Real Estate Investment Trust Returns: An Analysis of Volatility and Correlation

Abstract

This thesis examines commercial real estate investments by examining the returns on U.S. Real Estate Investment Trusts. First, efforts are made to ascertain whether the volatility of REIT returns is driven more by financial assets (i.e. stocks and bonds) or by direct real estate. Based on quarterly returns over the sample period of 1999 through 2014, results suggest that REIT return volatility is driven vastly by financial assets (especially stocks) and very little by direct real estate. Clayton and MacKinnon’s (2003) variance decomposition approach is applied to reach this conclusion. The second part of this thesis examines how diversification benefits of REITs change over time, using a sample period of September 2000 through December 2014. Engle’s (2002) model of Dynamic Conditional Correlation with Generalized Autoregressive Conditional Heteroskedasticity (DCC-GARCH) is used to examine correlation dynamics between REITs and common stocks and bonds. In addition, via time-varying correlations estimated by DCC-GARCH, Sharpe ratios in tangent portfolios and optimal asset weights in minimum-variance portfolios are examined to observe the diversification benefits of REITs when added to benchmark portfolios comprised of domestic and international stocks and bonds. Under a similar methodology to Huang and Zhong (2013), the results suggest that diversification benefits of REITs change significantly over time. Finally, out-of-sample performance of portfolio strategies is examined by comparing correlation structures estimated by DCC-GARCH, rolling-window correlation, historical correlation, and constant correlation. This thesis corroborates the finding of Huang and Zhong (2013) that DCC-GARCH provides the best correlation estimate to use in asset allocation when studying REITs. The second part of this thesis’ analysis is replicated on other test assets besides REITs, namely International Real Estate Investment Trusts (IREITs), commodities, and Treasury Inflation-Protected Securities (TIPS). Out-of-sample performance analysis suggests that DCC-GARCH provides the best correlation estimate to use in asset allocation when studying some, but not all, test assets.