Too-Big-To-Fail and the Modeling of Bank Distress

This study develops a model for predicting distress events among large banks. We conjecture that a bailout possibility induces different behaviors among small and large banks and therefore separate models may be need for early warning models. We indeed find that a failure prediction model for large banks is considerably different from that for small banks. Major bank-level fundamentals show opposite conjecture directions for large banks compared to small ones. The Tier 1 capital ratio, which is under scrutiny of regulators and investors, has almost no distress-prediction power among large banks. However, banks rescued by governments tend to maintain a lower Tier 1 ratio. Interestingly, the cost of funding in large banks is negatively correlated with the probability of failure, possibly reflecting the tendency of SIFIs to be rescued. In conclusion, early warning models for bank failures should be separate for small and large banks