Prediction of Corporate Bankruptcy: An Evaluation of Accounting-ratio-based Bankruptcy Models

Abstract

Bankruptcy is costly, especially for corporate management. Therefore, it is of great importance to identify relevant financial indicators that predict bankruptcy risk prior to an actual insolvency state. Many researchers have pointed out stationarity concerns of bankruptcy prediction models that negatively affect the accuracy of previously developed model. Thus, advocate for a constant re-estimation. This study proposes a new 6-factor logit model that has shown a higher forecasting ability in comparison to Altman’s (1983) Z’-score and Ohlson’s (1980) O-score models and its re-estimations. Nevertheless, re-estimations of the existing bankruptcy models’ coefficients and its cut-off scores also greatly enhance their accuracy. The analysis is based on private manufacturing firms in France from the period 2012 to 2019. The most important financial ratios in bankruptcy models are identified and used in a backward elimination procedure based on the Akaike’s Information Criterion (AIC) to derive the new proposed model. The accuracy of all investigated prediction models is evaluated by classification tables, receiver operating characteristic (ROC) curves, and Type I and Type II errors. This study also evaluates the interpretation of the models from a managerial perspective and places the Francespecific results in the context of existing literature.