## _wp0816

## Source details

**Canonical URL:** [_wp0816](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp0816.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp0816.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp0816.pdf.json)

---

### I. Introduction
- Islamic financial institutions growth:
  - From one institution in one country in 1975 to over 300 institutions operating in more than 75 countries (El Qorchi, 2005).
  - In Sudan and Iran, the entire banking system is currently based on Islamic finance principles.
  - Total assets of Islamic banks worldwide are estimated at about $250 billion, and are expected to grow by about 15 percent a year (Choong and Liu, 2006; Ainley and others, 2007).
- Contribution of the paper:
  - Claims to be the first cross-country empirical analysis of the role of Islamic banks in financial stability.
- Paper structure (as presented):
  - Section II: specifics of Islamic banking from a prudential perspective (see Appendix I).
  - Section III: methodology, variables, and data (Appendix II).
  - Section IV: empirical results.
  - Section V: conclusions and topics for further research.

### II. Specifics of Islamic Banking from a Prudential Perspective
- Key definitions and contract features:
  - Islamic or Shari’ah-compliant banking: provision and use of financial services and products that conform to Islamic religious practices and laws.
  - Prohibition against payment and receipt of interest at a fixed or predetermined rate.
  - Contracts rely on profit-and-loss sharing arrangements (PLS), purchase and resale of goods and services, and provision of services for fees.
  - In PLS modes, the rate of return is not known or fixed prior to undertaking the transaction.
  - In purchase-resale transactions, a mark-up is determined based on a benchmark rate of return, typically a return determined in international markets such as LIBOR.
  - Range of contracts available depending on investor rights in project management and timing of cash flows.
  - Islamic banks generally prohibited from trading in financial risk, and from financing production/trade in alcoholic beverages or pork, non-Islamic media, and gambling operations.
- Potential risk-increasing features (literature highlights):
  - PLS financing: shifts direct credit risk from banks to investment depositors but increases overall asset-side risk as banks become vulnerable to risks normally borne by equity investors.
  - Operational risk: increased by administration of PLS modes; includes legal risk and Sharī`ah compliance risk.
  - PLS cannot be made dependent on collateral or guarantees.
  - Product standardization is more difficult, increasing operational risk and legal contract uncertainty.
  - Fewer risk-hedging instruments historically and underdeveloped interbank and money markets; lender-of-last-resort access limited in some environments (though recent Islamic money market developments and implicit liquidity support have reduced these differences).
  - Non-PLS modes resemble conventional financing but can still carry elevated operational risk in some cases.
  - Investment deposits: capital value and rate of return are not guaranteed, potentially increasing moral hazard and incentives for risk taking and undercapitalization.
- Potential risk-mitigating features:
  - Ability to pass through negative asset shocks (e.g., Musharaka loss) to investment depositors (Mudaraba arrangement).
  - Risk-sharing on deposit side provides additional protection beyond book capital.
  - Shareholders’ responsibility for negligence or misconduct and more difficult access to liquidity can create pressures to be more conservative.
  - Investors typically do not have deposit insurance and thus have stronger incentives to exercise oversight over management.
  - Islamic banks traditionally hold a comparatively larger proportion of assets in reserve accounts with central banks or in correspondent accounts, providing buffers.
- Conclusion from Section II:
  - Whether Islamic banks are more or less stable than conventional banks is an empirical question that may differ across countries and banks; the remainder of the paper provides empirical evidence.

### III. Methodology and Data — Measuring Bank Stability
- Primary dependent variable:
  - Z-score defined as z ≡ (k + μ) / σ
    - k: equity capital and reserves as percent of assets
    - μ: average return as percent of assets
    - σ: standard deviation of return on assets (proxy for return volatility)
  - Interpretation: the number of standard deviations a return must fall to deplete equity under normality assumption; higher z-score implies lower probability of insolvency.
- Applicability and critique for Islamic banks:
  - Criticism: investment accounts may act like equity, so book capital and reserves may understate financial strength.
    - Investment accounts often linked to pre-agreed maturity (from one month upwards) and can generally be withdrawn with one month’s advance notice.
    - Example profit-sharing: 80 percent to depositors and 20 percent to the bank.
    - In extreme cases, a bank with only restricted investment accounts could resemble a mutual fund.
  - Counterarguments: conventional banks can also pass risks to customers; additional protection layers are reflected in banks’ returns and capital and thus in z-scores; short withdrawal features and pre-determined profit-sharing reduce factual differences.
  - Authors’ stance: capital and reserves are a reasonable proxy to assess default risk; use of z-score is justified.
- Preliminary tests:
  - Pairwise comparisons of z-scores between Islamic and commercial banks.
  - Subdivision by size with cut-off at total assets of US$ 1 billion between small and large banks.

### III.B Regression Analysis — Model and Key Variables
- General panel model structure (notation preserved as in source):
  - Dependent variable: z_{tji} for bank i in country j at time t.
  - Model includes bank-specific lagged variables B_{t-1,ji}, time-varying industry-specific variables I_{t-1,j}, bank type interactions T_s and I_{t-1,js}, macroeconomic variables M_{t,j}, country dummies C_j, year dummies D_t, and residual ε_{tji}.
- Key variables of interest:
  - Islamic bank dummy: equals 1 if bank is Islamic, 0 if commercial. Negative coefficient implies Islamic banks are relatively weaker (lower z-scores).
  - Market share of Islamic banks by assets: calculated for each year and country and interacted with Islamic and commercial bank dummies to test systemic-level impact.
    - A negative sign for interaction of Islamic banks’ market share and Islamic bank dummy indicates that a higher share of Islamic banks reduces their soundness (reduces z-scores).
- Bank-level controls:
  - Asset size in U.S. dollars billion.
  - Loans over assets.
  - Cost-income ratio.
  - Income diversity measure following Laeven and Levine (2005); income diversity for Islamic banks defines net interest income as the sum of positive and negative income flows associated with PLS arrangements (International Monetary Fund, 2004).
  - Interaction of income diversity with Islamic bank dummy to capture differences in business orientation.
- Rationale: controls account for differences in size, asset composition, cost efficiency, and income structure between Islamic and other banks.

