本研究使用近來新發展的「已實現共變異數」 (realized covariance) ，以2004年12月至2010年12月，不動產投資信託 (REIT) 、大型股、小型股、債券、黃金之指數股票型基金 (ETF) 日資料為研究樣本，檢驗 REIT 與其它資產的動態資產配置效益。經由估計一系列的24月滾動窗口複迴歸模型，觀察 REIT 變異數的動態組成結構；並使用平均數-變異數模型，觀察各資產之最佳投資權重的變化趨勢以及替代性。實證結果顯示，在樣本期間中， REIT 與股票有強烈的關聯性，尤其是大型股。此外，實證結果也顯示 REIT 與股票的關聯性在2008年10月以後有進一步地提升。與Case et al. (2010) 的結論一致，本研究指出在2000年代， REIT 與股票無法提供良好的風險分散效果，此與 NAREIT 的官方主張，亦即 REIT 與股票有良好的風險分散效果相互衝突。 In this study, we use the recently developed realized measures of covariance to examine the dynamic asset allocation benefits between real estate investment trusts (REITs) and related assets in exchange traded funds (ETFs). We use daily ETF data of REITs, large cap stocks, small cap stocks, bonds, and gold from December 2004 to December 2010 as our samples. By estimating a series of rolling multiple regressions for 24-month window, we perform the variance decomposition of REITs; by using the mean-variance model, we observe the trends of optimal portfolio allocations of each asset and the substitutability among assets. The results show that REITs are more linked with stocks recently during the sample period, to large cap stocks especially. In line with Case et al. (2010), we find that REITs and stocks would not provide that great diversification effects in 2000s; our result conflicts with Chou et al. (2009) and the official arguments of the NAREIT society.
