在我國土地開發市場中，開發者應偏好整合坵塊、以大面積土地進行開發，惟相關報導指出臺北市近年以狹小基地開發之鳥籠建案有所增加，遂形成實際與預期不符的矛盾情況。本文從土地供給角度出發，認為除了地價昂貴之外，土地整合成本亦對開發基地規模有決定性的影響，且由於容積獎勵制度的施行，當開發者取得大小足以開發的土地後，可能不再進行土地整合、而改以資本替代之。 為驗證上述推測，本研究以民國83至102年第三種住宅區新建住宅之使用執照存根為基礎，建立開發個案之資料庫；並利用集群分析方法，對開發個案之土地、資本配置型態進行歸納。經過統計，歷年均有近40%的新建住宅屬於基地狹小之鳥籠建案，且集群分析實證結果指出在空間上，市中心至郊區的住宅開發型態呈現單獨小規模開發至整合大規模開發的遞移變化，代表土地整合的難題已長期存在於臺北市住宅開發市場中，其中又以土地較零碎的市中心更為明顯。 此外，本文另以基地面積、公告現值與政策虛擬變數對實際容積率進行線性迴歸，結果顯示整併土地確實有提升容積總樓地板面積的規模經濟效果，但效果卻不如政策性容積獎勵來得明顯；且經過模擬亦得知若欲達到同樣的樓地板面積增量，申請政策性獎勵所需的成本將遠低於進行土地整合。是以在取得合理規模之基地後，土地整合已非必要，再加上多樣化卻缺乏總量管制的容積獎勵，開發者將傾向透過政策性手段來達到擴大資本開發規模之目的，遂造成住宅開發型態呈現空間垂直的成長。 / In Taiwan, land developers are expected to prefer assembling small plots into a large lot for development. However, it is widely reported recently that development projects of building on a small plot have been increasing in Taipei. This phenomenon is inconsistent with the above expectation. From the perspective of land supply, this study argues that apart from land prices, costs of land assembly also have a decisive influence on the lot size for land developments. Also, because of the bulk reward regulations, developers might be less willing to conduct land assembly. Instead, they choose capital intensive means which presented as building high-rise housings once they obtain a site. To verify the argument, this study establishes a database of housing projects based on the residential usage licenses from 1994 to 2013, and observes allocation patterns of land and capital by using cluster analysis. Statistically, there are about 40% of housing projects that were built on plots smaller than 330 m2 every year. The empirical result of cluster analysis shows that housing development patterns change from downtown to suburbs. In downtown area, most developers build houses on a small single plot. However, they assemble plots to get a bigger site in the suburbs. It means that housing development market of Taipei has been entangled by land assembly problems for a long time. This issue explicitly exists in downtown area where lands are much more fragmented. In addition, this study uses site area, current land value and a policy dummy as independent variables and building density as dependent variable to conduct a linear regression. The result shows that, although land assembly increases the floor area, its effect is not as large as that of policy incentives. Based on this result, the comparison analysis of land assembly and policy incentives also points out that the application of bulk reward costs less than land assembly, while they can enjoy the same floor area. Therefore, it is believed that developers tend to choose increasing housing scale by applying bonuses instead of assembling lands. This trend makes the housing development forms vertically expand consequently.