Working Paper: NBER ID: w16178
Authors: Andrew Ellul; Vijay Yerramilli
Abstract: In this paper, we investigate whether U.S. bank holding companies (BHCs) with strong and independent risk management functions have lower enterprise-wide risk. We hand-collect information on the organizational structure of the risk management function at the 74 largest publicly-listed BHCs, and use this information to construct a Risk Management Index (RMI) that measures the strength of organizational risk controls at these institutions. We find that BHCs with a high RMI in the year 2006 (i.e., before the onset of the financial crisis) had lower exposure to private-label mortgage-backed securities, were less active in trading off-balance sheet derivatives, had a smaller fraction of non-performing loans, and had lower downside risk during the crisis years (2007 and 2008). In a panel spanning the 9 year period 2000--2008, we find that BHCs with higher RMIs have lower enterprise-wide risk, after controlling for size, profitability, a variety of risk characteristics, corporate governance, CEO's pay-performance sensitivity, and BHC fixed effects. This result holds even after controlling for any dynamic endogeneity between risk and internal risk controls. Overall, these results suggest that strong internal risk controls are effective in restraining risk-taking behavior at banking institutions.
Keywords: risk management; banking; financial crisis
JEL Codes: G01; G21; G32
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| strong internal risk controls (G38) | lower exposure to private-label mortgage-backed securities (G21) |
| strong internal risk controls (G38) | lower risky trading assets (G19) |
| strong internal risk controls (G38) | healthier loan portfolios (G51) |
| one standard deviation increase in RMI (C29) | 0.43 standard deviation decrease in downside risk (G19) |
| one standard deviation increase in RMI (C29) | 0.62 standard deviation decrease in tail risk (C46) |
| one standard deviation increase in RMI (C29) | 0.6 standard deviation decrease in aggregate risk (C46) |