以大样本数据的cohort study探讨消费性小额信用贷款的逾期因素

A Cohort Study of Credit Risk of Small-scale Consumers Loans in Large Sample Size

金融机构获利的稳定,健全与持续的授信质量影响很大,全球性新巴塞尔资本协议(New Basel Capital Accord)的推动及主管机关要求金融机构改善资产质量及合并政策的影响下,使台湾的金融业进入激烈合并竞争的时代。消费性贷款随着社会趋势的发展日益增大,为了探讨消费性小额授信户逾期还款行为。这里运用台湾地区两家商业银行1996—2001年间的授信户样本数据,采用retro-prospective cohort study以修正前学者采用Case-control study的限制,及配合新巴塞尔资本协议(New Basel Capital Accord)下金融全球化的规定。结果发现样本数大小会影响logis-tic模型的显著性因素及适合度。另外也发现Logistic regression模型预测能力,其逾期案件预测结果正确预测率达70.0%,66.95%及71%,另在整体正确预测率从67.3%和69.85%增至72.1%。因此,以大样本retro-prospective cohort study所发现的影响因素及预测能力的稳定性,较符合新巴塞尔资本协议规定及科学程序。

The stability of profits made by financial organs is subject to the sustainability of their credit. As the New Basel Accord is to be implemented globally, the banking industry will encounter increasing competition and need to improve the quality of the assets. Many banks will be confronted with the dilemma of reducing default rates and increasing return rates （on assets and on equity）. Since small-scale consumers loans are relatively profitable and their risks are more diversified, they are preferred in accreditation over business loans. Because the cohort study re- flects the study design of the real date, which is stipulated in the New Basel Capital Accord, the above findings can serve as reference for assessing credit risk, whereas these findings would not be obtained in the previous case-control studies. This research intends to use a retro-prospective cohort to analyze the significance of factors affecting default behavior. Unlike previous case-control studies, the cohort study can assess some elements of credit risk. There are two findings in this study. First, the sample sizes would affect the number and the choice of the explanatory variables in logistic regression models. Second, the predicted power increases significantly with the sample sizes.