National Individual Consumption Loan Credit Rating-models And Empirical Study

Since 1990's, with the rapid development of China's economy, the improvement of residents' income level, the transformation of the consumption idea as well as the impetus of the national credit consumption policy, various individual consumption loans such as credit card, individual auto loan, the durable consumable loan, student loan, individual housing mortgage loan have merged one after another. Although the balance of individual loan of China's commercial bank doesn't account for large amount in the total credit loan, its growth rate is very rapid. The increasing amount of individual credit loan creates demand for the control of individual credit risk. However, due to the incompetence of individual credit data, the faultiness of credit rating technology and the shortage of professional credit rating analyst, the supply of credit rating can not meets the demand in our country. The rapid growth of individual credit loan makes striking contrast with the outdated credit risk management and measurement, so it is an urgent task to establish a risk management system adapting to the growth of consumer credit.The foreign experience indicated that, as an effective individual credit risk management tool, credit rating model played a very important role in the commercial banks' credit decision-making. How can we learn from the foreign advanced experience to make use of the existing individual credit data in the commercial banks' information systems so as to build a individual credit rating model with certain forecasting ability and certain reference value in credit decision-making and also suitable for our national condition, which is very meaningful. This thesis makes some discussions on how to build national individual credit rating models.Firstly, the thesis introduces the basic concepts of credit rating, and then on the basis of the literature review of credit rating home and abroad we point out the shortcomings of national credit ratings. Meanwhile, we propose some betterment methods to improve the traditional models, which are the creative points of this thesis. Afterward, we introduce these new points to some traditional models used in individual credit rating, including the traditional statistical methods such as Discriminant analysis, Logistic regression , as well as the newly developed non-statistical methods such as neural network, multivariate adaptive regression splines etc. Secondly, we use a certain city commercial bank's individual housing installment loans in Zhejiang Province to build several credit rating models. After that, we make comparison among these models in the aspects of accuracy, robust, and explanation ability. Finally, according to the characteristics of different models, we propose a new two-stage hybrid model on the base of utilizing the strongpoint of each model. The empirical analysis proves that the new model is an alternative individual credit rating model with better prediction ability.