Net interest margin = total interest revenues / interest bearing assets - total interest expenditures / interest bearing liabilities.
- Bank-specific determinants (B):
  - Market concentration: Herfindahl index (sum of squares of individual bank market shares using total assets). Theoretical ambiguity: higher concentration could imply greater market power (higher margins) or efficiency-driven lower margins (efficient-structure hypothesis).
  - Operating costs: operating expenses / total assets — higher values expected to increase margins.
  - Risk aversion: equity-to-total assets ratio — ambiguous relationship with margins.
  - Credit risk: loan loss provisions / net loans — higher credit risk increases required margin.
  - Liquidity: liquid assets / total assets — more liquid banks may have higher margins to compensate opportunity cost.
  - Size: log(total assets) — larger banks expected to have lower margins if scale economies present.
- Macroeconomic determinants (M):
  - Economic activity: real GDP growth rate — margins expected to decline with higher growth.
  - Inflation: percentage change of CPI — higher inflation expected to widen margins.
- Institutional determinants (I):
  - KKZ index (Kaufmann et al., 2002): composite of voice and accountability, political stability, government effectiveness, regulatory burden, rule of law, freedom from graft — higher values expected to lower margins.
  - Rule of law (Kaufmann et al., 2002): higher values expected to lower margins.
  - Control of corruption (Kaufmann et al., 2002): higher values expected to lower margins.
  - Regulatory quality (Kaufmann et al., 2002): higher values expected to lower margins.
- Regulatory determinants (R):
  - Fraction of entry denied: fraction of bank entry applications denied from World Bank banking regulation surveys (1999, 2003, 2008).
    - Average share of denied entries: LICs 71 percent, EMs 39 percent — indicates lower competition in LICs.
  - Activity restrictions: average scale of restrictions across securities, insurance, real estate, owning shares in non-financial firms; scaling: unrestricted = 1, restricted = 2, prohibition = 3; indicator ranges from 1 to 2.5 — higher values expected to increase margins via limited scope economies.
  - Reserve requirements: dummy = 1 if reserve requirements exist, 0 otherwise.
    - Reserve requirements prevalence: LICs 57 percent, EMs 82 percent.
    - Reserve requirements typically non-remunerated or remunerated below market rates, imposing a tax passed to customers and boosting margins.
- Table 1 provides descriptions and sources for all variables (table not reproduced here).

C. Descriptive statistics (Table 2 summary points)
- Interest margins larger in median LICs than in median EMs.
- Banking industry characteristics:
  - Median Herfindahl index: LICs 21.4, EMs 13.7 (LICs more concentrated).
  - Capitalization (median equity-to-assets): LICs 10.2, EMs 12.2 (LICs lower capitalization).
  - LICs have smaller banks and higher credit risk (loan loss provisions measure) relative to EMs.
- (Full descriptive statistics reported in Table 2.)

*Source: Excerpt from the IMF working paper (pages provided) — analysis, data descriptions, and empirical framework as presented in the source document.*

### 1.0 in EMs), higher opportunity costs (liquidity ratio is 27.7 against 25.4 in EMs). All these

### _wp12140 - 1.0 in EMs), higher opportunity costs (liquidity ratio is 27.7 against 25.4 in EMs). All these

### Bank-specific determinants
- Larger banks tend to have lower margins; a standard deviation increase in log of total assets in LICs (1.4) reduces the margin by 1.9 percentage points, and in EMs (2.2) reduces the margin by 5.0 percentage points.
- A higher fraction of liquid assets boosts bank margins (banks compensate extra costs of holding liquid funds by charging higher margins).
- More risk averse/better capitalized banks require lower margin (reluctance to engage in more profitable but riskier lending).
- Higher credit risk is associated with larger margins (banks require higher profits to compensate for risk).
- Less efficient banks exhibiting larger operating costs charge higher margins.
- Market concentration:
  - Enters positively and significantly in both LIC and EM specifications (columns (1) and (2) in Table 4).
  - When time fixed effects are included (columns (3) and (4)), impact becomes insignificant in EMs but remains significant in LICs.
  - Economic magnitude in LICs: a standard deviation increase in market concentration in LICs (23.6) raises the margin by 1.6 percentage point.

