Banking and Interbank Markets.

Starting from the stylized fact that banks hold capital ratios in excess of the regulatory minimum capital requirement, this report examines the determinants of banks' capital ratio in Germany from an internal balance sheet perspective. It borrows heavily from the empirical banking and corporate finance literature in choosing cost and revenue variables that affect the trade-off which banks face in setting their optimal capital ratio. A vast panel dataset comprising balance sheet and income statement metrics of all German banks obtained from the reporting system of Deutsche Bundesbank allows for a comprehensive analysis of each of the three bank pillars in Germany: cooperative banks, savings banks and all other commercial banks.
As the preliminary descriptive analysis shows, on average, other banks have a consistently higher capital ratio than cooperative and savings banks. Yet, cooperative and savings banks are more homogenous in their level of the capital ratio than all other banks. This dichotomy between cooperative and savings banks on the one hand and other banks on the other hand is reflected in the estimation results of the static fixed effects model. The findings suggest that larger, less profitable and less forward-looking cooperative banks and savings banks, which rely more on deposits, are more invested with collateral and issue more subordinated debt, tend to hold a lower capital ratio, ceteris paribus. Profitability appears to be an important determinant of cooperative banks' capital ratio as expected from their common practice of retaining profits (Gewinnthesaurierung). The results remain robust with respect to the definition of profitability. In the case of savings banks, profitability however only significantly impacts savings banks' capital ratio when proxied by post-tax net profit to total assets. Most of the variation in the capital ratio of both cooperative banks and savings banks is explained by bank-specific time-invariant fixed effects.
The expected determinants of banks' capital ratio have much weaker joint explanatory power when other banks are considered. Risk and collateral do not significantly affect the capital ratio of other banks. Market funding exerts a disciplining influence on their capital management.
The more profitable other banks are the lower their capital ratio is predicted to be. Robustness checks with respect to different measures of the capital ratio and of ex post risk proxies corroborate these findings. The results are moreover invariant to the estimation techniques and controls employed in the static model, i.e. with and without bank FE, time FE and statetime FE.
Despite its simplicity vis-à-vis more complex dynamic panel data models, the static FE model is able to explain differences in capital ratios based on four factors. First, bank specific timeinvariant fixed effects as measured by the rho statistic constitute the main driver behind varying capital ratios. Second, bank (pillar) specific time-variant variables derived from the balance sheet and income statement have significant explanatory power in the case of cooperative and savings banks. Third, time fixed effects, supported by Wald-tests of joint significance and a higher LF, significantly improve the model fit. Finally, state-time fixed effects control for further variation between banks as evidenced by the higher LF when state-time interactions are included. Together these findings imply that banks' choice of optimal capital ratio is much more idiosyncratic than expected. Future data mining activities could possibly yield better balance sheet proxies for the determinants that capture more of the variation between and across bank pillars.
Note also that FE estimation is not superior to the more complex system-GMM estimation. The latter may be able to capture important dynamics in banks' capital adequacy management.