Failure prediction: Sensitivity of classification accuracy to alternative statistical methods and variable sets

The purpose of this study was to assess the sensitivity of the prediction of business failure to the use of different variable sets and different statistical methods. This objective was accomplished by comparing the performance of failure prediction models using four alternative variable sets on firms which failed from 1966–1975. The four sets of variables were those used by Altman (1968), Deakin (1972), Blum (1974), and Ohlson (1980). A linear discriminant model, a quadratic discriminant model, and a logit model were developed for each of the four variable sets. For a given variable sets, the linear and logit models had comparable misclassification rates and performed at least as well as the quadratic models. Using linear discriminant analysis or logit analysis, all four variable sets performed comparably.