因應新巴塞爾資本協定倡導的內部模型法，個別銀行想要計提住宅貸款業務的資本需求額，需自行估算違約機率、違約損失率與違約暴險額。故本研究希冀找出影響住宅貸款戶異常還款、還款狀態轉移的因素藉以協助銀行估計違約機率，並且嘗試發掘估計違約損失率之顯著影響因子，幫助放款部門能夠依照違約風險高低給予不同的放款利率加減碼，找出住宅貸款業務的市場區隔，建立更為客觀的授信政策。 根據本文的實證結果顯示，影響住宅貸款異常還款及還款狀態轉移的因素尤以「目前貸款態勢」的利率差異與貸款往來年限最具解釋能力，利率差異越大、貸款往來年限越久的客戶，比較容易延長遲付時間，並惡化其逾付情形，故銀行在擬定催收策略時，可加重該變數的影響性，增加催收效率。此外，比較實際還款狀態機率值與模型預測值時，發現樣本銀行即使經歷總體環境不好的年度，仍能控制使其逾期現象不惡化反而下降，顯見銀行有做好住宅貸款業務的風險管理。最後在驗證住宅貸款違約損失率的影響變數，發現多與借款者選擇違約的行為及景氣良寙相關，而與個別貸款合約內容，如:貸款成數、期限、金額及抵押品座落地點較無關聯。 As the New Basel Capital Accord supports the internal ratings based approach, individual banks need to estimate probability of default（PD）, loss given default（LGD） and exposure at default by themselves in order to set capital requirements of residential mortgages. Because of the reason, the purpose of this paper are to find the significant factors on abnormal payments and loan states transitions to help estimate PD,and factors of estimating LGD to assit in pricing differentials to compensate for risks,and making strategic credit-granting decisions. According to the empirical findings, interest differential and vintage of loan are more useful in explaining resolution of sample delinquencies than other variables . They are positively and significantly related to the probability of delinquencies and defaults and transitions.Thus financial institutions can add the importance of them when making collection policies.Besides, by comparison of actual loan state probabilities and model forecasts reaveals that sample bank can control the delinquent conditions even under bad economics.Finanily, the behaviors of borrowors choosing defaults and prosperity are siginificant factors of estimating LGD, whereas information on the loan contract are not.