### Variable Definitions, Lag Structure, and Estimation
- Income diversity measure:
  - Defined as (incomeoperatingOther − Net interest income) / incomeoperatingTotal − 1. Higher values correspond to a higher degree of diversification.
- Country-level controls:
  - GDP growth rate, inflation rate, and exchange rate depreciation included (same value for all banks in a given country).
  - Herfindahl index ranges from 0 to 10,000 (for a system with only one bank).
  - Governance indicator: average of 6 measures (voice and accountability, political stability, government effectiveness, regulatory quality, rule of law, control of corruption) averaged across years 2004, 2002, 2000, 1998, and 1996 into a single index per country.
- Lagging and robustness:
  - All bank-specific and macroeconomic variables, the Herfindahl index, the Islamic banks’ market share and its interactions are lagged; robustness check restricts explanatory variables to contemporaneous effects.
- Estimation methods:
  - Primary: pooled ordinary least squares (OLS).
  - Robust estimation: Hamilton (2002) technique to downweight observations with large residuals; standard errors via pseudovalues approach (Street, Carroll, and Ruppert, 1988).
  - Additional methods: fixed effects regressions and median least squares regressions.

### Data and Sample Coverage
- Data source: BankScope database.
- Jurisdictions included (initial sample of 20): Bahrain, Bangladesh, Brunei, Egypt, Gambia, Indonesia, Iran, Jordan, Kuwait, Lebanon, Malaysia, Mauritania, Pakistan, Qatar, Saudi Arabia, Sudan, Tunisia, United Arab Emirates, West Bank and Gaza, and Yemen.
- Observations and sample period:
  - Up to 520 observations for 77 Islamic banks.
  - 3,248 observations for 397 commercial banks.
  - Sample period: 1993 to 2004.
- Sample selection and coverage notes:
  - Systems where Islamic banks accounted for more than 1 percent of total assets in at least one year in 1993–2004.
  - Islamic banks covered in starting sample have total assets of US$253 billion as of 2004 (consistent with earlier estimate of “about US$250 billion”).
  - About 49 percent of Islamic banks and about 62 percent of commercial banks fall into the large bank category (assets over US$ 1 billion).
  - BankScope coverage varies by country; for most countries in sample BankScope covers 80–90 percent of banking systems in terms of total assets.
  - Lebanon and Kuwait excluded from regression analysis due to lack of unconsolidated observations for Islamic banks, reducing country coverage from 20 to 18.
  - Consolidated data used when available; otherwise unconsolidated data used.
  - Cross-checks performed against Institute of Islamic Banking and Insurance and IBF.net.
  - Data-definition issues exist (e.g., capital definition, measuring equivalent of interest income); paper relied largely on BankScope definitions with cross-checking and outlier exclusion.

### Descriptive Statistics and Z-score Overview (selected exact figures from Table 2)
- Z-score ranges and outliers:
  - z-score ranges from -81 to 203,347, indicating presence of outliers; robust estimation methods used.
  - Outlier treatment: averages shown for full sample and excluding 1st and 99th percentiles.
- Table 2 — Averages across banks, 1993–2004 (preserved exactly as in source):
  - All Banks:
    - Z-Score: Commercial 132.9; Islamic 27.2
    - Z-Score excl. outliers: Commercial 18.1; Islamic 20.2**
    - Loans/ Assets: Commercial 0.52; Islamic 0.52
    - Cost/ Income Ratio: Commercial 0.56; Islamic 0.60**
    - Income Diversity: Commercial 0.42; Islamic 0.42
    - Assets in Billion USD: Commercial 2.5; Islamic 22.9
  - Large Banks:
    - Z-Score: Commercial 24.5; Islamic 13.2***; Z-Score excl. outliers: Commercial 19.5; Islamic 12.9***
    - Loans/ Assets: Commercial 0.52; Islamic 0.56**
    - Cost/ Income Ratio: Commercial 0.55; Islamic 0.65***
    - Income Diversity: Commercial 0.42; Islamic 0.39
    - Assets in Billion USD: Commercial 6.06; Islamic 7.53**
  - Small Banks:
    - Z-Score: Commercial 200.1; Islamic 35.8
    - Z-Score excl. outliers: Commercial 17.2; Islamic 25.0***
    - Loans/ Assets: Commercial 0.51; Islamic 0.50
    - Cost/ Income Ratio: Commercial 0.57; Islamic 0.58
    - Income Diversity: Commercial 0.42; Islamic 0.41
    - Assets in Billion USD: Commercial 0.3; Islamic 0.3
- Descriptive interpretations:
  - Excluding outliers, Islamic banks’ z-scores are on average higher than commercial banks’.
  - Small Islamic banks have higher z-scores than small commercial banks; large Islamic banks have lower z-scores than large commercial banks.
  - In full sample, extreme outliers among small commercial banks can reverse simple comparisons.
  - Large Islamic banks have on average higher loan-to-asset ratios and higher cost-to-income ratios than large commercial banks.
  - No significant difference in income diversity between Islamic and commercial banks (large or small).
  - Islamic banks in sample are on average bigger than commercial banks; difference significant for large banks.

### Alternative z-score Measures and Robustness (selected exact figures from Table 3)
- Alternative measures (Volatility of Z-Scores, % points) preserved exactly:
  - All Banks:
    - Modified Z-Score: Commercial 72.63; Islamic 53.02***
    - Downward: Commercial -3.38; Islamic -4.93***
    - Upward: Commercial 3.21; Islamic 5.52***
  - Large Banks:
    - Modified Z-Score: Commercial 93.62; Islamic 36.92***
    - Downward: Commercial -3.06; Islamic -4.13***
    - Upward: Commercial 2.66; Islamic 2.51
  - Small Banks:
    - Modified Z-Score: Commercial 59.02; Islamic 64.88
    - Downward: Commercial -3.57; Islamic -5.14***
    - Upward: Commercial 3.59; Islamic 7.60***
- Robustness conclusions from alternative measures:
  - Large Islamic banks are less stable than both small Islamic banks and large commercial banks.
  - Small Islamic banks are about as stable or more stable than small commercial banks.

