REIT idiosyncratic risk

Abstract

Investors are told to hold a well‐diversified portfolio; when everyone does so, idiosyncratic risk is diversified away and does not enter the pricing equation in equilibrium. This study finds that the idiosyncratic risk of real estate investment trusts (REITs) appears to have an upward time trend during the vintage REIT era (1980–1992) and appears to trend downward during the new REIT era (1993–2006). This study also finds that this pattern appears to coincide with a reversion in the relation between REIT idiosyncratic risk and the excess returns of REITs. Specifically, during the vintage REIT era, the excess return of REITs is positively related to REIT idiosyncratic risk. After 1993, the excess return of REITs is negatively related to REIT idiosyncratic risk.

Keywords:

• REITs
• idiosyncratic risk
• predictability
• risk‐return tradeoff
• time trend

Acknowledgments

The authors thank Seow Eng Ong (the Editor) and two anonymous referees for their helpful comments and suggestions. The authors also thank Kenneth French for providing factor returns.

Notes

1. Rather remotely related to this study is the work of Chaudhry et al. (2004), who examine the determinants of REIT idiosyncratic risk.

2. It is well documented that REITs experienced a structural change in the early 1990s (Glascock, Lu, and So, 2000). One of the main goals of the Revenue Reconciliation Act of 1993 is to facilitate investment in the real estate industry by institutional investors.

3. The concepts of new and vintage REIT eras are discussed in Downs and Patterson (2005).

4. Callender, Devaney, Sheahan, and Key (2007), Ellis, Wilson, and Zurbruegg (2007), and Lee and Devaney (2007) provide recent literature reviews on the risk diversification of real estate portfolios.

5. Jirasakuldech et al. (2009) argue that a positive relationship between REIT market volatility and REIT market return is consistent with the theoretical foundation of the capital asset pricing model (CAPM) and arbitrage pricing theory. The authors’ statement is true only if REITs are completely segmented from stock markets, whereas the existing literature appears to suggest that REITs are integrated with stocks (Cauchie and Hoesli, 2006; Clayton et al., 2007; Ling and Naranjo, 1999). Under the paradigm of financial integration, the intertemporal capital asset pricing model of Merton (1973) predicts an intertemporal trade off between the returns and the variance of the market portfolio. Because the volatility of the REIT market, relative to the volatility of the market portfolio, is at best sector‐specific, the standard theory of asset pricing does not immediately render a testable implication for the relation between REIT sector returns and the variance of REIT sector returns.

6. The one‐period‐ahead regression in Equation (2) is intuitive and widely used in the literature. One can view the current explanatory variables as proxies/estimates for the expected values of the variables; thus, one‐period‐ahead tests address the theorized contemporaneous relations between REIT excess returns and risk measures.

7. Goyal and Santa‐Clara (2003) demonstrate that the effect of idiosyncratic risk constitutes about 85% of the stock counterpart of . Thus, we view and as proxies for REIT idiosyncratic risk.

8. Campbell et al. (2001) demonstrate that the REIT idiosyncratic risk measures defined in Equations (5) and (6) provide good approximations for idiosyncratic risk.

9. Campbell and Yogo (2006), Lewellen (2004), Stambaugh (1999), and Torous, Valkanov, and Yan (2005), among many others, demonstrate that a highly persistent predictive variable may lead to inference difficulties, such as inefficient estimates of regressions.

10. For example, if an idiosyncratic risk measure follows an AR(1) process, then the innovations can be obtained by running the following regression: .

11. This study also experiments with the use of daily value‐weight returns on the REIT portfolio and the market portfolio. The unreported results are qualitatively similar.

12. This study also experiments with the use of a natural logarithm to transform the explanatory variables. The unreported results are qualitatively similar.

13. We also experiment with the original data; i.e., without the transformation. Our conclusions are not sensitive to data treatment.

14. Chiang, Lee, and Wisen (2005) discuss this method in detail. The authors use the method to investigate the time trend in REIT betas.

15. Ooi et al. (2009) present a graphical description of this decreasing pattern in REIT‐specific volatility over the period 1990–2005. This study provides formal tests to substantiate the authors’ observation.

16. The Newey–West (1987) t‐statistics are used throughout the paper and are in parentheses. The actual implementation of the regression specification in Equation (7) is based on the squared roots of predictors. This study follows Jiang and Lee (2006) and chooses six lags based on the Akaike (1974) Information Criterion and the Schwarz (1978) Bayesian Criterion. The study also experiments with the use of four, five, seven, and eight lags for robustness check. The unreported results based on alternative lag lengths are qualitatively similar.

17. We also experiment with adding the momentum factor of Carhart (1997). The unreported results are virtually the same.