租買選擇和家庭組成都是影響住宅需求的主要考慮。若一個家庭決定選擇購屋，則是否需要貸款，貸款多少，是另外一個必須同時決定的考量。雖然國內已有相關文獻分別探討租買選擇、貸款需求與世代組成對住宅需求的影響，但受限於資料，這些文獻一直都缺乏對上述這三個因素交互影響的聯合分析。本研究主要目的，是採用1996年台灣的家計收支調查資料，利用幾個不同層次的logit模型，同時分析這三個因素對住宅需求的交互影響關係。 本文先採用條件式logit模型與多項式logit模型來進行估計，然後再利用Hausman檢定來測試這二種模型的適當性。由於兩種模型的適用性都未通過檢驗，我們接著再建立一個巢式logit模型，並利用充分訊息最大概似法(FIML)進行估計，結果發現FIML的設定方式符合巢式logit模型的要求。在FIML的估計之下，我們發現家戶總所得、家戶組成、與所得組成對於租買選擇及貸款選擇都有顯著的影響。 Tenure choice and household composition are important factors in determining housing demand. If a household decides to own a dwelling unit, consequently, he has to consider whether and how much to borrow from a bank. Though there are some literatures in Taiwan considering the influence of tenure choice, household composition, and mortgage on housing demand, it is lack of a systematic analysis on the three factors. The purpose of this paper is using the 1996 Survey on Family Income and Expenditure of DGBAS and applying several levels of logit model to estimate the impacts of the above three factors on housing demand simultaneously. A conditional logit model and a multinomial logit model are estimated first and the Hausman tests are applied to check the effectiveness of the two logit models in describing the data. Since the IIA hypothese does not hold completely under the two types of specifications, we then extend to a nested logit model and apply the full information maximum likelihood (FIML) method to consider these three factors on housing demand. The results show that the model's requirement of the nested logit specification is fulfilled under the FIML estimation. Moreover, it is found that household income, composition of generation, and income earners composition all play significant roles on the behavior of tenure choice and housing loan demand.