A Study on the Significant Factors for Default in Residential Mortgages

[[abstract]]我國於2002 年通過「金融資產證券化條例」後，銀行業資產中佔大宗之不動產抵押債權將可重新包裝後以證券之型態售出，在目前國內金融資產證券化機制中尚無保證或保險之配套要求下，影響不動產抵押債權評價最重要的影響因子為借款人違約，也因此引發市場對不動產抵押債權價格衡量及風險評估之重視。本研究旨在應用羅吉斯迴歸模型探討影響國內不動產抵押債權違約之因素及其重要性。在蒐集資料時特別考量國內某大型行庫之住宅貸款「已結案」資料，以免低估違約機率，並採用Akaike’s information criterion（AIC），Bayesian information criterion（BIC） 及接受者作業特徵曲線（ROC curve ）統計方法探討總體經濟、借款人特性及貸款契約等因素之顯著性比較分析。實證結果顯示AIC 及BIC 訊息指標可改善之前文獻使用羅吉斯迴歸模型時，類別自變數無法與其他影響因子相互比較，及連續型自變數缺乏一致性標準化單位時無法比較之缺點；更進一步發現，加入「總體經濟因素」對住宅抵押貸款逾期的影響後，在模型適合度及模型之預測上皆高於單獨應用「借款者特質及貸款契約條件」因素，故亦驗證外商銀行參酌「總體經濟因素」做為決定房屋貸款市場進退機制之重要要指標有其合理性與實用性。[[abstract]]This study explores three factors affecting mortgage default behavior, including “characteristics of borrowers”, “loan contract” and “macro-economic indices”. We collected 2,658 terminated mortgage loans from a large commercial bank in Taiwan, and employed Logistic regression model for testing the significance of variables, and used Akaikes’ information criterion (AIC), Bayesian information criterion (BIC) and receiver operating characteristics (ROC) curve for comparison. Results show that “macro-economic indices” can serve as an excellent predictor for default behavior. This conclusion also explains why some foreign banks in Taiwan ceased their mortgage business in the mid 1990s while the economy in Taiwan was experiencing a decline