Predicting Bank Failures: A Comparison of On- and Off-Site Monitoring Systems

Abstract

On-site examinations are regulators' primary tool for monitoring the financial condition of federally insured depository institutions. In this paper, we assess the speed with which the information content of the supervisory rating assigned during bank exams - the CAMEL rating - decays. This is an important issue because cost and regulatory burden considerations often cause CAMEL ratings to be assigned relatively infrequently. As a benchmark for information content, we use econometric forecasts of bank failures generated by applying a probit model to publicly available accounting data. When compared with all CAMEL ratings available at a given point in time, the econometric forecasts provide a more accurate indication of failure. Further analysis reveals that this overall finding reflects the tendency for a CAMEL rating's information content to deteriorate noticeably beginning in the second or third quarter after the rating initially was assigned.

Keywords: bank examination, bank failure, CAMEL rating, early warning system

JEL Classification: E58, G21, G28