An Empirical Study of Credit Rating Model to Evaluate Auto Loans

• Main Authors: SU, YU-HSING, 蘇育興
• Other Authors: LIANG, LING-CHING SOPHIE
• Format: Others
• Language: zh-TW
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/15570089642196069644

碩士 === 國立臺北大學 === 合作經濟學系 === 91 === With the increase of consumer’s credit extension and personalized financial management, installment has become a strategy of car promotion. Among consumer loan products, the features of auto loans are middle risk and return. While the banking environment becomes more competitive and the gap of interest is decreasing, more banking managers participate the auto loan market. In order to minimize risks so as to cut down dues and earn more profit, it becomes more important for banks to develop and apply loan credit rating model. The purpose of this paper is building prediction model to evaluate the risk of auto loans. We collected 312 normal samples and 86 bad-loan samples (principal and interest overdue for one month or longer) from the northern branch office of a newly established bank in Taiwan from 2000 to 2001, applying Logistic regression analysis in order to establish a final and efficient auto loan credit rating model. Among that, credit rating model is used to classify borrowers into regular group and default group and used for bank as a multi-level discrimination to investigate borrower’s credit more flexibly. The empirical study led to the following findings: （1）There are six significant factors in this model: repayment method, living years, relationship between borrower and guarantor, loan to value ratio, credit history, occupation. We could use six significant factors to forecast credit risk in auto loans. （2）The of importance of six significant factors are ordering as following: repayment method, credit history, loan to value ratio, living years, occupation, relationship between borrower and guarantor. This order could help practical evaluation in the baking management. （3）Appling Logistic regression to analyze database, the predictive correct rate reach 95.5%. There is a little different result between bad-loan and normal-loan accuracy of prediction: 89.5% for bad-loan cases and 96.5% for normal-loan cases. In summary, in order to improve the credit quality and enhance the bank''s competitive advantage, it is important to develop a specific, simple, and fair loan credit rating model. The result of this paper shows that adding some factors out of the credit form can greatly improve the ability of predicting and accuracy of Logistic regression model. According to above, it shows the model constructed by this procedure is useful to all of banks and finance companies.