REIT Returns: Insights from Behavioural Asset Pricing

While extensive literature focuses on explaining REIT returns from the perspective of fundamentals and risk factors, little attention has been paid to the impact of investor behaviour on REIT pricing. Specifically, non-rational investor behaviour causes mispricing in the REIT market, and such inefficiencies will be corrected over time. This paper, built on Daniel et al. (2020), aims to introduce two behavioural asset pricing factors. We encapsulate the behaviours that affect REIT prices in long- and short-term behavioural factors. Our long horizon factor (FIN) is based on issuance/repurchase activities. It is a catch-all for many possible sources of overconfident investors’ misperception. On the other hand, the post-earnings announcement drift (PEAD) in the REIT market is considered as the short-term behavioural factor, reflecting high-frequency systematic pricing movement caused by limited investor attention to earning related information. The research questions we aim to address are threefold: first, do the two behavioural asset pricing factors, FIN and PEAD, contribute to explaining REIT returns? Does encompassing both into a conventional linear asset pricing model enhance the model’s performance? Second, which factor or factors among multiple asset pricing factors exhibit stronger explanatory power for REIT returns? Third, in non-linear machine learning models, such as the random forest model, do the performance of the asset pricing factors and of the model consistent with the results obtained from conventional linear models? This paper utilises CRSP Ziman US REIT monthly data from January 2003 to December 2021. We constructed and tested various real estate specific asset pricing factors including the two aforementioned behavioural factors, the real estate factor (RE), the liquidity factor, and the Fama-French (2015) five-factor customised by REIT data only, against the conventional stock market asset pricing models. The tests used cover the R-squared contributions, the jointalpha test, the Gibbons, Ross, and Shanken (GRS) test, and the spinning test. Our results show that: In both the parametric asset pricing model as well as the non-parametric random forest model, the inclusion of two behavioural factors, FIN and PEAD, positively affected the performance of the model. FIN and PEAD complement the information that the other factors do not possess and improve the model in better explaining REIT returns. The market, liquidity, and size factors based on the REIT data show strong explanatory power for REIT returns, while FIN and PEAD show weaker but still significant explanatory power for REIT returns. From an empirical point of view, the findings from the data confirm that REIT investors are not entirely rational. Overconfident (captured by FIN) and inattention (captured by PEAD) behaviours exist in the REIT market and have an impact on REIT returns.