Previsão de insolvência: uma abordagem do Modelo de COX

Abstract

The general aim of this study is to use the Cox regression model to identify risk factors for insolvency in Brazilian companies listed on B3 (Brasil, Bolsa, Balcão). Predicting factors that influence insolvency is extremely important for companies and, therefore, various studies have been carried out with the aim of identifying possible risk factors for insolvency. In an attempt to help identify these risk factors, 523 companies were initially selected from the Economatica® database. The companies were selected between 2011 and 2020. Companies that left B3 before January 2011 and those that did not have the information of interest were excluded. The final matrix is a timeless cross-section, i.e. the companies are at different times, but for each company there is data for only one year. The information used was only collected for one year, with the follow-up time recorded and the existence of insolvency or not characterized using the models evaluated. The methodology used was a survival analysis technique called the Cox model. The Cox model was best fitted by the Altman and Kasznar models, with covariates significant at 5%. Most of the covariates used did not contribute significantly to understanding company insolvency in the Cox model. Among the covariates that were statistically significant in the adjusted models, average stockholders' equity and short-term related party liabilities remained in the adjustment of the final model for several insolvency indicators. In addition to the two covariates mentioned above, financial leverage, the net debt/total debt ratio, general liquidity and share sales are other covariates that predict insolvency when using the Cox model in this study.