次貸危機對全球金融市場及不動產造成極大的衝擊，本文八檔國內投信所發行的國內區域型及全球型REITs為研究對象，評估傳統風險值模型(Value at Risk, VaR)與極端值模型(Extreme Value Model)的績效。實證上運用不同的波動率模型計算變異－共變數法的風險值，極端值理論則是以一般化柏拉圖極端值模型(GPD)分別計算在99%、97.5%與95%信賴水準下的風險值，並透過回溯測試及壓力測試來檢測模型績效。實證結果顯示，全球型REITs不論採歷史波動率或GARCH波動率估計風險值，均較國內區域型為佳。而應用GPD模型估計風險值，亦發現全球型的績效較國內區域型為佳。最後的壓力測試結果亦顯示出，所有的模型皆通過三倍標準差所求出的風險值。

The subprime mortgage crisis has had a great impact on global financial and real estate markets. Based on 8 domestic and global Real Estate Investment Trusts (hence REITs), we compare the performance of variance-covariance and extreme value Value-at-Risk model. In empirical study, we consider different volatility models in variance-covariance Value-at-Risk model. We also apply GDP extreme value model for estimating Value-at-Risk under the confidence level of 99%, 97.5% and 95%, respectively. Then, we check the performances of all Value-at-Risk models by back-testing and stress testing. The empirical results show that global REITs have a better performance with historical volatility according to the GARCH model than local REITs. The GPD has the same results. Finally, the stress test shows that the entire model passes the three times standard deviation Value at Risk.