Predicting Bank Loan Defaults: Statistical Versus Bank Loan Officers’ Evaluation

Abstract

This empirical study compares various ways of forecasting company failure.  Multiple discriminant analysis (MDA) was used to come up with a mathematical model using financial accounting ratios as variables to predict business firms’ capability to repay loans, thus classifying firms as “non-failing” and “failing”.  Four ratios were identified as good discriminators between would-be failures and non-failures.  These ratios, plus the current ratio, were used by bank loan officers in the Philippines and the United States for application in their financial evaluations.  The study came up with interesting results in the performance of the MDA model and the loan officers in failure prediction.

The overall predictive ability of the MDA model was higher than that of the loan officers.  In particular, the MDA model fared better in predicting the failing firms.  There is a basic agreement between the Filipino and U.S. lending officers except that the Filipino officers were slightly more accurate in predicting failing firms than their American counterparts.  Experience during the early 1980s made the Filipino loan officers more sensitive to this issue.