| 摘要: | 我國的信用卡業務已有相當顯著的成長。然信用卡不僅只是一種支付工具，也屬於免擔保的信用融資，對發卡銀行言，風險很高。中華信評於2003年底也指出，信用卡逾放損失通常會延後二年浮現。可見，除有效地篩選客戶避免違約發生外，對於逾繳帳款的處理也應被銀行高度重視。基於此，本研究旨在探討信用卡業務的徵審過程，如何影響持卡戶違約，並且欲瞭解影響持卡戶每月繳款狀態移轉的因素為何。本文以國內某知名金融機構的信用卡客戶為研究對象，並將研究分成兩個部份：(1)分層隨機抽取3萬戶：核卡戶24,000筆與拒絕戶6,000筆，核卡戶又細分為23,398筆正常戶與602筆違約戶。參照Jacobson-Roszbach(2003)的作法，利用Bivariate Probit 迴歸模型解決選樣偏誤問題後，試圖探討申請者或持卡戶的人格特質，擷取自金融聯合徵信中心的信用資訊，及與發卡銀行往來關係三大構面20項研究變數，如何顯著影響信用卡業務的徵審過程與持卡戶違約；(2)整合Smith-Lawrence(1996)、Lin-Yang(2000)，以及羅文綺(2002)的研究，以2002年7月至2004年2月，共20期為研究期間，分層隨機抽取3萬筆開戶日期為2001年1月1日至2001年12月31日的持卡戶，並納入持卡戶的人格特質、與發卡銀行往來關係，及繳款歷史三大構面22項變數，利用馬可夫鍊機率移轉矩陣與Multinomial Logit迴歸模型，進一步剖析影響持卡戶繳款狀態移轉的因素。 主要實證結果發現：(1)學歷與職位兩項變數對信用卡審核呈正向關係，卻對持卡戶違約呈負向關係，顯示銀行為提高收益，未視降低違約風險為第一要務。銀行徵審時除以學歷、職位為衡量標準外，應更加注意財力相關證明，以免造成職位相當，經濟水準差異極大而出現違約的窘境；(2)核卡額度與卡別並非Bivariate Probit 模型的關鍵因素，表示樣本銀行傾向不浮濫核予額度以獲得更高報酬，此與Jacobson-Roszbach(2003)論調不謀而合；(3)輕微逾期的可能回收率近9成，然使用循環信用償還逾繳款者，將是影響未來回收率高低的關鍵，且管理者應主動避免持卡戶發生長期逾繳，勝於事後催討行動；(4)北區、中區、信用額度越高之持卡戶，習慣全額繳清消費額，南區持卡戶則充分利用循環信用方式繳足最低額，值得列南區持卡戶為推展信用卡業務的主要目標；(5)金卡戶容易發生輕微的遲繳，但這些欠款也較容易收回。反觀普卡戶，雖是信用卡業務循環利息的主要收入來源，一旦發生逾繳，持續積欠帳款的機率也較高，回收墊付款的機率相對較低。 Credit card market is well-developed during these 30 years in Taiwan. However, credit card is not only the paying tool, but a kind of non-asset backed loan which would make the banks suffer from great credit risk. Also, according to the research from Taiwan Ratings at end of 2003, loss given default from credit card usually will not appear until two years later, meaning that except telling good from bad applicants, banks should put emphasis on how to deal with the delinquent payment, too. The contribution of this research is to find out how the card granting mechanism affects cardholders’ default as well as what kind of factors would affect the monthly payment transitions. Dividing the thesis from two parts: (1)Stratified random sampling 6,000 rejected and 24,000 approved applicants including 23,398 good and 602 bad cardholders. Based on Jacobson-Roszbach(2003),we use Bivariate Probit Model to solve the sample selection bias problem and discuss the factors both influence the granting mechanism and cardholders’ default; (2)Integrating the concept of Smith-Lawrence(1996) and Lin-Yang(2000), we sampled monthly data for 30,000 cardholders, who open their credit card during 2001,from July 2002 to February 2004 to have a research on the issue of monthly payment transitions by Multinomial Logit Regression and Markov Chain. The main results of this research as following described: (1)Two variables, education and job position ,have opposite coefficient in the loan granting and default equations. That means the bank was not minimizing credit risk, but willing to increase the revenue more; (2)Loan size and the class of credit card both have no significant influence on default risk showing that the bank wouldn’t grant bigger loans because they are riskier which is the same with Jacobson-Roszbach(2003)’s result; (3)Around 90% of slightly delinquent payment loans will transit to normal payment ones, but the cardholders using the involving interests to avoid default would greatly influence the recovery rate. The bank should prevent these loans from getting worse instead of collecting the loans when they have already been delinquent seriously; (4) As we mention cardholders’ living area, no one would tend to use the involving interests to pay the monthly bill but the ones living in south of Taiwan, so it ‘s worthy to focus on this kind of cardholders to expand the earnings; (5) Gold cardholders are more likely slightly delinquent, however, the recovery rate is higher. Moreover, although the income of involving interests is mainly from general cardholders, the probability of remaining staying in default conditions is also greater and that would cause the lower recovery rate for the bank. |