### Macroeconomic determinants
- In LICs, macroeconomic variables have no significant impact on interest margins, neither individually nor jointly (columns (1)-(3) in Table 5).
- In EMs, higher inflation has a significant and positive impact on margins; higher output growth tends to reduce margins (columns (4)-(6) in Table 5).
- Economic magnitudes in EMs:
  - A standard deviation increase in output growth (4.8) reduces the margin by 1.9 percentage points.
  - A standard deviation increase in inflation (16.8) raises the margin by 2.5 percentage points.
- Market concentration remains significant in LICs with macro controls; in EMs, market concentration becomes insignificant once macroeconomic variables are included.
- Interpretation: macroeconomic fluctuations affect margins in EMs beyond bank-specific factors; in LICs macro risks appear already factored into bank-specific determinants.

### Institutional determinants
- Strong negative association between institutional quality and interest margins: composite KKZ index, rule of law, control of corruption, and regulatory quality all have negative and significant impacts on margins in both LICs and EMs.
- Inclusion of any institutional variable renders market concentration insignificant in both LICs and EMs.
- Economic magnitudes for the composite KKZ index:
  - A standard deviation increase in the KKZ index reduces the margin by 4.1 percentage points in LICs.
  - A standard deviation increase in the KKZ index reduces the margin by 6.9 percentage points in EMs.
- Example: if Ghana's composite KKZ index (0.06) increased to South Africa's level (0.41), margins in Ghana would be pushed down by 3.3 percentage points.
- Inclusion of institutional characteristics diminishes the impact of operational costs (bank efficiency) and bank size (scale effects); these become insignificant in both LICs and EMs.

### Regulatory determinants
- Fraction of bank entries denied is strongly and positively associated with margins in both LICs and EMs:
  - A standard deviation increase in fraction of entries denied raises the margin by 1.6 percentage points in LICs and by 1.9 percentage points in EMs.
- Restrictions on non-traditional banking activities (securities underwriting, insurance, real estate, ownership in non-financial firms) do not have a significant impact on bank interest margins in LICs.
- Presence of reserve requirements does not seem to affect interest margins in LICs.
- In EMs, restrictions on non-bank financial activities and presence of reserve requirements have significant positive impacts on margins with economic magnitudes of 1.7 and 0.9 percentage points, respectively.
- Introduction of regulatory determinants:
  - Washes away significance of market concentration in EMs specifications.
  - Market concentration remains significant in LICs specifications.
- Interpretation: regulatory constraints may drive market concentration in EMs; in LICs, market structure retains dominant explanatory power.

### Robustness checks
- Replacing concentration index with bank-specific market share (of assets) yields a highly significant positive coefficient for market share; macro variables remain insignificant; institutional indices and fraction of entries denied remain positive and highly significant.
- Significance of market share does not vanish in specifications with institutional variables (unlike market concentration), underscoring bank market power importance for margins.
- Using total loans and customer deposits to calculate market concentration (instead of total assets) produces positive and significant coefficients for concentration; finding holds across assets- or liabilities-based measures.
- Regional interactions (market concentration × region dummies for Asia, CIS, MENA, Sub-Saharan Africa; comparator Western Hemisphere):
  - Impact of market concentration on margins is significantly larger in CIS relative to Western Hemisphere.
  - Some specifications show a lightly lower impact for Asia and MENA, but coefficients are only marginally significant (10 percent confidence level).
  - Interaction terms for Africa are insignificant (impact comparable to Western Hemisphere).
- Country-level regressions (averaging bank-level data by country-year with country fixed effects) confirm positive impact of market concentration; some environmental variables (inflation, reserve requirement ratio) come out significant with unexpected signs.

