金管會宣布二零一五年九月一日起,將降低雙卡利率的上限(現金卡利率與信用卡循環利率),各現金卡或信用卡業務機構所提出之最高利率不得超過15%。這一項修正,促使銀行面臨顧客管理的風險增加,原本面對較高利率的高風險持卡者,將擁有較低的循環利率,提高呆帳風險或違約風險。預借現金者屬於高風險與高報酬的合併者,回顧先前信用卡相關文獻,學者多以「違約因素」探討為方向,也點出「預借現金」是造成延遲付款的變數之一,卻少有學者研究信用卡預借現金者的逾期行為。不同的消費者擁有不同的行為模式,即使同為逾期違約者也有相異之處,如同產品市場定位,針對既定目標與族群以進行行銷策略的發想,因此本研究目的在於探討信用卡之預借現金者中延遲付款的消費行為差異,並加以分群,以及其影響差異因素為何,也將建構一預測模型以供未來觀察值的預測與分群。本研究之探勘工具為「二階段集群分析」以及「區別分析」。 研究結果發現,以「信用額度」、「預借現金總額」、「預借簽帳比」、以及「曾延遲付款與否」做為分群變數,得到最佳分群數為3群─曾延遲付款1群(周轉不良)與不曾延遲付款2群(緊急借用、理財專家)。進一步以「區別分析」探討哪些因素會影響信用卡預借現金者中曾延遲付款與否的行為,實證發現,人口統計變數(年齡、教育程度、居住地區)以及信用卡相關變數(卡片等級、信用額度、預借現金次數、預借現金總額、曾延遲付款與否)皆會影響信用卡預借現金者在延遲付款上的行為,並依這些區別變數建立預測模型以預測未來觀察值的行為。
Financial Supervisory Commission (FSC) announced that from September 1, 2015, the interest rate of cash cards and revolving interest rate of credit cards shall not exceed 15% per annually. This financial laws will increase the default risk of card holders. Originally, high interest rate fits high risk holders, however, now is changing, high risk holders will face lower interest rate and make the default rate higher. Previously, lots of papers focus on the delinquent factors, and point out “cash advance” is one significant factor of the default behavior. However, less papers focus on the delinquency of credit card cash-advance user. So, this thesis devotes to investigate that the difference between the delinquent credit card cash advance user and those who do not with cluster analysis. And use discriminant analysis to build a model for future predicting. In the result, with “credit amount”, “amount of cash advance”, “cash advance / payment”, and “delinquency” as cluster variables, the investigate turns out the best cluster numbers is three. Further, using the discriminant analysis, demographic variables, such as age, education, living area, and credit card variables, such as class of card, credit amount, numbers of cash advance, amount of cash advance, and delinquency, are the significant variables on the behavior of cash advance of credit card holders. And use those significant variables to set up a predictable model for observing the behavior of future customers.