Islamic Banking System: The Empirical Framework (2024-2025) PDF

Empirical Banking Islamic Banking System THE EMPIRICAL FRAMEWORK Stefano Caiazza 2024-2025  Research Questions 1) What is the profitability of Islamic banks? 2) What is the risk profile of Islamic banks? 3) How did Islamic banks behave in the Financ...

Empirical Banking Islamic Banking System THE EMPIRICAL FRAMEWORK Stefano Caiazza 2024-2025  Research Questions 1) What is the profitability of Islamic banks? 2) What is the risk profile of Islamic banks? 3) How did Islamic banks behave in the Financial crisis (and, in general, during and post crises)?  Total assets of the three largest bank in Europe Total assets of Islamic banks worldwide by region 1. HSBC Holdings, UK – €2,860 billion 2. BNP Paribas, France – €2,850 billion 3. Crédit Agricole Group, France – €2,540 billion Total assets of the three largest bank in the US 1. JPMorgan Chase – $3,670 billion 2. Bank of America – $3,050 billion 3. Citigroup – $2,420 billion Total assets of the three largest bank in China Industrial and Commercial Bank of China Limited – $5,740 billion China Construction Bank – $5,020 billion Source: Statista Agricultural Bank of China – $4,9020 billion Distribution of global Islamic banking assets in of 2021, by country Source: Statista Size and profit for 125 Islamic Banks in 2022 Source: BankFocus  1. Introduction The global financial crisis has not only shed doubts on the proper functioning of conventional ‘‘Western’’ banking but has also increased the attention on Islamic banking, as some observers have pointed to their superior performance during the crisis (Hasan and Dridi, 2010). Paper by Beck, Demirguc-Kunt and Merrouche (BDM) compares the business model, efficiency, asset quality, and stability of Islamic and conventional banks, using an array of indicators constructed from the balance sheet and income statement data across a sample of 22 countries with both Islamic and conventional banks. BDM gauge the relative performance of both bank groups during local banking crises and the recent global financial crisis. BDM paper thus sheds light on an important debate. While proponents of Shariah-compliant financial services point to clear differences in business models of Islamic and conventional banks and higher efficiency and stability of Islamic banks, critics argue that (i) conventional and Islamic banks might be different in form but are similar in substance and/or (ii) Islamic banks do not have any advantages in efficiency and stability.  2. Sharia-compliant products and agency problems Explanation of the main aspects of Shariah for banks. Agency problems. Conventional banks: The debt contract with deterministic monitoring (in case of default) (Diamond, 1984) or stochastic monitoring (Townsend, 1979) is optimal for financial intermediation in a World with a large number of savers and entrepreneurs. In addition, banks face a maturity mismatch between deposits, demandable on sight, and long-term loans, which can result in bank runs and insolvency (Diamond and Dybvig, 1983). Diamond and Rajan (2001) argue that it is precisely the double agency problems banks face, with depositors monitoring banks, that disciplines banks in turn to monitor borrowers, while Government interventions such as deposit insurance distort such equilibrium. Islamic banks: On the one hand, the equity-like nature of savings and investment deposits might increase depositors’ incentives to monitor and discipline the bank. At the same token, the equity-like nature of deposits might distort the bank’s incentives to monitor and discipline borrowers as it does not face a threat by depositors of immediate withdrawal. Similarly, the equity-like character of partnership loans can reduce the necessary discipline imposed on entrepreneurs by debt contracts (Jensen and Meckling, 1976). On the other hand, the equity character of banks’ asset side of the balance sheet and thus higher risk might also increase the uncertainty of depositors’ return and increase the likelihood of both uninformed and informed bank runs. This is exacerbated by the restrictions that banks face on terminating partnership loans or restricting them in maturity.  3. Data and methodology BDM use a sample that comprises only countries with both conventional and Islamic banks, which allows them to control for any unobserved time-variant effect by introducing country- year dummies. They only include banks with at least two observations and countries with data on at least four banks. They eliminate outliers in all variables by winsorizing at the 1st and 99th percentiles within each country. They also double-check the categorization of Islamic banks in Bankscope with information from Islamic Banking Associations and country-specific sources.  3. Data and methodology Their main analysis covers a period between 1995 and 2009 and includes 510 banks across 22 countries, of which 88 are Islamic banks. In addition, BDM use a sample of 209 listed banks across 21 countries for the period 2005–2009 to assess the impact of the global financial crisis on the stock market performance of Islamic and conventional banks.  