台灣地區餘屋存量影響因素之分析An Analysis on the Determinants of Housing Inventories in Taiwan

民國89年底台灣地區戶口普查結果顯示，台灣地區空屋數量為123萬戶，與張金鶚等(2000)估計的數據十分接近。雖然國內空屋情況很嚴重，但國內相關研究並不多見，而針對建商手中餘屋存量研究的分析則更少。由於後者手中的餘屋直接在市場銷售，其存量對於市場的影響可能會比一般家計單位手中的空屋更為嚴重。本研究利用土地銀行的問卷資料，針對768家與土銀有往來的建商的餘屋狀況進行分析。為避免因樣本選擇不當可能產生的估計誤差，我們利用Heckman二階段估計法，先估計建商可能出現餘屋的機率，然後再進一步分析產生餘屋的原因。結果我們發現推案量、推案城市、市場狀況、餘屋型態、及預售方式等都是影響餘屋的顯著因素。此外，inverse Mill's ratio的係數也十分顯著，表示Heckman二階段估計法對於修正樣本選擇可能產生的偏誤的確有明顯的效果。
關鍵詞：餘屋、空屋、Heckman二階段估計法

According to the Population Census of Taiwan in 2000, there are 1.23 million units of vacant houses in Taiwan, which is quite near to Chang, et al (2000)'s estimation. Though the problem of excess dwelling units is quite serious, there are little studies on relevant topics in Taiwan, not mention studying specifically on housing inventories held by the housing developers. Since the developers' housing inventories have a direct impact on housing market and housing price, it is meaningful to pay more attention to study the determinants of the developers' housing inventories in Taiwan. Applying a data set from the Land Bank of Taiwan, this paper analyzes the situation of housing inventories of 768 housing developers in Taiwan. In order to avoid the problem of sample selection bias, we apply the Heckman two-stage model to estimate the probability whether a developer will have housing inventories or not. Then, we estimate the impact of the determinants on the developers' housing inventories. We find that all of total sales, location, local market situation, type of housing units, and selling method have significant impacts on housing inventories. Furthermore, we also find that the estimated coefficient of inverse Mill's ratio is also significant, which shows that we might get a serious sample selection bias if we do not use the Heckman two-stage model to correct our sample.
Key words: housing inventory, vacant house, Heckman two-stage model