碩士 / 國立臺灣海洋大學 / 航運管理學系 / 105 / For the time being, the market of real estate remains at the downturn; nonetheless, competition among banks has turned ever fiercer for mortgage business is considered as one of the major source of revenue of banks. As such, the quality of mortgage credit matters greatly the stability of financial environment and sustainable operation of banks. To reduce the loss incurred from bad debts and increase revenue, it is necessary that a comprehensive set of risk evaluation model on mortgage credit to assess credit risk, thus helping to trim down the generation of non-performing loan.
In view of the both local and oversea scholars, they have, mostly, resorted to either merely Logistic regression or discriminant analysis for study. But for scenarios in research of mortgage default, only some of the scholars would, concurrently, employ both measures in order to compare whichever method should come out with better prediction rate. As for this study, it has, for the first time, utilized these two research methods to conduct screening significant variables. With the build-up of such evaluation model, it should, then, founded as its basis for credit risk measurement of mortgage loan business, in the hope that it can be used to enhance credit quality as well as the derived loss of bad debts from risk of overdue default.
This study has taken the branch of a professional bank on domestic real estate as its subject of study, while cases of mortgage loans completed of drawdown processes totaling 306 cases are employed as its population ranging from November 2006 to November 2016. Among these cases, there are 60 cases of households which have already exceeded loan payment for more than 1 month, while 60 other cases chosen at random are within time limits, totaling 120 cases. Furthermore, 50 households are selected from each of the former 50 and latter 50 cases of households aforementioned, making use of Logistic regression as well as discriminant analysis to set up a set of risk evaluation model of mortgage credit. And results are obtained as based on this substantive study:
1.Characteristic variables that should affect mortgage default of case bank would, according to the intensity of their impact, rank respectively as "provision of withholding certificate or payroll transfer and other substantive information," "times of new business inquiry for the last three months with Joint Credit Information Center," "amount of this loan," "location of the collateral," "grace period of principal," and "times of revolving credit rate of credit utilized within a year." Among these variables, the two items as "provision of withholding certificate or payroll transfer and other substantive information" and "times of revolving credit rate of credit utilized within a year" are of the independent variables taken into the prediction model for the time as based on practical experiences, given with much significance.
2.If Logistic regression is employed only, the accuracy rate of overall classification should achieve 55%; however, if only discriminant analysis is utilized, the accuracy rate of overall classification will go up to 75%. Nonetheless, if both methods are employed, being the methodology used in this study, the accuracy of overall classification should reach 80%, telling that the prediction model built up in this study should fare with better accuracy of overall classification as compared to other analysis models.
| Identifer | oai:union.ndltd.org:TW/105NTOU5301031 |
| Date | January 2017 |
| Creators | Wu, Chia-Ming, 吳嘉明 |
| Contributors | Chou, Heng-Chih, 周恆志 |
| Source Sets | National Digital Library of Theses and Dissertations in Taiwan |
| Language | zh-TW |
| Detected Language | English |
| Type | 學位論文 ; thesis |
| Format | 59 |
Page generated in 0.0013 seconds