Improving model averaging in credit risk analysis

摘要

In this paper we compare classical and Bayesian Model Averaging (BMA) models for logistic regression in credit risk. We also investigate regression models for rare event data, using Generalised Extreme Value regression (GEV). On the basis of a real data set, we show that Bayesian Model Averaging models outperforms classical regression in terms of percentage of correct classifications and related performance indicators. We show that Bayesian Model Averaging is a technique designed to help account for the uncertainty inherent in the model selection process, something which traditional statistical analysis often neglects. In credit risk, by averaging over many different competing models, BMA incorporates model uncertainty into conclusions about parameters and prediction. In this paper we report also the empirical evidence achieved on a real data sample containing rare events provided by a rating agency.