Statistical modelling in credit rating

Abstract

Nowadays, investors are more willing to make investment on exchange market to gain much more superior returns. There is an increasing demand on credit information of corporations. However, there are only a few of Hong Kong corporations have been evaluated by credit rating agencies. Therefore, U.S. non-finance corporations listed in New York Stock Exchange are adopted in model development. Various individual statistical methods, namely multiple discriminant analysis, ordinal logit model, multinomial logit model, ordinal probit model, and neural network, and combining forecast introduced by Kamstra, Kennedy and Suan (2001) have been adopted to predict S&P’s credit ratings. In extending the Kamstra, Kennedy and Suan (2001) combined forecast method to combine the probability forecasts in probability space; it is found that the modified KK method in probability space outperforms the individual classification methods and the original KK combining forecast methods. In addition, another well-known credit assessment corporation – KMV Corporation, has developed another model to evaluate the default risk of public company. It is found that KMV model has advanced power in predicting default risk than S&P credit rating.

Date of Award	15 Jul 2005
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Miu Han Iris YEUNG (Supervisor)