### Regression Findings — Main Results (preserved phrasing and significance)
- Main empirical conclusions:
  - (i) small Islamic banks tend to be financially stronger than small commercial banks;
  - (ii) large commercial banks tend to be financially stronger than large Islamic banks;
  - (iii) small Islamic banks tend to be financially stronger than large Islamic banks.
- Regression specifics:
  - Large Islamic banks tend to be less stable than large commercial banks: Islamic dummy predominantly negative and significant at the 1 percent level in large-bank regressions (specifications (5) to (8) in Tables 4 and 5).
  - Small Islamic banks tend to be more stable than small commercial banks: Islamic dummy consistently positive across regressions for small banks and significant at the 5 percent level in all OLS regressions and one robust regression.
  - Full-sample regressions less clear-cut due to opposing signs for large vs. small banks.
- Control-variable effects (signs and significance as reported):
  - Loans/Assets: higher loan-to-asset ratios associated with lower z-scores; coefficient consistently negative with two exceptions; significant in eleven robust specifications and in four OLS specifications.
  - Cost-to-income ratio: consistently negative link to z-scores; sign consistently significant except in several regressions for small banks.
  - Assets (size): z-scores tend to increase with bank size for large banks, but decrease with size for small banks.
  - Income diversity: greater income diversity tends to increase z-scores in large Islamic banks; interaction Income Diversity * Islamic Bank shows positive and significant coefficients in some large-bank specifications (examples reported in Table 4: 20.677 (0.001)***, 23.606 (0.000)***, 23.606 (0.000)***, 15.296 (0.017)**).
  - Governance: governance variable tends to be positive in all regressions in which entered and strongly positive in some specifications (examples: Governance 8.624 (0.055)* in Table 4; Governance 9.118 (0.005)*** in Table 5).
  - Herfindahl index: impact significantly negative in some regressions; effect less clear in robust estimations.
  - Exchange Rate Depreciation (-1): depreciation associated with higher banking risk (negative impact on z-scores); examples: OLS -0.063 (0.006)*** in Table 4; robust -0.015 (0.080)* in Table 5.
  - Real GDP Growth and Inflation: not significant in most specifications.
- Presence of Islamic banks in system:
  - In OLS, higher presence of Islamic banks has a negative impact on z-scores in large commercial banks, but a positive impact on z-scores in commercial banks in general; results weaker in robust estimation.
  - A higher presence of Islamic banks tends to weaken Islamic banks’ own z-scores in some regressions.
- Robustness checks summary:
  - Excluding Iran and Sudan did not materially change main results.
  - Using only unconsolidated or only consolidated data yielded consistent but weaker results.
  - Restricting to contemporaneous explanatory variables did not materially change main results.
  - Distinguishing majority government-owned vs. other banks showed no impact on main result (data caveat noted).

### Conclusions, Policy-Relevant Implications, and Further Research
- Interpretation of main findings:
  - Islamic banks may be relatively more stable at small scale but less stable at large scale after adjusting for size, balance sheet structure, and system-wide variables.
  - Possible mechanisms: difficulty scaling credit risk monitoring for PLS arrangements; limited standardization aggravating adverse selection and moral hazard at larger scale; small banks focusing on low-risk investments and fee income, while large banks may engage in more PLS business.
- Policy-relevant implication:
  - Mergers of small Islamic banks may improve efficiency (Yudistira, 2004), but to capture efficiency benefits prudential risks need attention because larger Islamic banks—other things equal—tend to have greater prudential risks.
- Limitations and caveats:
  - Preliminary nature of findings due to data caveats: incomplete coverage, use of unconsolidated statements in part, exclusion of Islamic windows/branches, inability to distinguish types of Islamic contracts (PLS vs. other), and inability to fully account for investment account types and return equalization funds.
- Suggested directions for further research:
  - Extend sample to more countries and banks by filling BankScope gaps with other data sources.
  - More comprehensive data collection to address data issues identified.
  - Analyze financial stability impacts of other forms of Islamic finance beyond fully-fledged Islamic banks.

### Appendix I — Background and Contracts (summary)
- Islamic banking consistent with Sharia; prohibits riba (interest).
- Contracts often PLS or purchase-resale; PLS returns not known or fixed ex ante.
- Islamic law generally prohibits trading in financial risk and investing in haram activities.

### Appendix II — Sample, Classification, and Variables (summary)
- Initial sample of 20 countries; regression analysis uses 18 due to data gaps for Lebanon and Kuwait.
- Aggregate asset figures: Islamic banks in panel have total assets of US$253 billion as of 2004.
- Classification: BankScope classification cross-checked with other sources; consolidated data used when available.
- Size classification: large banks defined as assets > US$1 billion.
- Variable list (selected):
  - Z-score: z≡(k+μ)/σ (authors’ calculations based on BankScope).
  - Assets_bln: total assets (U.S. dollars billion).
  - Loans_Assets, Cost_Income, Income Diversity, Herfindahl Index, GDP Growth, Inflation, Exch. Rate Depreciation, Islamic Bank Dummy, Share of Islamic Banks, Governance.

### Appendix Table 6 — Islamic Banking Basic Terminology (selected entries)
- Amana (Demand deposits): guaranteed in capital value, earn no return.
- Bay mu’ajal: deferred payment sale; price agreed at sale and cannot include charges for deferring payment.
- Bay salam: pre-payment, deferred delivery.
- Ijara: lease or lease purchase.
- Istisna: deferred payment, deferred delivery manufacturing contract.
- Ju’ala: service charge.
- Kifala: surety/pledge.
- Mudaraba: trustee finance contract; profits shared, financial losses borne by rabb-ul-mal.
- Murabaha: mark–up financing.
- Musharaka: equity participation; profits and losses shared by capital contributions.
- Qard Hassana: zero-return loans; banks may charge service fee not related to amount or maturity.

