房地產市場在台灣經濟發展上佔有重要地位，而土石流災害為台灣人民居住之潛在威脅，政府部門除致力於土石流災害防治外，亦公開其相關勘查作業結果之資料，故本研究將探討土石流災害對鄰近房地產交易價格造成之影響。 本研究之資料選取台灣經濟新報TEJ之不動產實價登錄交易資料，資料期間為2017年1月至2021年11月。其中房屋交易資料之選用地區，經考量土石流與房屋資料之完整性，分別選取屬於六都之台北市、新北市、桃園市，與非屬六都的基隆市、新竹縣、南投縣、宜蘭縣，共七個縣市。此外本研究選用國家災害防救科技中心所提供之土石流潛勢溪流資料，利用地理資訊系統(Quantum GIS)計算其中每筆交易資料與土石流發生地之距離，並考慮其他可能影響房價之房屋與鄰里特徵屬性變數。 本研究之樣本選取使用所有樣本與選擇距離土石流兩公里內之兩種樣本，並使用半對數特徵價格模型，含直線距離模型與分距離間距模型，兩種模型對樣本進行迴歸分析。其中直線距離模型實證結果顯示，在使用所有樣本時，大部分縣市房屋與土石流之距離對於房價會有正面影響，即房屋距離土石流災害愈遠，其房屋價格愈高；而使用選擇樣本時，七縣市房屋與土石流之距離對於房價皆有正面影響，且影響力更強。分距離間距模型之實證結果顯示，使用所有樣本或選擇樣本時，除宜蘭縣外，距離土石流一公里內之房屋價格皆相較其他區域之房價為低。兩種模型中，房屋與鄰里特徵屬性變數之實證結果，符合房屋距離土石流災害愈近，其房價愈低之預期，並具有統計顯著性。 ;This study selected the actual estate transaction data from Taiwan Economic Journal (TEJ) from January 2017 to November 2021. Seven counties were selected that offered a complete data set of landslide and housing information. These counties include municipal cities of Taipei City, New Taipei City, and Taoyuan City, and non-municipal cities including Keelung City, Hsinchu County, Nantou County, and Yilan County. In addition, the distance between each transaction and the site of the landslides was calculated by Quantum GIS, and other housing and neighborhood characteristic variables that may affect the housing price were also considered. This study uses a semi-logarithmic characteristic pricing model, including a straight-line distance model and a fractional distance model for the empirical studies. The empirical results of the linear distance model showed that the distance between houses and landslides had a positive effect on housing prices in most of the counties when all samples were used. The study shows that the greater the distance is between the house and the landslide, the higher the price of the house. This data is apparent in all seven counties used. The empirical results of the spacing models show that the house prices within one kilometer from the landslides are lower than those in other areas in all counties researched except for Yilan County. The empirical results of the house-neighborhood characteristic variables in both models are consistent with the expectation that houses are exponentially cheaper as the distance to the landslide decreases.
