PREDICTION OF THE INSOLVENCY COX MODEL: AN APPLICATION FOR RISK ANALYSIS OF BRAZILIAN PUBLIC COMPANIES

Abstract

The first studies on failure prediction were developed in the 1930’s. However, the subject gained impulse only in the 1960’s, when statistical techniques started to be used. In Brazil, the first works on the subject date from the 1970’s. It is worth pointing out that the statistical technique used in the majority of these works was the multivariate linear discriminant analysis. In an attempt to contribute to the theme, this work proposes to test a model of prediction of corporate bankruptcy based on the model developed by Cox (1972). This model differs from those based on logit, probit and discriminant analysis techniques, because it provides not only the probability of an event happening in the future, but also an estimate of the time until it occurs. The results show that only two indicators were considered relevant to calculate the risk of bankruptcy: SHORT TERM BANKING LOANS/CURRENT ASSETS and RETURN ON EQUITY (ROE). The model demonstrated that it is possible to identify, in advance, whether or not a corporation is likely to go bankrupt and that the Cox Proportional Hazards Model can be used as a tool in the task of predicting the risks of bankruptcy faced by corporations operating at Bovespa.