*Source: _wp0816 - References (excerpts provided from the paper’s Introduction, Section II, and Section III).*

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

### _wp0816 - References..............................................................................................................

### I. Introduction
- Islamic financial institutions have grown from one institution in one country in 1975 to over 300 institutions operating in more than 75 countries (El Qorchi, 2005).
- In Sudan and Iran, the entire banking system is currently based on Islamic finance principles.
- Total assets of Islamic banks worldwide are estimated at about $250 billion, and are expected to grow by about 15 percent a year (Choong and Liu, 2006; Ainley and others, 2007).
- Existing literature largely compares instruments and regulatory challenges of Islamic and commercial banking; empirical analysis of Islamic banks' role in financial stability is limited.
- This paper claims to be the first cross-country empirical analysis of the role of Islamic banks in financial stability.
- Paper structure:
  - Section II: specifics of Islamic banking from a prudential perspective (see Appendix I for more detail).
  - Section III: methodology, variables, and data (Appendix II provides more detail).
  - Section IV: empirical results.
  - Section V: conclusions and topics for further research.

### II. Specifics of Islamic Banking from a Prudential Perspective
- Definition and contract features:
  - Islamic or Shari’ah-compliant banking: provision and use of financial services and products that conform to Islamic religious practices and laws.
  - Prohibition against payment and receipt of interest at a fixed or predetermined rate.
  - Contracts rely on profit-and-loss sharing arrangements (PLS), purchase and resale of goods and services, and provision of services for fees.
  - In PLS modes, the rate of return is not known or fixed prior to undertaking the transaction.
  - In purchase-resale transactions, a mark-up is determined based on a benchmark rate of return, typically a return determined in international markets such as LIBOR.
  - Range of contracts available depending on investor rights in project management and timing of cash flows.
  - Islamic banks generally prohibited from trading in financial risk, and from financing production/trade in alcoholic beverages or pork, non-Islamic media, and gambling operations.
- Potential risk-increasing features highlighted in the literature:
  - PLS financing shifts direct credit risk from banks to investment depositors, but increases overall asset-side risk as banks are vulnerable to risks normally borne by equity investors.
  - Operational risk is crucial: defined as loss risk from inadequate or failed internal processes, people and systems, or external events, including legal risk and Sharī`ah compliance risk; administration of PLS modes increases operational risk.
  - PLS cannot be made dependent on collateral or guarantees to reduce credit risk.
  - Product standardization is more difficult due to multiplicity of financing methods, increasing operational risk and legal contract uncertainty.
  - Islamic banks can use fewer risk-hedging instruments and traditionally have operated in environments with underdeveloped interbank and money markets and limited lender-of-last-resort access; significance of these differences has decreased due to recent developments in Islamic money market instruments and Islamic lender-of-last-resort modes and implicit liquidity support in most countries.
  - Non-PLS modes are less risky and resemble conventional financing but still carry risks, such as elevated operational risk in some cases.
  - Investment deposits: capital value and rate of return are not guaranteed, potentially increasing moral hazard and incentives for risk taking and undercapitalization.
- Potential risk-mitigating features:
  - Islamic banks can pass through negative asset shocks (e.g., Musharaka loss) to investment depositors (Mudaraba arrangement).
  - Risk-sharing on deposit side provides additional protection beyond book capital.
  - Shareholders’ responsibility for negligence or misconduct and more difficult access to liquidity can create pressures to be more conservative, reducing moral hazard and risk taking.
  - Investors typically do not have deposit insurance and thus have stronger incentives to exercise oversight over management.
  - Islamic banks traditionally hold a comparatively larger proportion of assets in reserve accounts with central banks or in correspondent accounts, providing buffers.
- Conclusion from Section II:
  - Whether Islamic banks are more or less stable than conventional banks is an empirical question that may differ across countries and banks; the rest of the paper aims to provide empirical evidence.

### III. Methodology and Data
#### A. Measuring Bank Stability
- Primary dependent variable: the z-score as a measure of individual bank risk.
- Z-score definition and interpretation:
  - z ≡ (k + μ) / σ
    - k: equity capital and reserves as percent of assets
    - μ: average return as percent of assets
    - σ: standard deviation of return on assets (proxy for return volatility)
  - The z-score measures the number of standard deviations a return must fall to deplete equity, under the assumption of normality of bank returns.
  - A higher z-score corresponds to a lower upper bound of insolvency risk; higher z-score implies lower probability of insolvency.
- Applicability to Islamic banks:
  - Z-score focuses on insolvency risk—relevant across institution types.
  - Criticism: investment accounts in Islamic banks may act like equity, so book capital and reserves may understate financial strength.
    - Investment accounts often linked to pre-agreed maturity (from one month upwards) and can generally be withdrawn with one month’s advance notice.
    - Profits distributed per pre-determined ratio (example provided: 80 percent to depositors and 20 percent to the bank).
    - In extreme cases, a bank with only restricted investment accounts could resemble a mutual fund.
  - Counterarguments:
    - Conventional banks can also pass risks to customers via deposit and loan rate adjustments; additional protection layers are ultimately reflected in banks’ returns and capital and thus in z-scores.
    - Short withdrawal features and pre-determined profit-sharing reduce factual differences with conventional funding.
  - Authors’ view: capital and reserves are a reasonable proxy to assess default risk; on balance, arguments favoring z-score use are stronger.
- Preliminary tests:
  - Pairwise comparisons of z-scores between Islamic and commercial banks.
  - Subdivision by size: large versus small banks, using total assets of US$ 1 billion as the cut-off point between small and large banks.

