| Summary: | 碩士 === 東吳大學 === 經濟學系 === 94 === Using a commercial bank’s database on its product of the small business custom loan, this study is to analyze the default probability of that product with the logit model and then attempts to build an effective credit scoring system. There are several major findings in the study. First, we find that six kinds of variables might affect the default odds and help estimate probability of default (industry groups, years in business, sales, checking account amount in the past three months, enterprises payment record on other loan products inside bank, and inquiries made to JCIC in the past three months). Second, based upon the PD estimation, the model found that the seven items on the bank’s traditional credit-screening statement have resulted in significantly different average PD. These items include industry groups, type of business, years in business, capitalization, sales, debt amount, the ratio of income by debt, and the initial weights in the statement. However, these average PD are inconsistent with their weights initially designed in the statement. Therefore, it is apparent that the traditional credit-screening statement designed by the bank’s credit management has been insufficient in measuring the risk of small business custom loan. Moreover, there exist obvious discrepancies in the average PD for two extra items: the number of the useful credit cards and the sources of client. They should be incorporated into the credit scoring system as a part of the effective credit risk management. Finally, the study found that the target enterprises which are characterized by bigger size and longer years in business, and engage in manufacturing and trade business have higher covert default risk. These results also contradict to the weights initially expected in the credit-screening statement. Therefore, the traditional credit-screening statement designed in accordance with experiential rules of experts could be modified with the quantitative analysis as suggested in the study to develop an effective credit scoring system for small business loan.
|
|---|
