Table 2.3 presents the estimation of the model given by Equation (2.1). The first and second columns are the pooled OLS estimation where errors are clustered over countries to account for possible within country serial correlation in the combined error termαi+it. The second column includes country legal origin dummies to control for the institutional background of the economy.
The focus of this study is on the fixed effects model given by column three as discussed in the previous section, but still I present the pooled OLS estimations for comparison. The fixed effect models predict a strongly significant and negative effect of all three variables on the stringency of bank capital regulations.10 The negative coefficient on the GDP growth variable provides empirical support to the theoretical result in Kara (2013). He shows that high return countries choose less stringent bank capital regulations because higher returns allow a country to take on more risk through less stringent regulations. The fixed effects model estimates that one percentage point increase in the GDP growth rate reduces the stringency of capital regulation by 8.3% standard deviation. This coefficient is not only statistically highly significant, but it is also economically meaningful. The following exercise helps us to see the economic significance of the coefficient: what would be the necessary change in the GDP growth rate for an emerging country to have the capital stringency level of an advanced country? The mean of the standardized index is 0.18 for the advanced countries, and it is
10
-0.07 for the emerging countries. Therefore, the difference between the two means is equal to 0.25 standard deviation and a t-test shows that it is statistically greater than zero. The coefficient on the GDP growth variable implies that, everything else constant, if emerging countries had 3 percentage points lower GDP growth rate, they would choose the same level of bank capital stringency as the developed countries (0.25/0.083'3). A larger and economically more significant effect is obtained when a dynamic model is estimated in the next section.
Table 2.3: Results for the Statics Model
(1) (2) (3)
VARIABLES Pooled OLS Pooled OLS Fixed Effects
GDP growth -0.067*** -0.069*** -0.083*** (0.004) (0.004) (0.000) Government-owned banks -0.004 -0.003 -0.020*** (0.334) (0.388) (0.002) Concentration ratio -0.003 -0.003 -0.019** (0.424) (0.414) (0.013) Legal origin UK -0.046 (0.912) Legal origin FR -0.221 (0.568) Legal origin GE -0.157 (0.690) Constant 0.419* 0.584 1.657*** (0.097) (0.268) (0.000) Observations 246 245 246 R-squared 0.058 0.066 0.171 Number of code 81
1 The dependent variable is the standardized value of the capital stringency index.
The range of the standardized index is [−2.25,1.22]. For the pooled OLS regres- sions, standard errors are clustered at the country level to take into account the highly likely within country correlation in error terms.
2
GDP growth, government-owned banks and concentration ratio are expressed in percentages. For example, three percent real GDP growth rate is expressed as 3.0 in our data.
3 GDP growth is 3 year average growth rate for yearst, t−1 andt−2 whereas the
capital stringency index represents the state of the capital regulation the end of yeart.
4
Legal origin variables are dummies that do not change over time, and hence they are dropped in the fixed effect regressions as a result of the within transformation.
5
p-values in parentheses. *** p<0.01, ** p<0.05, * p<0.1
The negative coefficient on government-owned banks supports the theoretical prediction in Dell’Ariccia and Marquez (2006). They show that if a regulator is more concerned about the
banking sector profits as opposed to the financial stability, it will choose less stringent capital regulations. I use government ownership of banks as a proxy for the weight on bank profits in a regulator’s objective function. The fixed effects results show that one percentage point increase in government ownership of banks ratio leads to 2% standard deviation decrease in the stringency of bank capital regulations. Using the same exercise from the previous paragraph, the coefficient implies that, everything else constant, if emerging countries had 12.5 percentage points lower government-owned bank ratios, they would choose the same level of bank capital stringency as the developed countries (0.25/0.02'12.5). Since government ownership of banks varies between 0% and 80% in our sample, the estimated effect is economically reasonable in addition to being statistically highly significant.
Lastly, the negative coefficient on concentration ratio shows that regulators relax strin- gency of bank capital regulations as the banking sector becomes more concentrated. This could happen if the regulators tend to associate higher concentration ratio with less incentives for excessive risk taking in the banking sector. Therefore, the result supports the “concentration- stability” hypothesis in the long theoretical and empirical divide about the effect of concen- tration in the banking sector on financial stability. The fixed effects result shows that one percentage point increase in the concentration ratio reduces the stringency of capital regula- tions by 1.9% standard deviations. Again let us ask: what would be the necessary change in concentration ratio for an emerging country to have the capital stringency level of an advanced country? The coefficient implies that, everything else constant, if emerging countries had 13 percentage points lower concentration ratio, they would choose the same level of bank capital stringency as the developed countries. (0.25/0.019 ' 13). This effect is also economically meaningful given that the concentration ratio varies between 15% and 100% in our sample.
Estimating a pooled OLS model increases standard errors significantly: government-owned banks and concentration ratio become insignificant whereas the p-value of GDP growth rises from 0.000 to 0.004. However, the pooled OLS results are inconsistent if the independent variables are correlated with fixed country effects, which is quite likely in this setup. I provide them here only for comparative reasons. Also, the second column in Table 2.3 shows that none of the legal origin dummies have a significant effect on the stringency of capital regulations.