#### B. Regression Analysis
- General panel model estimated:
  - z_{tji} = α + β s_{...} + γ M_{...} + δ C_j + φ I_{t-1} + φ T_{s-1} + ... + λ BT_{...} + ε_{tji}
  - (Model notation as presented in source: tjittjjtjstjistjsssstjtjitji DCMTBITTIBz ,,1,1,,1,1,1,,, επλπφφδγβα+++++++++= ∑∑∑∑∑ −−−−− )
  - Dependent variable: z-score z_{tji} for bank i in country j at time t.
  - B_{t-1,ji}: vector of bank-specific variables.
  - I_{t-1,j}: time-varying industry-specific variables.
  - T_s and I_{t-1,js}: type of banks and interaction between type and some industry variables.
  - M_{t,j}: macroeconomic variables; C_j: country dummies; D_t: yearly dummies; ε_{tji}: residual.
- Key variables of interest:
  - Islamic bank dummy: equals 1 if bank is Islamic, 0 if commercial. A negative coefficient would indicate Islamic banks are relatively weaker (lower z-scores).
  - Market share of Islamic banks by assets (calculated for each year and country) is interacted with Islamic and commercial bank dummies to test systemic-level impact.
    - A negative sign for interaction of Islamic banks’ market share and Islamic bank dummy would indicate that a higher share of Islamic banks reduces their soundness (reduces z-scores).
- Control variables included:
  - Bank-level: asset size in U.S. dollars billion; loans over assets; cost-income ratio.
  - Income diversity measure following Laeven and Levine (2005) to capture degree of diversification from net interest income–type activities to other activities.
    - For Islamic banks, net interest income is generally defined as the sum of positive and negative income flows associated with PLS arrangements (International Monetary Fund, 2004).
  - Interaction of income diversity variable with Islamic bank dummy to capture business orientation differences.
- Rationale for controls: to account for differences in size, asset composition, cost efficiency, and income structure between Islamic and other banks.

*Source: _wp0816 - References (excerpts provided from the paper’s Introduction, Section II, and Section III).*

### Appendix II provides a description of the variables, and Table 2 provides summary statistics for the key bank-

### _wp0816 - Appendix II provides a description of the variables, and Table 2 provides summary statistics for the key bank-

### Variable definitions and key measures
- Income diversity measure:
  - Defined as (incomeoperatingOther − Net interest income) / incomeoperatingTotal − 1.
  - Higher values correspond to a higher degree of diversification.

### Country-level controls
- Macroeconomic cycle adjustments:
  - Include GDP growth rate, inflation rate, and exchange rate depreciation (same value for all banks in a given country).
- Market concentration:
  - Herfindahl index defined as the sum of squared market shares (in terms of total assets) of all banks in the country.
  - Index range: 0 to 10,000 (for a system with only one bank).
  - No strong prior on its impact because literature contains contrasting views:
    - Allen and Gale (2004): more concentrated markets likely more stable.
    - Mishkin (1999): more concentrated systems characterized by increased risk-taking by banks.
- Governance:
  - Governance indicator compiled by Kaufmann, Kraay, and Mastruzzi (2005).
  - Average of 6 measures: voice and accountability, political stability, government effectiveness, regulatory quality, rule of law, and control of corruption.
  - Averaged across available years 2004, 2002, 2000, 1998, and 1996 into a single index per country.
  - Captures cross-country differences in institutional developments that might affect banking risk.

### Lag structure and variable timing
- All bank-specific and macroeconomic variables, the Herfindahl index, the Islamic banks’ market share and its interaction with the Islamic and commercial bank dummies are lagged to capture possible past effects on banks’ individual risk.
- Robustness check: also test restricting explanatory variables to contemporaneous effects.

### Estimation methods
- Primary technique:
  - Pooled ordinary least squares (OLS).
- Robust estimation:
  - Use Hamilton (2002) robust estimation technique to downweight observations with large residuals via an iterative process.
  - Standard errors calculated using the pseudovalues approach (Street, Carroll, and Ruppert, 1988).
- Additional methods for sensitivity:
  - Fixed effects regressions.
  - Median least squares regressions (minimizes the median square of residuals rather than the average to reduce effect of outliers).

### Sample robustness and sub-sample analysis
- Robustness with respect to sample selection:
  - Re-estimate regressions for different bank sizes.
  - Sub-samples:
    - Large banks: total assets of more than US$1 billion.
    - Small banks: all others.

### Data
- Source: BankScope database (commercially available).
- Coverage:
  - Islamic and commercial banks in 20 banking systems with a non-negligible presence of Islamic banks.
  - Jurisdictions included (alphabetically ordered): Bahrain, Bangladesh, Brunei, Egypt, Gambia, Indonesia, Iran, Jordan, Kuwait, Lebanon, Malaysia, Mauritania, Pakistan, Qatar, Saudi Arabia, Sudan, Tunisia, United Arab Emirates, West Bank and Gaza, and Yemen.
- Observations:
  - Up to 520 observations for 77 Islamic banks.
  - 3,248 observations for 397 commercial banks.
  - Sample period: 1993 to 2004.

*Source: _wp0816 - Appendix II provides a description of the variables, and Table 2 provides summary statistics for the key bank-*

### Appendix II for additional information on sample coverage).

### _wp0816 - Appendix II for additional information on sample coverage)

### Sample coverage and BankScope data issues
- Sample includes systems where Islamic banks accounted for more than 1 percent of total assets in at least one year in 1993–2004.
- Islamic banks covered in the starting sample have total assets of US$253 billion as of 2004, consistent with an earlier estimate of “about US$250 billion”.
- About 49 percent of the Islamic banks and about 62 percent of the commercial banks fall into the large bank category (assets over US$ 1 billion).
- BankScope coverage varies by country; for most countries in the sample BankScope covers 80–90 percent of the banking systems in terms of total assets.
- Two countries were excluded from regression analysis because BankScope lacks unconsolidated observations for Islamic banks (Lebanon and Kuwait), reducing aggregate-country coverage from 20 to 18 countries.
- Consolidation: only about 1/3 of relevant BankScope observations are consolidated; the rest are unconsolidated. Consolidated data are used when available; otherwise unconsolidated data are used.
- Cross-checks: BankScope data on Islamic banks were cross-checked against lists from the Institute of Islamic Banking and Insurance and IBF.net (sources noted in original).
- Data-definition issues exist for Islamic banks (e.g., capital definition, measuring equivalent of interest income); the paper relied largely on BankScope definitions, with basic cross-checking and outlier exclusion.

