An Early Warning System of Financial Distress Using Multinomial Logit Models and a Bootstrapping Approach

Abstract

This study adopts multinomial logit models to separately measure the extent to which financial ratios and corporate governance signal the likelihood of "slight distress events" and "reorganization and bankruptcy." The results show that corporate governance variables are closely related to the occurrence of "slight distress events." The estimated misclassification costs of the 1,000 resamples generated through bootstrapping procedures are statistically lower for a model that makes use of corporate governance (CG model) than one without corporate governance (non-CG model) at all cutoff points in 2009, and cutoff points from 0.11 to 0.27 in 2008. Since corporate governance is incrementally useful in predicting financial distress, the CG model's predictive ability improves as two corporate governance factors are considered: ownership ratio of insiders and pledge-ownership ratio of insiders.