### Conclusions and policy implications
- Evidence of significantly higher interest margins in LICs compared to EMs.
- Decomposition: higher median margin in LICs mainly explained by greater profitability of banks (possibly due to less competitive environment) and higher credit risk (possibly reflecting a weaker environment).
- Econometric findings:
  - Margins in LICs are more responsive to market structure (competition) than in EMs.
  - Margins in LICs are less responsive to macroeconomic and regulatory environment compared to EMs.
  - Institutional setting has a dominant impact on margins in both groups, often overshadowing market concentration and bank-specific controls.
  - Bank characteristics explain substantial within-country variation: margins increase with higher credit portfolio risk, lower bank capitalization, and smaller bank size.
- Policy recommendations implied by findings:
  - Foster banking competition in LICs to reduce interest margins.
  - Relax restrictions to bank entry to help lower intermediation costs.
  - Improve informational, contractual, and enforcement frameworks (broad national institutions) to play a key role in lowering interest margins in LICs.

*Source: IMF working paper section contained in the provided PDF content.*

### REFERENCES

### _wp12140 - REFERENCES

### Bibliographic references (selected)
- Allen, L. (1988), “The Determinants of Bank Interest Margins: A Note,” Journal of Financial and Quantitative Analysis, 23(2), pp. 231–35.
- Angbanzo, L. (1997), “Commercial Bank Net Interest Margins, Default Risk, Interest-Rate Risk and Off-Balance Sheet Banking,” Journal of Banking and Finance, 21, pp. 55–87.
- Barajas, A., Steiner, R., and Salazar, N. (2000), “The Impact of Liberalization and Foreign Investment in Colombia’s Financial Sector,” Journal of Development Economics, 63(1), pp. 157–96.
- Barth, J., Caprio, G., and Levine, R. (2001; 2004), works on bank regulation and supervision and bank regulation effectiveness.
- Beck, T., Demirguc‑Kunt, A., and Levine, R. (2000), “A New Database on Financial Development and Structure,” World Bank Economic Review (September, 2000).
- Brock, P., and Rojas‑Suarez, L. (2000), “Understanding the Behavior of Bank Spreads in Latin America,” Journal of Development Economics, 63(1), pp. 113–34.
- Claessens, S., Demirguc‑Kunt, A., and Huizinga, H. (2001), “How Does Foreign Entry Affect Domestic Banking Markets?” Journal of Banking and Finance, 25, pp. 891–911.
- Demirguc‑Kunt, A., Laeven, L., and Levine, R. (2004), “Regulations, Market Structure, Institutions, and the Cost of Financial Intermediation,” Journal of Money, Credit, and Banking, 36(3), pp. 593–622.
- Fungacova, Z., and Poghosyan, T. (2011), “Determinants of Bank Interest Margins in Russia: Does Bank Ownership Matter?” Economic Systems, 35, pp. 481–95.
- Ho, T., and Saunders, A. (1981), “The Determinants of Bank Interest Margins: Theory and Empirical Evidence,” Journal of Financial and Quantitative Analysis, XVI(4), pp. 581–600.
- Jayaratne, J., and Strahan, P. (1996), “The Finance-Growth Nexus: Evidence from Bank Branch Deregulation,” Quarterly Journal of Economics, 111, pp. 639–70.
- Kaufmann, D., Kraay, A., and Zoido‑Lobaton, P. (2002), “Governance Matters II: Updated Indicators for 2000/01,” World Bank Policy Research Working Paper No. 2772.
- Maudos, J., and Fernandez de Guevara, J. (2004), “Factors Explaining the Interest Margin in the Banking Sectors of the European Union,” Journal of Banking and Finance, 28, pp. 2259–281.
- Poghosyan, T. (2010), “Re-Examining the Impact of Foreign Bank Participation on Interest Margins in Emerging Markets,” Emerging Markets Review, 11(4), pp. 390–403.
- Rajan, R., and Zingales, L. (1998), “Financial Dependence and Growth,” American Economic Review 88, pp. 559–86.
- Randall, R. (1998), “Interest Rate Spreads in the Eastern Caribbean,” IMF Working Paper 98/59.

