現金卡是屬於高風險且高報酬的消金業務，而目前發卡行針對核卡准駁之依據，均依照審查人員經驗之判斷。針對持卡戶潛在的信用風險，如私人借貸或還款意願薄弱等風險，單靠制式的書面信用評分表，其評估效果卻是極其有限；若是再加上審查人員的情緒或其他因素影響，其風險更大為提高。本研究以正常及逾期授信戶各二百組樣本透過逐步迴歸、t檢定及卡方檢定篩選顯著之變數並採用羅吉斯迴歸(Logistic Regression)分析，建立可供參考之模型。 實證結果顯示具有女性、已婚、年齡愈大、持多卡(含現金卡及信用卡)、曾使用信用卡預借現金、非義務役軍人等特質之申請人，逾期比率偏低；男性、年齡偏高仍單身者，逾期比率偏高，其預測準確率達74%。 Belonging to the affairs of the consumptive behavior and finance, cash cards have high risk as well as high reward. Currently, the approval and the disallowance of the card issue by the bank follow the experience and the assessment of the bank censor. Nevertheless, due to the potential risk of the card-holders, the result of the assessment is extremely limited if the bank only follows the written tabulations of the credit assessment. In addition, the emotion of the bank censor and other factors will even raise the risk. This study uses the analysis of Logistic Regression to conduct a proper model to provide the bank with real assistance in processing the issue of the card and to redeem the negligence and insufficient experience of the bank sensor. The empirical result manifests that the applicant who is female, is married, is senior, possesses many cards (including cash cards and credit cards), used the credit cards to borrow cash before, or is non-conscripted soldier, has relatively lower payment overdue rate. The result of this analysis is important for evaluating credit risk, which provides the bank with a proposal of an early warning system for cash cards.
