Financial Distress Prediction - A Comparison of Statistic-, Econometric- and Artificial Intelligence Approaches

This paper shows the evolution of financial distress prediction models of the past four decades. Special attention is paid to linear discriminant analyses, logistic regression analyses and neural networks. Based on accounting data of 50 UK industrial firms, prediction models are estimated using the mentioned classification methods. A fivefold cross-validation procedure is applied to reduce sampling bias and two methods are applied to minimize costs of misclassification. The paper finishes with a comparison of the models' predictive accuracy