隨著房地產市場熱絡,金融機構家數增多,銀行業為配合建商取得整批分戶貸款的業務及其他消費者的需求,除了加快核貸速度,提供更好的服務品質以符合客戶的需求,因此如何快速篩選出客戶的狀況,並幫助授信人員作出正確的決策,以減少房屋貸款逾期的發生。故本研究基於銀行授信審核及承作條件觀點,依貸款戶所提供之資料及授信審核時所考慮的各項變數進行實證分析,期能發現影響房屋貸款逾期發生之主要因素,提供授信部門作有系統的風險控管,以降低房貸違約風險及金融機構的損失,進而增加授信品質、速度及效率,提升銀行實質的收益及整體經營績效。 本研究以桃園地區某銀行於2007年至2011年所承做672件房屋貸款之授信案件為研究對象(逾期戶94件、正常戶578件),透過卡方檢定與羅吉斯迴歸模式分析,影響房屋貸款案件逾期發生因素與機率,實證結果顯示學歷、職業別、年所得、負債比、擔保品鑑估價值、貸款金額、貸款成數、寬緩期等因素,對房屋貸款授信案件發生逾期有顯著的影響,本研究實證模型對正常戶、逾期戶及整體房屋貸款案件之預測準確率分別為94.6%、71.2%及91.3%;若自變數採用未經調整的原始數字,將使預測準確率提高,尤其逾期戶的預測準確率將提高為78.7%。
The number of financial institutes is increasing as the estate market is heating up. For getting the entire builders mortgage loans business and other consumer needs, the financial industry will quickly approval loans for providing better service quality to meet consumer’s needs. Therefore, it is the issue that how to quickly screen out client’s situation and help loan officers to make the right decisions in order to decrease the overdue of mortgage loans. This study wants to investigate the affecting overdue factors to distinguish its implicit risk and know its attribute for giving the quantized indicator and the basis of making correct decision in order to reduce loan risk, non-performing loan ratio of mortgage loans. Therefore, financial institutes can raise their physical income and whole operational performance by means of increasing the quality, speed, efficiency of bank credit granting. Chi-Square Testing and Logistic Regression model will be applied to analyze factors and the probability of affecting mortgage loans overdue based on the data, 672 mortgage loans (578 of normal and 94 of abnormal) of a bank located at Taoyuan district during 2007 – 2011. The study found that seven significant factors, such as receive an education degree, occupational class, annual income, debt ratio, estimated value of collateral, loan amount, loan percentage, grace period have apparently influence on overdue probability of mortgage loans. The prediction accuracy rate of normal, abnormal and total mortgage loan cases is 94.6%, 71.2%, and 91.3% respectively. The prediction accuracy of mortgage loans will be improved if the data of independent variables were unmodified. Especially the prediction accuracy of abnormal mortgage loans will increase to 78.7%.