Probability Of Default Model And Empirical Study In Commercial Banks

According to the requirement of the New Basle Capital Accord, it is necessary to carry out ERM(the enterprise-wide risk management) for those banks applying IRB(internal rating-based approaches). To attain this aim, banks must set up their own ERM. Among all risks banks face, credit risk is of course the most important. This paper focuses on a model for credit risk management-PD model that is widely applied in practical exercises, and further explores empirical study.Moody’s KMV model is a well-known model for public firms in terms of internal credit rating, and is also world-wide recognized by industry, more details about this model will be illuminated in Chapter Ⅱ. As for those private firms whose right information is not easily accessible, in comparison, KMV model is not applicable. With the development of statistical methods, western banks began to engage on internal rating model research, a number of statistical and econometric tools have been introduced in the field of credit rating, such as Discriminant Analysis, Linear Regression Analysis, Mathematical Programming Method, Nearest Neighbor Methods, Logistic Regression, Expert System, Decision Tree, Artificial Neural Network, and Genetic Algorithm. However, Logistic regression method is no doubt the most popular model in the industry, almost all the banks are in favor of applying it not only in corporate rating, but also in retail rating. For example, Then Standard&Poor’s MEU model is a sort of mutation of logistic regression, while Moody’s RiskCalTM prefers Probit regression.This paper pay more attention on Logistic Regression method, and in the end the historical credit data from some American commercial bank will be taken advantage for empirical study, in order to rationally understand this model.