PREDICTIVE CAPACITY OF INSOLVENCY MODELS BASED ON ACCOUNTING NUMBERS AND DESCRIPTIVE DATA

Abstract

In Brazil, research into models to predict insolvency started in the 1970s, with most authors using discriminant analysis as a statistical tool in their models. In more recent years, authors have increasingly tried to verify whether it is possible to forecast insolvency using descriptive data contained in firms’ reports. This study examines the capacity of some insolvency models to predict the failure of Brazilian companies that have gone bankrupt. The study is descriptive in nature with a quantitative approach, based on research of documents. The sample is omposed of 13 companies that were declared bankrupt between 1997 and 2003. The results indicate that the majority of the insolvency prediction models tested showed high rates of correct forecasts. The models relying on descriptive reports on average were more likely to succeed than those based on accounting figures. These findings demonstrate that although some studies indicate a lack of validity of predictive models created in different business settings, some of these models have good capacity to forecast insolvency in Brazil. We can conclude that both models based on accounting numbers and those relying on descriptive reports can predict the failure of firms. Therefore, it can be inferred that the majority of bankruptcy prediction models that make use of accounting numbers can succeed in predicting the failure of firms.