Methodological Issues in Forecasting Corporate Failure: A Review

Abstract

Forecasting corporate failure has been a hot topic during for more than eighty years. From the univariate model of Beaver (1966), and the Multivariate Discriminant Analysis model of Altman (1968) to models based on Logit, Probit, Artificial Neural Networks (ANN), Bayesian models, Fuzzy models, Genetic Algorithms (GA), Decision trees, Support vector machines, Knearest neighbor, Hazard and Hybrid, model building has evolved during this period with the focus of enhancing prediction accuracy. The literature can be classified into three main methods, namely; statistical methods, intelligent techniques and theoretical approaches to forecast corporate failure. The paper aims to contribute to the existing literature by analyzing methodological problems in the above three areas. A systematic review is performed on 76 articles spanning a period from 1966 to 2018 in scholarly reviewed journals. The results on the SLR indicates that there has been significant prior work in the areas of forecasting corporate failure, but there lacks a sound theoretical view for a highly accurate, simple and widely used model.