Conditional market beta for REITs: A comparison of modeling techniques

There has accumulated strong evidence in the literature that market beta (β) is time varying. This paper contributes to the literature by studying how to best model the time varying beta for REITs. We include several commonly used methods and evaluate their performances in terms of in-sample beta estimates and out-of-sample beta forecasts. We apply these methods to U.S. equity REITs. Our results overwhelmingly suggest that the state space model is the best performer. Such a conclusion is supported by different evaluation criteria and robust to different sample splitting. Our findings have direct financial implications. The forecasted betas (preferably through the state space model) can be used in many applications such as estimating the cost of capital for the purpose of capital budgeting involving REITs, identifying equity REIT mispricing, evaluating the performance of managed REIT portfolios, etc.