Dependent Variables Business orientation Loans / Deposits Fees / Operating Income Non Deposits Funds / Total Funds Efficiency Overhead cost = Total Operating Costs / Total Assets Cost to Income Asset Quality Loan loss provision / Total gross loans Loss reserves / Total gross loans Non-performing loans (NPLs) / Total gross loans Bank stability Maturity match = Liquid assets / (deposit + and short-term funding) Z-score = [ROA + (Equity / Asset)] / σROA ROA Equity / Asset Controls Size = Log (Total assets) ; Fixed Assets / Total Assets ; Non-loan earnings assets / Total Assets  Z-Score Let π≡profits, A≡assets, E≡equity where π is a random variable (r.v.). Default occurs when -π>E and the probability of default  π E − EA P ( −π ≥ E ) → P  − ≥  → P ( − ROA ≥ EA ) → P ( ROA ≤ − EA ) P ( ROA ≤ − EA ) = ∫ φ ( ROA ) dROA  A A z −∞ If ROA is normally distributed: P ( ROA ≤ − EA) = ∫ N ( 0,1) dz −∞ − EA − µ ( ROA) with z = σ ROA z is the number of standard deviations below the mean by which ROA would have to fall in order to eliminate equity.  − EA − µ ( ROA )   EA + µ ( ROA )  P ( ROA ≤ − EA ) = Φ   = 1 − Φ    σ ROA   σ ROA  Where Φ is the standard normal cumulative distribution function. By Chebyshev’s inequality, we can calculate the upper bound of the probability of insolvency EA + µ ( ROA ) P ( ROA ≤ − EA ) ≤ Z 2 where Z ≡ σ >0 ROA Since Z-score is inversely proportional to the upper bound of the probability of default, Z-Score can be view, than it is a measure of the overall bank stability. Higher Z-Score implies a higher degree of solvency and therefore it givers a direct measure of bank stability  Estimating Z-Score (past observations). If EA is a r.v., Z-Score should also account for σEA (past 3 observations). σROA is biased (last observation for ROA and σROA over the whole sample). What r.v. is being estimates? (modified version of Z2). Both mean and sd are computed over the whole sample. 3. Data and methodology  Model Specification 1 where Bank is one of measures of business orientation, efficiency, asset quality, and stability of bank i in country j in year t; B is a vector of time-varying bank characteristics; Cj×Yt are country-year-fixed effects (22 countries × 14 years = 308 CY-fixed effects); I is a dummy taking the value one for Islamic banks and zero otherwise; εi,t is a white-noise error term. Error terms are clustered at the bank level, i.e., correlation among the error terms across years within banks.  Results Table 3 Table 4  Model Specification 2 where Bank is one of our measures of business orientation, efficiency, asset quality, and stability of bank i in country j in year t B is a vector of time-varying bank characteristics I is a dummy taking the value one for Islamic banks and zero otherwise Cj×Yt are country-year-fixed effects (22 countries × 14 years = 308 CY-fixed effects) Cj× Ii is the interaction term between the Islamic bank dummy and the country dummies (22 countries × 2 = 44 fixed effects) Another interaction term is calculated as a product between the Islamic bank dummy (Ii) and the size dummy dummies (2 × 6 = 12 fixed effects) εi,t is a white-noise error term. Error terms are clustered at the bank level, i.e., correlation among the error terms across years within banks.  Results Large banks: size > the 75th percentile Medium banks: the 25th percentile ≤ size ≤ the 75th percentile Small banks: size < the 25th percentile Results Results Robustness Check: local crisis (ratios) Robustness Check: The financial crisis (stock market returns) Robustness Check: the Financial crisis (stock market returns) Dependent: Stock market (quarterly)  Conclusions Few significant differences in business models. Islamic banks are less efficient… …but have with respect to conventional banks: Higher intermediation ratio Higher asset quality Better capitalization Islamic banks perform better during capitalization and asset quality crises and are less likely to disintermediate than conventional banks. Variations across countries and Islamic banks of different sizes are important.  Comments Good comments in the text Outliers are probabily still presents  Comments Do they use a pooled estimator?  Comments Table 3 Model is misspecified Table 4 Adjusted R-squared  Comments And the single variables? “As we interact the Islamic bank dummy with all 22 country dummies, we drop the Islamic bank dummy itself». Large, Medium, and Small size banks: arbitrary classification “The results in Table 5 show that large Islamic banks have higher cost-income ratios than conventional banks and lower non-deposit funding (significant at Is it trues? the 10% level), while large Islamic banks do not differ from conventional banks along any of the other dimensions” (page 441).  Comments “In Panel B, we replace the Islamic bank dummy with two interaction terms of the bank dummy with dummies indicating country-year pairs with the Islamic bank share above and below the median market share of Islamic banks”. How is it possibile to use 2 complementary dummies (H-L)?  Comments And the single constitutive terms? I am interested in comparing before and after the crisis  Comments And the single constitutive terms?  Stata reghdfe reghdfe is a generalization of areg (and xtreg,fe, xtivreg,fe) for multiple levels of fixed effects (including heterogeneous slopes), alternative estimators (2sls, gmm2s, liml), and additional robust standard errors (multi-way clustering, HAC standard errors, etc). reghdfe depvar [indepvars] [if] [in] [weight] , absorb(absvars) [options] ivreghdfe depvar [indepvars] [endogenous=instrument(s)] [if] [in] [weight] , absorb(absvars) [options]