This thesis is a collection of three essays with a common theme on the riskiness of banks. The first essay (Chapter 2) examines how banks' discretionary accounting practices affect banks' stock trading liquidity and stock price crash risk. Bank managers can use discretionary loan loss provisioning practices to obscure information on banks' performance and riskiness, thus elevating information uncertainty among investors. We find empirical evidence confirming that banks with greater opacity, as measured by the discretionary loan loss provisioning (DLLP) practice, are associated with greater stock trading illiquidity and higher stock price crash risk.
The second essay (Chapter 3) examines depositors' tendency to apply financial discipline to banks that adopt discretionary accounting practices. We find that uninsured deposit growth is negatively associated with banks' DLLP. The findings suggest that depositors punish banks' discretionary accounting practices. We also find that within the non-too-big-to-fail (non-TBTF) banks, depositors react sensitively to DLLP during both the 2008 financial crisis and non-crisis periods. However, for those so-called too-big-to-fail (TBTF) banks, depositors react sensitively to DLLP practices only during the non-crisis period, but not during the crisis period. In addition, banks subject to U.S. government bailout actions during the 2008-2009 crisis period received less depositor discipline than other banks.
The third essay (Chapter 4) examines whether tail risk for respective real (i.e., non-financial) sub-sectors appears to be affected by risk emanating from the financial sector. We use the Conditional Value-at-Risk ("CoVaR") approach to examine which real sub-sectors are more vulnerable to risk emanating from the financial sector by measuring the change in CoVaR in each real sub-sector when the financial sector moves from a normal state to a stressed state. We find that the spillover effect from the financial sector to real sub-sectors reached the highest levels in our sample period before the 2008 financial crisis and dramatically decreased once the financial crisis had started. By comparing the back-testing results of our VaR and CoVaR estimations, we also find that the CoVaR measure can be a more reliable indicator of risk during very volatile market periods than the traditional VaR measure.