### Input data overview (from BankScope)
- Table 1 (authors’ calculations based on BankScope data) presents counts:
  - All Banks:
    - Commercial: Number of banks 1153; Number of observations 1,016
    - Islamic: Number of banks 397; Number of observations 3,248
    - Consolidated / Consolidated and Unconsolidated columns present in source table.
  - Large Banks (assets > 1 billion USD):
    - Commercial: Number of banks 892; Number of observations 676
    - Islamic: Number of banks 461; Number of observations 1,190
  - Small Banks (assets ≤ 1 billion USD):
    - Commercial: Number of banks 261; Number of observations 340
    - Islamic: Number of banks 517; Number of observations 2,058
- Note: Large (small) banks defined as having assets larger (smaller) than 1 billion USD.

### Pairwise comparisons and descriptive statistics (z-scores and other variables)
- Z-score variability: z-score ranges from -81 to 203,347, indicating presence of outliers; robust estimation methods are used in regressions.
- Outlier treatment: averages shown for full sample and excluding 1st and 99th percentiles of z-score distribution.
- Table 2 (Averages across banks, 1993–2004) key figures (preserved exactly as in source):
  - All Banks:
    - Z-Score: Commercial 132.9; Islamic 27.2
    - Z-Score excl. outliers: Commercial 18.1; Islamic 20.2**
    - Loans/ Assets: Commercial 0.52; Islamic 0.52
    - Cost/ Income Ratio: Commercial 0.56; Islamic 0.60**
    - Income Diversity: Commercial 0.42; Islamic 0.42
    - Assets in Billion USD: Commercial 2.5; Islamic 22.9
  - Large Banks:
    - Z-Score: Commercial 24.5; Islamic 13.2***; Z-Score excl. outliers: Commercial 19.5; Islamic 12.9***
    - Loans/ Assets: Commercial 0.52; Islamic 0.56**
    - Cost/ Income Ratio: Commercial 0.55; Islamic 0.65***
    - Income Diversity: Commercial 0.42; Islamic 0.39
    - Assets in Billion USD: Commercial 6.06; Islamic 7.53**
  - Small Banks:
    - Z-Score: Commercial 200.1; Islamic 35.8
    - Z-Score excl. outliers: Commercial 17.2; Islamic 25.0***
    - Loans/ Assets: Commercial 0.51; Islamic 0.50
    - Cost/ Income Ratio: Commercial 0.57; Islamic 0.58
    - Income Diversity: Commercial 0.42; Islamic 0.41
    - Assets in Billion USD: Commercial 0.3; Islamic 0.3
- Descriptive findings:
  - For sample excluding outliers, Islamic banks’ z-scores are on average higher than commercial banks’.
  - Small Islamic banks have higher z-scores than small commercial banks (indicating higher stability); large Islamic banks have lower z-scores than large commercial banks.
  - In full sample (with outliers), small commercial banks include some with extremely high z-scores which can reverse simple comparisons.
  - Large Islamic banks have on average higher loan-to-asset ratios than large commercial banks (reflecting Islamic banking constraints on investment in bonds/T-bills).
  - Large Islamic banks have higher cost-to-income ratios than large commercial banks (consistent with literature on efficiency).
  - No significant difference in income diversity between Islamic and commercial banks (large or small).
  - Islamic banks in the sample are on average bigger than commercial banks; difference significant for large banks.

### Robustness checks and alternative z-score measures
- Alternative z-score definitions explored (Table 3):
  - Modified z-score: capitalization plus ROA over absolute value of downward volatility of ROA.
  - Downward (upward) volatility of z-scores: sample average of differences between bank-specific yearly z-score and its mean when below (above) bank-specific mean.
- Table 3 figures (Volatility of Z-Scores, % points), preserved exactly:
  - All Banks:
    - Modified Z-Score: Commercial 72.63; Islamic 53.02***
    - Downward: Commercial -3.38; Islamic -4.93***
    - Upward: Commercial 3.21; Islamic 5.52***
  - Large Banks:
    - Modified Z-Score: Commercial 93.62; Islamic 36.92***
    - Downward: Commercial -3.06; Islamic -4.13***
    - Upward: Commercial 2.66; Islamic 2.51
  - Small Banks:
    - Modified Z-Score: Commercial 59.02; Islamic 64.88
    - Downward: Commercial -3.57; Islamic -5.14***
    - Upward: Commercial 3.59; Islamic 7.60***
- Robustness check results align with regular z-score findings:
  - Large Islamic banks are less stable than both small Islamic banks and large commercial banks.
  - Small Islamic banks are about as stable or more stable than small commercial banks.
- Other checks: alternative financial soundness measures (e.g., nonperforming loans) are not viable as comprehensive measures for this study; credit ratings sample for Islamic banks too small.