### Figures — distributions and comparisons
- Figure 1: Comparison of Implicit Net Interest Margins in LICs and EMs
  - Plot notes: Whiskers indicate minimum and maximum values; box edges denote 25th and 75th percentiles; line splitting box denotes median.
  - Legend entries: Spread, LIC, EM (plot “excludes outside values”).
- Figure 2: Percentile Distribution of Net Interest Margin Determinants in LICs and EMs
  - Panels shown for Operational costs, Loan loss provisions, Profitability, Non-interest income.
  - Each panel includes axes with factor scales (e.g., Operational costs factor 0 5 10 15 20 25; Loan loss reserves factor -20 - ? 24; Profitability factor -50 0 5 10; Non-interest income factor -10 -5 0 5 10 15) and legend: LIC, EM (excludes outside values).
- Figure 3: Median Interest Margins in LICs and EMs by Countries
  - Two panels labeled LICs and EMs with p 50 of SPREAD axis; country-level medians listed (e.g., LICs: YEMEN, GAMBIA, UGANDA, NIGERIA, ZAMBIA, ... BANGLADESH; EMs: PARAGUAY, JAMAICA, BRAZIL, LEBANON, VENEZUELA, ... TUNISIA).

### Table 1 — Variable definitions and sources (verbatim key variables)
- Net interest margin: Difference between the ratio of total interest revenues and interest bearing assets, and the ratio of total interest expenditures and interest bearing liabilities — BankScope.
- Market concentration: Herfindahl index (total assets) — BankScope.
- Operating costs: Ratio of total operating expenses to total assets — BankScope.
- Risk aversion: Ratio of total equity to total assets — BankScope.
- Credit risk: Ratio of loan loss provisions to total loans — BankScope.
- Liquidity: Ratio of liquid reserves to total assets — BankScope.
- Size of operations: Logarithm of total loans — BankScope.
- Real GDP growth: Real GDP growth rate — IMF, WEO.
- Inflation: Percentage change in consumer price index — IMF, WEO.
- KKZ index: Composite of country scores in the areas of voice and accountability, political stability, government effectiveness, regulatory burden, rule of law, and freedom from graft — Kaufmann et al. (2002).
- Rule of law; Control of corruption; Regulatory quality — Kaufmann et al. (2002).
- Fraction of entries denied; Activity restrictions; Reserve requirements — Barth et al. (2001, 2004).

### Table 2 — Descriptive statistics (as presented)
- Table row formatting (as in source):
  - Net interest margin14.710.914.615.49.921.1
  - Market concentration30.921.024.017.113.313.5
  - Operating costs5.04.35.39.04.318.8
  - Risk aversion15.010.263.916.812.230.2
  - Credit risk2.21.15.32.21.08.7
  - Liquidity33.227.727.830.925.442.4
  - Size of operations12.312.21.513.013.02.3
  - Real GDP growth5.35.63.54.85.14.8
  - Inflation9.37.121.07.85.516.8
  - KKZ index-0.7-0.60.4-0.3-0.40.5
  - Rule of law-0.7-0.80.4-0.4-0.60.6
  - Control of corruption-0.7-0.80.4-0.4-0.50.6
  - Regulatory quality-0.6-0.50.5-0.1-0.30.5
  - Fraction of entries denied71.3100.041.539.221.137.7
  - Activity restrictions1.31.30.41.31.00.3
  - Reserve requirements0.61.00.50.81.00.4
- Column labels indicate LICs and EMs; entries above preserve original numeric formatting from the source table.

