Using neural networks vs. multiple discriminant analysis to forecast bond rating changes

This paper presents the results of a study comparing the ability of neural network models and multiple discriminant analysis (MDA) models to predict bond rating changes and to exam if segmentation by investment grade improves classification. Data was collected on more than 900 bonds that had their Standard and Poor's Corporation rating changed during the period 1997 to 2002. This was matched this dataset with corresponding firms which had the same initial bond rating but which did not change. The correspondence was based on the firms being in the same industry, having the same rating at the time of the change (the time frame was one month) and the same approximate asset size (within 20%). This relatively stringent set of criteria reduced the data set to 282 pairs of companies. A neural network model and a multiple discriminant analysis were used to predict both a bond change and the general direction of a movement from a particular bond rating to another bond rating. The predictive variables were financial ratios and rates of change for these ratios. In almost all cases, particularly for the larger sample studies, the neural network models were better predictors than the multiple discriminant models. The paper reviews, in detail, performance of the respective models, strengths and limitations of the models – particularly with respect to underlying assumptions- and future research directions.