| Summary: | 碩士 === 國立中興大學 === 應用經濟學系所 === 95 === This research mainly explored the credit of borrowers in a financial institution house loan department. The influencial variables from literatures were identified and classsified into the credit five P principle structure planes. To establish the borrower credit evaluation model, a logistic regression method was used to select the important influencial factors first, and then the credits of the borrowers were divided into different credit ranks, each with a corresponding normal (default) anticipated probability. So that we can take the evaluation as the base of giving loans. By applying this model, the more effective loan evaluattion, the reduced credit risks and the enhanced quality of house loans approval can be achieved.
The empirical study showed that there were six variables with significant correlations at the level of 10%. The four variables of gender, loan to value ratio, card whether circulation credit and debt burden ratio reveal a negative correlation with the normal repayment probability. The income and education level variables show a positive correlation with the normal repayment probability.
Variables of gender and education level were significant in People structure plane. Variables of income, debt burden ratio, and card whether circulation credit were significant in Ability to pay structure plane. Only loan to value ratio was significant in Creditor''s rights Protection structure plane.
This research findings are consistent with the default risk elementary theory. The debt burden ratio, card whether circulation credit, and income three variables may represent the borrower''s "Ability to Pay" and the loan to value ratio variable may represent "Equity" of house.
Moreover, this research may induce significant variables into three categories simply, the Ability to pay (including debt burden ratio, card whether circulation credit, income), Willingness to pay (including loan to value ratio) and others (including gender, education level). The Logit regression prediction with the accuracy of 97.3% shows high value of our model.
Finally, by dividing the samples of 476 house loan households into risk danger classes of A to F according to their probabilities of breaking the contract, and clustering characteristics of each class, the financial institution house loan department can effectively decide the loan amount and price.
|
|---|