### Correlations (Table 3) — selected relationships (preserve matrix entries exactly as presented)
- LICs: Net interest margin correlations with selected variables:
  - Market concentration 0.15, Operating costs 0.14, Credit risk 0.21, Liquidity 0.21, Size of operations -0.13, Real GDP growth 0.10, Inflation 0.04, KKZ index -0.06, Rule of law -0.10, Control of corruption -0.09.
- EMs: Net interest margin correlations with selected variables:
  - Market concentration -0.01, Operating costs 0.19, Credit risk 0.17, Liquidity 0.30, Size of operations -0.18, Real GDP growth -0.18, Inflation 0.12, KKZ index 0.00, Rule of law -0.09, Control of corruption 0.02.
- Full correlation matrices are presented in the source (LICs and EMs panels).

### Estimation results — Bank-specific determinants (Table 4, key coefficients)
- Dependent variable: net interest margin. Fixed effects estimator, explanatory variables lagged.
- LICs (columns reported):
  - Market concentration 0.0698*** (t/econ magnitude 1.6) ; 0.0852*** [0.017] [0.020]
  - Operating costs 0.1400 [0.095]; 0.1206*** [0.020]
  - Risk aversion -0.0253*** [0.007]; -0.0656*** [0.010]
  - Credit risk 0.3091*** [0.034]; 0.1276*** [0.018]
  - Liquidity 0.0480*** [0.013]; 0.0264*** [0.007]
  - Size -1.3495*** [0.291]; -2.2041*** [0.216]
  - Constant examples: 25.8169*** [4.006]; 41.3708*** [3.050]
  - Observations: 2,187 (LICs); Number of banks: 341; R-squared 0.105.
- EMs (columns reported):
  - Market concentration reported in alternative columns (e.g., 0.0626*** [0.018]; 0.0240 [0.022]).
  - Operating costs 0.2536** [0.098]; 0.1303*** [0.020].
  - Risk aversion -0.0355*** [0.008]; -0.0663*** [0.010].
  - Credit risk 0.2864*** [0.034]; 0.1166*** [0.018].
  - Liquidity 0.0649*** [0.013]; 0.0322*** [0.007].
  - Size -0.7934* [0.419]; -0.8595*** [0.299].
  - Observations: 11,446 (EMs); Number of banks: 2,069; R-squared 0.036.

### Estimation results — Controlling for macro variables (Table 5, key coefficients)
- LICs (specifications (1)-(3)):
  - Market concentration ~0.0755***, 0.0768***, 0.0769*** [0.019].
  - Operating costs 0.2588** to 0.2650*** [0.102–0.103].
  - Risk aversion -0.0354*** to -0.0358*** [0.008].
  - Credit risk 0.2897*** to 0.2908*** [0.035].
  - Liquidity 0.0626*** to 0.0631*** [0.014].
  - Size -0.7869* to -0.7993* [0.432–0.434].
  - GDP growth 0.0144; Inflation -0.0066.
  - Observations: 2,050; Number of banks: 317; R-squared 0.129.
- EMs (specifications (4)-(6)):
  - Market concentration coefficients near 0.0033 to -0.0258 (not significant).
  - Operating costs 0.1436***, 0.1170***, 0.1141*** [0.022].
  - Risk aversion -0.0933*** to -0.0800*** in some specs.
  - Credit risk 0.1184*** to 0.1017***.
  - Liquidity 0.0664*** to 0.0576***.
  - Size -1.8814*** to -1.0482***.
  - GDP growth -0.4806*** to -0.3890*** in some EM specs.
  - Inflation 0.1630*** to 0.1465*** in some EM specs.
  - Observations: 9,054–9,039; Number of banks: 1,230; R-squared up to 0.098.

