Industry and Time Specific? An Examination of Bankruptcy Forecasting

Abstract

The purpose of this research is to determine whether bankruptcy forecasting models are subject to industry and time specific effects. A sample of 15,848 firms was obtained from the Compustat and CRSP databases, spanning the time period 1950 to 2013, of which 396 were bankrupt. Using five models from the literature, we evaluated whether these effects impacted on the forecasting capability of the models. We found that three of the models demonstrated varying levels of success depending on the decade and industry in question, with Altman being affected significantly. The Zmijewski model outperformed both Altman and Ohlson, and also had the lowest overall change in predictive performance. In addition, based on the results of the older models it appears that the inclusion of market information increases the volatility of model performance.