With One Shot, Which Bullet Would You Use? : Market versus Accounting Data in Bankruptcy Prediction - Part III (The Strong Case of the Use of Neural Networks in Bankruptcy Prediction)

In this paper, we test the predictive power of neural networks to predict corporate bankruptcy. In contrast with the previous literature, we not only use nontraditional models and employ accounting ratios, but also market and microstructure variables. The most important findings are that market variables alone are best at correctly classifying bankrupt firms, the multi-layer perceptron is the preferable neural network model due to the nonlinear relation between predictive variables and bankruptcy, and the multi-layer perceptron has higher classification rates for all different input sets and years