### Regression analysis — methodology and key regression outcomes
- Regression approaches: OLS (Table 4) and robust estimation (Table 5) that downweights observations with large residuals; also fixed effects and median least squares estimations (consistent results reported).
- Main regression findings (preserved language and significance statements):
  - Large Islamic banks tend to be less stable than large commercial banks: Islamic dummy predominantly negative and significant at the 1 percent level in regressions for large banks (specifications (5) to (8) in Tables 4 and 5).
  - Small Islamic banks tend to be more stable than small commercial banks: Islamic dummy consistently positive across regressions for small banks and significant at the 5 percent level in all OLS regressions and one robust regression.
  - Full-sample regressions are less clear-cut due to opposite signs for large vs. small banks.
- Control variable effects (summary of signs and significance as reported):
  - Loans/Assets: higher loan-to-asset ratios tend to be associated with lower z-scores; coefficient consistently negative with two exceptions; significant in eleven robust specifications and in four OLS specifications.
  - Cost-to-income ratio: consistently negative link to z-scores; sign consistently significant except in several regressions for small banks.
  - Assets (size): z-scores tend to increase with bank size for large banks, but decrease with size for small banks.
  - Income diversity: greater income diversity tends to increase z-scores in large Islamic banks; interaction Income Diversity * Islamic Bank shows positive and significant coefficients in some large-bank specifications (e.g., large-bank OLS specifications where Income Diversity * Islamic Bank shows 20.677, 23.606, 23.606, 15.296 with p-values (0.001)***, (0.000)***, (0.000)***, (0.017)** respectively in Table 4).
  - Governance: governance variable tends to be positive in all regressions in which entered and is strongly positive in some specifications (e.g., Governance 8.624 with p-value (0.055)* in Table 4; Governance 9.118 with p-value (0.005)*** in Table 5).
  - Herfindahl index: impact significantly negative in some regressions (consistent with literature linking higher concentration to lower stability); effect less clear in robust regressions.
  - Macroeconomic variables:
    - Exchange Rate Depreciation (-1): depreciation associated with higher banking risk (negative impact on z-scores); significant in several specifications (e.g., OLS: -0.063 with p-value (0.006)*** in Table 4; robust: -0.015 with p-value (0.080)* in Table 5).
    - Real GDP Growth and Inflation: not significant separate effects in most specifications.
  - Presence of Islamic banks in system:
    - In OLS, higher presence of Islamic banks has a negative impact on z-scores in large commercial banks, but a positive impact on z-scores in commercial banks in general; results weaker in robust estimation.
    - A higher presence of Islamic banks tends to weaken Islamic banks’ own z-scores in some regressions.
- Robustness checks reported:
  - Excluding Iran and Sudan did not materially change main results.
  - Running regressions using only unconsolidated or only consolidated data yielded consistent but weaker results.
  - Restricting explanatory variables to contemporaneous effects did not materially change main results.
  - Additional checks distinguishing majority government-owned vs. other banks showed no impact on main result (caveat on data availability).

### Conclusions and suggested directions for further research
- Main empirical conclusions (preserved phrasing):
  - (i) small Islamic banks tend to be financially stronger than small commercial banks;
  - (ii) large commercial banks tend to be financially stronger than large Islamic banks;
  - (iii) small Islamic banks tend to be financially stronger than large Islamic banks.
- Interpretation and plausible explanations:
  - Islamic banks may be relatively more stable at small scale but less stable at large scale, after adjusting for bank size, balance sheet structure, and system-wide variables.
  - Possible reasons: complexity of adjusting credit risk monitoring as Islamic banks grow; limitations on standardization in monitoring profit-loss arrangements leading to aggravated adverse selection and moral hazard at larger scale; small banks may concentrate on low-risk investments and fee income, while large banks may engage in more PLS business.
- Policy-relevant implication noted:
  - Mergers of small Islamic banks may improve efficiency (as suggested by Yudistira, 2004), but to reap efficiency benefits prudential risks need attention because larger Islamic banks—other things equal—tend to have greater prudential risks.
- Limitations and caveats:
  - Preliminary nature of findings due to data caveats: incomplete coverage, reliance in part on unconsolidated statements, exclusion of Islamic windows/branches, inability to distinguish types of Islamic contracts (e.g., PLS vs. other investments), and inability to fully account for investment account types and return equalization funds.
- Suggestions for further research:
  - Extend sample to a larger number of countries and banks by filling BankScope gaps with other data sources.
  - More comprehensive data collection to address data-related issues identified in Section III.
  - Analyze financial stability impacts of other forms of Islamic finance beyond fully-fledged Islamic banks.

### Appendix I (brief background excerpt)
- Islamic banking: system of banking consistent with Islamic law (Sharia) and Islamic economics.
- Prohibitions and contract forms:
  - Islamic law prohibits usury (riba) — the collection and payment of interest.
  - Contracts often based on profit-and-loss-sharing arrangements (PLS) or purchase and resale of goods and services.
  - In PLS modes, the rate of return on financial assets is not known or fixed prior to the transaction.
  - Islamic law generally prohibits trading in financial risk (seen as a form of gambling) and prohibits investing in businesses considered haram (e.g., alcohol, pork, un-Islamic media).

*Source: authors’ calculations and text based on BankScope data as presented in the supplied PDF content.*

