2.5 Empirical Results
2.5.3 Commingling, Cross-Subsidization and Bank Risk-Taking
Hypothesis 2 argues that under a common deposit insurance scheme, Islamic depos- itors can be affected by commingled funds when their banks become insolvent. In other words, under a conventional deposit insurance system, Islamic funds would be invested in interest-bearing assets (commingling). In case of bank insolvency, Is- lamic depositors would be cross-subsidized by the commingled funds. A sufficient condition to test this hypothesis is to consider how Islamic depositors react to bank insolvency before and after the reform. If Hypothesis 2 is indeed correct, one must observe a decrease in the deposit supply in response to an increase in bank insolvency
by depositors for pre-treatment observations. Also, the degree of disciplining against bank insolvency must decline for post-treatment observations.
Table 2.5 shows the estimation results of equations (2.3) and (2.4). Coefficient es- timates of Z−score demonstrate that before the treatment, Islamic depositors tend to withdraw their funds and ask for higher returns in response to an increase in the probability of bank insolvency (shown in columns (1) and (2), respectively). Interest- ingly, the coefficient estimates of the interaction term,Treatment∗Z−scorein columns (1) and (2) indicate that there is a significant decline in the degree of disciplining de- pending on the status of the treatment. That is, the degree of disciplining in response to an increase in the probability of bank insolvency decreases significantly after the announcement of a separate Islamic deposit insurance scheme.
In short, the results from Table 2.5 verify the hypothesis that depositors of Islamic banks are concerned about being compensated by the commingled funds in case their banks become insolvent. After the treatment, however, as the possibility of commin- gling and cross-subsidization vanishes, depositors of Islamic banks also stop engaging in disciplining through reducing the deposit supply in response to an increase in bank insolvency risk.
The economic intuition behind the estimation results presented in Table 2.5 might not be straight forward. Ultimately, the table presents the results on the effect of the treatment depending onZ−score. To be able to interpret the estimated coefficients in a meaningful way, I concentrate on the changes in deposit growth and return along the distribution of Z−score. Moving from 75th percentile (51.38) to 25th percentile (17.14) of the Z−score distribution, treatment variable increases deposit growth by 1.2 percentage points and decreases deposit return by 0.8 percentage points.
So far, I have not touched upon the issue of whether the treatment variable has an effect on risk-taking of individual banks. This issue is particularly relevant in my study for the following reason: If the new deposit insurance scheme provided an ad- equate regulatory discipline forcing Islamic banks to take on less financial risk, the evidence on depositors’ rewarding their banks might simply be due to a decrease in
Table 2.5: Evidence on commingling and cross-subsidization
(1) (2) (3) (4)
DEPG DEPR DEPG DEPR
Z-score 0.00101∗∗ -0.000353∗∗ 0.000905∗∗ -0.000319∗ (0.000306) (0.000104) (0.000298) (0.000136) Treatment ×Z-score -0.000317∗∗∗ 0.000162∗ -0.000313∗∗∗ 0.000153∗ (0.0000552) (0.0000764) (0.0000431) (0.0000615) DEPG(−i)ct 0.0493∗ (0.0205) DEPR(−i)ct 0.0751 (0.114)
Bank FE Yes Yes Yes Yes
Bank Fundamentals Yes Yes Yes Yes
Year FE Yes Yes Yes Yes
Macro Controls Yes Yes Yes Yes
Wild Bootstrap-t p-value 0.044 0.092 - -
N 294 292 294 292
R2 0.472 0.788 0.497 0.791
Notes: The sample period is 2007-2015. This table reports the estimation results from equa- tions (2.3) and (2.4). The dependent variable in columns (1) and (3) is DEPG. The dependent variable in columns (2) and (4) is DEPR. Treatmentrefers to the announcement of a separate Islamic deposit insurance scheme. All columns include bank fixed effects, year fixed effects, bank fundamentals and macro controls. Bank fundamentals areReturnonAssets,LoanQuality, Liquidity, Size and Capital AssetRatio. Macro controls include HH I, GDPperCapitaGrowth, In f lation,PopulationGrowth. DEPG(−i)ctandDEPR(−i)ctare respective mean values ofDEPG
and DEPR of each country and each year excluding each respective bank i itself. Standard errors are shown in parentheses and are clustered by country. Wherever shown, country-level clustered standard errors are wild bootstrapped. Wild bootstrapped p-values for the coeffi- cient of Treatment∗Z−score variable is reported. *, **, *** indicate significance at the 10%, 5% and 1% levels, respectively.
Table 2.6: Sharia-compliant deposit insurance and bank risk-taking
(1) (2) (3) (4) (5)
Z-score Loan Quality CAR ROA Liquidity
Treatment -4.699 0.00235 -0.0193 -0.00981 0.0709
(4.107) (0.0108) (0.0101) (0.0120) (0.0453)
Bank FE Yes Yes Yes Yes Yes
Bank Fundamentals No No No No No
Year FE Yes Yes Yes Yes Yes
Macro Controls Yes Yes Yes Yes Yes
Wild Bootstrap-t p-value 0.65 0.77 0.32 0.95 0.43
N 347 319 347 347 347
R2 0.951 0.638 0.878 0.482 0.618
Notes: The sample period is 2007-2015. This table reports the estimation results from yi,c,t = αi +αt+β1Treatmentc,t +β2MacroControlsc,t +ei,c,t. The dependent variables are
Z−score,LoanQuality,CAR(capital asset ratio),ROA(return on assets),Liquidity. Treatment refers to the announcement of a separate Islamic deposit insurance scheme. All columns in- clude bank fixed effects, bank fundamentals, year fixed effects and macro controls. Bank fundamentals include Size. Macro controls include HH I, GDPperCapitaGrowth, In f lation, PopulationGrowth. Standard errors are shown in parentheses and are clustered by coun- try. Wherever shown, country-level clustered standard errors are wild bootstrapped. Wild bootstrapped p-values for the coefficient of Treatment variable is reported. *, **, *** indicate significance at the 10%, 5% and 1% levels, respectively.
risk-taking. Although the reduced-form model takes care of the direct effects of reg- ulatory discipline (deposit demand side), the indirect effects manifesting themselves through financial risk-taking cannot be taken care of. Especially in the price equa- tion, this effect can show itself through a traditional risk-return relationship. Thus, whether such a scenario is present is an empirical concern, which I address in Table 2.6.
I use five dependent variables, which are largely accepted as factors capturing the publicly observable risk of banks (Martinez-Peria et al. (2001)). The coefficient of interest is still the treatment variable that measures if there was a significant change in terms of risk-taking differentially for treated banks after the announcement of a separate Islamic deposit insurance scheme. Given that all of these publicly observable risk variables could be affected by the treatment variable and confound the treatment, I exclude them as controls for one another in the regressions. Therefore, Table 2.6 presents the results without BankFundamentalsas controls.
As Table 2.6 suggests, none of the risk variables seem to be significantly changed after the treatment. Hence, the indirect effects of regulatory discipline affecting the supply of deposits through a reduction in banks’ financial risk-taking seem not to be present.