Financial Distress prediction in the Netherlands: An Application of Multiple Discriminant Analysis, Logistic Regression and Neural Network

The main goal of this thesis is to adjust existing prediction models to achieve an accuracy of 80% or higher in the Netherlands. This thesis gives an application of the three most common used financial distress prediction methods. The Multiple Discriminant Analysis (MDA), Logistic Regression (Logit) and Neural Networks (NN) method are used to predict whether a Dutch company becomes bankrupt. The model of Altman (1984) is tested, re-estimated and improved in this study. The models are compared with the Logit and NN equivalents to asses which model has the best performance in the Netherlands. The models are tested on a sample of 125 matched pairs of Dutch bankrupt and non-bankrupt companies. The results indicate that adjusted prediction models perform better than the model of Altman (1984).