Bank risk classification and optimal regulatory choice

Abstract

A theory of bank regulation is formed in this study by choosing an optimal classification scheme so as to minimize specific costs with assumed fixed regulatory instruments and relative costs. The likelihood of failure of a financial institution can be estimated using the financial data of that institution. Previous research studies have attempted to predict the probability of failure by using one year of data or lagged data. In these studies, a bank on a failure trajectory was counted as a nonfailure until it actually failed. The estimation was biased in favor of nonfailures, meaning that a failing bank was more likely to be classified as a survivor. This study develops a multinomial ordered logit model which uses several years of data to classify banks into a larger number of categories. Instead of just predicting banks that will fail in the following year, it can predict the probability of failure within multiple time periods. The major empirical results of this study state that the probability of failure of financial institutions can be estimated using a multinomial ordered logit model. Financial ratios based on capital, assets, total loans, nonaccruing loans, loans 90 days past due and net income were found to be significant variables in predicting failure probabilities. The results present evidence that banks can be classified into high or low risk categories which could be used by regulators to minimize the costs of regulation and bank failure. Better predictive ability would allow regulators to take action sooner to assist banks in maintaining solvency and reduce the number of failures and their associated costs.