### Appendix Table 6 provides an overview of the basic terminology of Islamic banking.

### Appendix Table 6: Islamic Banking — Basic Terminology

### Basic terminology (Table 6)
- Amana (Demand deposits)  
  - Deposits held at the bank for safekeeping purpose. They are guaranteed in capital value, and earn no return.
- Bay mu’ajal (Pre-delivery, deferred payment)  
  - The seller can sell a product on the basis of a deferred payment, in installments or in a lump sum. The price of the product is agreed upon between the buyer and the seller at the time of the sale, and cannot include any charges for deferring payment.
- Bay salam (Pre-payment, deferred delivery)  
  - The buyer pays the seller the full negotiated price of a product that the seller promises to deliver at a future date.
- Ijara (Lease, lease purchase)  
  - A party leases a particular product for a specific sum and a specific time period. In the case of a lease purchase, each payment includes a portion that goes toward the final purchase and transfer of ownership of the product.
- Istisna (Deferred payment, deferred delivery)  
  - A manufacturer (contractor) agrees to produce (build) and to deliver a certain good (or premise) at a given price on a given date in the future. The price does not have to be paid in advance (in contrast to bay salam). It may be paid in installments or part may be paid in advance with the balance to be paid later on, based on the preferences of the parties.
- Ju’ala (Service charge)  
  - A party pays another a specified amount of money as a fee for rendering a specific service in accordance with the terms of the contract stipulated between the two parties. This mode usually applies to transactions such as consultations & professional services, fund placements and trust services.
- Kifala  
  - It is a pledge given to a creditor that the debtor will pay the debt, fine or liability. A third party becomes surety for the payment of the debt if unpaid by the person originally liable.
- Mudaraba (Trustee finance contract)  
  - Rabb -ul- mal (capital’ s owner) provides the entire capital needed to finance a project while the entrepreneur offers his labor and expertise. Profits are shared between them at a certain fixed ratio, whereas financial losses are exclusively borne by rabb -ul- mal. The liability of the entrepreneur is limited only to his time and effort.
- Murabaha (Mark–up financing)  
  - The seller informs the buyer of his cost of acquiring or producing a specified product. The profit margin is then negotiated between them. The total cost is usually paid in installments.
- Musharaka (Equity participation)  
  - The bank enters into an equity partnership agreement with one or more partners to jointly finance an investment project. Profits (and losses) are shared strictly in relation to the respective capital contributions.
- Qard Hassana (Beneficence loans)  
  - These are zero-return loans that the Qur’an encourages Muslims to make to the needy. Banks are allowed to charge borrowers a service fee to cover the administrative expenses of handling the loan. The fee should not be related to the loan amount or maturity.

*Sources: Errico and Farrahbaksh (1998) and El-Hawary, Grais, and Iqbal (2004).*

### Appendix II — Data issues (sample and classification)
- Initial sample covered banks in the following 20 countries and jurisdictions (alphabetically ordered): Bahrain, Bangladesh, Brunei, Egypt, Gambia, Indonesia, Iran, Jordan, Kuwait, Lebanon, Malaysia, Mauritania, Pakistan, Qatar, Saudi Arabia, Sudan, Tunisia, United Arab Emirates, West Bank and Gaza, and Yemen.
- Data source: BankScope database, provided by Bureau van Dijk; annual data on all Islamic and commercial banks in the database from the above 20 countries.
- Sample counts and period:
  - up to 520 observations for 77 Islamic banks
  - 3,248 observations for 397 comparable commercial banks
  - period of 1993 to 2004
- Exclusions and consolidation:
  - BankScope does not have unconsolidated observations for Islamic banks in Lebanon and Kuwait, so these countries are excluded from the regression analysis, thus bringing the number of countries on which the aggregate results are based from 20 to 18.
  - We use consolidated data when available, but when consolidated data are not available for a bank, we use unconsolidated data instead.
- Aggregate asset figures:
  - Islamic banks included in the panel have total assets of US$253 billion as of 2004, which is in line with the “about US$250 billion” estimated worldwide assets of Islamic banks (see, e.g., El Qorchi, 2005).
- Classification procedures:
  - BankScope classification used as a starting point: defines as Islamic banks that are members of the “International Association of Islamic Banks” plus 20 non-member banks considered to be “Islamic” by FitchRatings.
  - Cross-checked BankScope classification with FSAP exercises, banks’ websites, and lists from the Institute of Islamic Banking and Insurance and IBF.net.
- Size classification:
  - Large banks are defined as those with total assets of more than US$1 billion.
  - All other banks are classified as small banks.

### Variables used in the paper (descriptions and sources)
- Z-score  
  - Defined as z≡(k+μ)/σ, where k is equity capital as percent of assets, μ is average return as percent of assets, and σ is standard deviation of return on assets as a proxy for return volatility. Measures the number of standard deviations a return realization has to fall in order to deplete equity, under the assumption of normality of banks’ returns.  
  - Source: Authors’ calculations based on BankScope data.
- Assets_bln  
  - Total assets of a bank (In U.S. dollars billion).  
  - Source: BankScope.
- Loans_Assets  
  - Ratio of loans to assets (percent).  
  - Source: BankScope.
- Cost_Income  
  - Ratio of cost to income (percent).  
  - Source: BankScope.
- Income Diversity  
  - () incomeoperatingTotal incomeoperatingOtherincomeerestintNet− −1  
  - Source: Authors’ calculations based on Laeven and Levine (2005) and BankScope.
- Income Diversity* Islamic Bank Dummy  
  - Interaction of income diversity and Islamic bank dummy.  
  - Source: Authors’ calculations based on BankScope.
- Herfindahl Index  
  - Sum of squared market shares of banks in the system.  
  - Source: Authors’ calculations based on BankScope.
- GDP Growth  
  - Growth rate of nominal GDP, adjusted for inflation (in local currency).  
  - Source: IMF (International Financial Statistics).
- Inflation  
  - Year-on-year change of the CPI index (percent).  
  - Source: IMF (International Financial Statistics).
- Exch. Rate Depreciation  
  - Year-on-year change in the nominal exchange rate, local currency per U.S. dollars (percent).  
  - Source: IMF (International Financial Statistics).
- Islamic Bank Dummy  
  - Equals 1 for Islamic banks; 0 otherwise.  
  - Source: Authors’ calculations based on BankScope.
- Share of Islamic Banks  
  - Market share of Islamic banks in a country per year.  
  - Source: Authors’ calculations based on BankScope.
- Share of Islamic Banks * Islamic Bank Dummy  
  - Interaction of share of Islamic banks and the Islamic bank dummy.  
  - Source: Authors’ calculations based on BankScope.
- Share of Islamic Banks * Commercial Bank Dummy  
  - Interaction of share of Islamic banks and the commercial bank dummy.  
  - Source: Authors’ calculations based on BankScope.
- Governance  
  - Average of the six governance measures- voice & accountability, political stability, government effectiveness, regulatory quality, rule of law and control of corruption- across the available years 2004, 2002, 2000, 1998 and 1996 into one single index per country.  
  - Source: Authors’ calculations based on Kaufmann, Kraay, and Mastruzzi (2005).

*Italic: Appendix Table 6 and Appendix II material from the source PDF.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp0816.pdf_