### Estimation results — Institutional variables (Table 6, selected coefficients)
- LICs:
  - KKZ index -9.5314*** [1.701] (significant negative association).
  - Rule of law -3.0738*** [1.162].
  - Control of corruption -4.1745*** [1.062].
  - Regulatory quality -3.9 389*** and -1.8713** reported (source formatting preserved).
- EMs:
  - KKZ, rule of law, control of corruption, regulatory quality coefficients reported with negative and significant magnitudes (examples: KKZ and rule of law significance shown in EM columns).
- Credit risk and liquidity retain positive and significant coefficients in both LICs and EMs across specifications.
- Observations and R-squared vary by specification (LICs: Observations 1,397–1,684; EMs: Observations 7,403–9,019).

### Estimation results — Regulatory variables (Table 7, selected)
- Fraction of entries denied: 0.0381*** and 0.0503*** in LIC specifications ([0.010], [0.008]).
- Activity restrictions: coefficients reported (e.g., 1.1978; 4.9698*** [0.665] in an EM specification).
- Reserve requirements: example negative and positive coefficients reported (e.g., -0.3869; 2.4427***).
- Market concentration and operating costs often positive and significant in LIC specifications (e.g., Market concentration 0.0804*** [0.024]; Operating costs 0.2247 to 0.2579***).
- Observations: LICs 1,366–2,187; EMs 8,079–11,446.

### Robustness checks (Tables 8–12) — summary of robustness patterns
- Using Market Share instead of Market Concentration for LICs (Table 8):
  - Market share coefficients consistently positive and significant across many specifications (examples: 0.1172***, 0.1120***, 0.1404***).
  - Key bank-specific controls (risk aversion negative; credit risk positive; liquidity positive; size negative) remain broadly consistent.
- Using Loan Market Concentration for LICs (Table 9):
  - Market concentration (loans) positive and significant in many specifications (e.g., 0.0890***, 0.0795***).
  - Other signs and significance for operating costs, risk aversion, credit risk, liquidity, and size broadly consistent with main results.
- Using Deposit Market Concentration for LICs (Table 10):
  - Market concentration (deposits) positive and significant in many specifications (e.g., 0.0648***, 0.0536***).
- Interaction of Market Concentration with Regional Dummies (Table 11):
  - Market concentration main effect and interactions presented (examples: Market concentration 0.1511; Market concentration * Asia -0.2123; Market concentration * CIS 0.2739**).
  - Risk aversion, credit risk, liquidity maintain consistent signs.
- Using Annual Average Variables for LICs (Table 12):
  - Market concentration coefficients remain positive and often significant (e.g., 0.1353***, 0.1075***).
  - Risk aversion larger negative magnitudes in annual-average specifications (e.g., -0.0918*** to -0.1151***).
  - Liquidity, credit risk remain positive in many specifications.
  - Observations for annual-average models: e.g., Observations 523, 410, 337 depending on specification.

### Recurrent empirical patterns documented across tables
- Market concentration (or market share) is generally positively associated with higher net interest margins in LICs (multiple specifications show positive and often statistically significant coefficients).
- Operating costs are generally positively associated with net interest margins, with significance varying by specification and region.
- Risk aversion (equity/asset ratio) consistently has a negative coefficient (significant in many specifications), indicating higher capitalization is associated with lower margins.
- Credit risk (loan loss provisions/loans) consistently enters with a positive and significant coefficient.
- Liquidity (liquid reserves/total assets) consistently positive and often significant.
- Size (log loans) typically enters with a negative coefficient (often significant), indicating larger banks tend to have lower net interest margins.
- Institutional quality indicators (KKZ index, Rule of law, Control of corruption, Regulatory quality) display negative and in many cases statistically significant associations with net interest margins.
- Regulatory variables such as fraction of entries denied, activity restrictions, and reserve requirements show mixed but in some specifications significant positive associations with net interest margins.

*Italic line: Source: _wp12140 - REFERENCES (excerpt of figures, tables, and references as presented in the source PDF).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12140.pdf_