本文從不動產抵押貸款銀行的角度，以涉險值(Value at Risk)的觀念，衡量台灣四個縣市(台北市(縣)、台中市與高雄市)不動產市場的下檔風險；並藉由比較各縣市的涉險值大小，深入了解各區域不動產市場的風險。實證結果顯示，四縣市中除台中市外，其餘縣市的房價報酬率分配則可能為常態；其次，本文使用的三種模型在不同的信賴水準下，表現有些微差異，並 不存在單一模型能完全預測四縣市的房價風險；再者，各縣市的房價下跌風險高低依序為高雄市、台中市、台北市、台北縣，此一結果與一般認為台北市風險最高的認知，顯然大異其趣；最後，在最高的信賴水準下(99%)，極值理論法能正確的預測四個縣市極端的房價跌幅，因此不失為一種較能反映房價風險的預測方法。

The current study employs Value at Risk (VaR) to evaluate the downside risk of the real estate market in four areas (Taipei City, Taipei County, Taichung City and Kaohsiung City) in Taiwan, and compares the VaR estimates among them with in-depth measurements of real estate price risk in these metropolitan areas. The main empirical results show that, first, the distributions of house price returns in the four areas fall into two categories: non-normal fat-tailed distributions for Taichung City and normal distributions for the other three cities. Second, there is no universally appropriate VaR model that captures real estate risk in the four areas. Third, the risk levels of the four areas in order of size are Kaohsiung City, Taichung City, Taipei City, and Taipei County. Finally, under the highest level of confidence (99%), the model based on extreme value theory responds quickly to the changes in house price returns and provides correct VaR forecasts in these four areas compared to the other models.