Simple Forecasts of Bank Loan Quality in the Business Cycle

Abstract

Experience from models such as SEER suggests that bank financial condition predict bank failures. However, it has been difficult to find a relationship between macroeconomic variables and bank failures. This paper shows ways in which simple time-series techniques can be used to forecast financial conditions of banks. The models include macroeconomic variables in order to consider systemic cyclical factors in forecasting. In addition, analysis of regression residuals is used to obtain relatively early warnings of unusual performance. ; The empirical result suggest that a limited number of regional and national macroeconomic variables are often good predictors for problem-loan ratios, and that simple, bivariate VAR systems of one bank variable, one macroeconomic variable, and seasonal dummies can be quite effective. These variables include bankruptcy filings, farm income (particularly for states where farming has important role), state annual product, housing permits, and unemployment. Analysis of the residuals is shown to be an interesting tool to detect unexpected changes in past-due loans. Impulse-response functions are a result of VAR estimation, which can be used for scenario analysis.