Macroeconomic Uncertainty and Predictability of Real Estate Returns: The Impact of Asset Liquidity

Abstract

Recent research has shown that macroeconomic uncertainty is a significant factor that is contemporaneously incorporated into asset returns. Therefore, it should not have a role in predicting future returns. At the same time, separate research has demonstrated that illiquidity is related to future returns. We examine the interplay between these two dynamics in a commercial real estate setting, where (il)liquidity is a defining characteristic of the asset class. Empirical tests confirm the absence of return predictability for liquid assets (publicly traded property portfolios). However, we find significant return predictability predicated on ex ante macroeconomic uncertainty when we examine assets that are not as liquid (directly held property portfolios). Our findings are robust to several refinements, including adjustments for delays in the transaction closing process to establish transaction prices in the directly held market, controls for leverage inherent in publicly traded real estate asset returns, and pro-cyclical liquidity variation in private real estate markets.

Keywords:

• Commercial real estate
• macroeconomic uncertainty
• price return predictability
• liquidity
• parallel markets

Acknowledgments

We thank RCA, NAREIT, and the MIT Center for Real Estate for providing the real estate asset price level indices employed in this paper. We are grateful to Paul Brockman, Stephen Brown, David Geltner, Bill Hardin (editor), Mariya Letdin, David Ling, Sterling Yan, two anonymous reviewers, and participants at the 2022 Annual Conference of the American Real Estate Society for helpful comments and suggestions. Nayar and Price are grateful for financial support from the Hans Julius Bär and the Collins-Goodman chairs, respectively.

Notes

1 Also see Cooper and Gulen (2006), who are skeptical about papers reporting the presence of predictability using the out-of-sample estimations, which they suspect to be consistent with data snooping.

2 Priestley (2019) reports that the results of Rapach et al. (2016) are highly sensitive to inclusion of data from the recession of 2008; specifically, excluding that recessionary period negates any evidence of predictability.

3 Bali et al. (2017) characterize the macroeconomic uncertainty variable of Jurado et al. (2015) as “defined as the conditional volatility of the unforecastable component of a large number of economic indicators” (p. 472).

4 Additionally, in a recent overview of commercial real estate, Ghent et al. (2019) state that there are few to no systematic differentials between the quality of properties held by REITs and those directly held by private institutional investors.

5 For additional insight into liquidity and real estate markets, see Ametefe et al. (2016), who provide a nice overview.

6 https://www.reit.com/data-research/reit-indexes/ftse-NAREIT-us-real-estate-index-historical-values-returns

7 https://www.rcanalytics.com/our-data/rca/_cppi

8 The macroeconomic uncertainty data are available for download from Sydney Ludvigson’s website: https://www.sydneyludvigson.com/data-and-appendixes. We are grateful for the availability of the data.

9 We obtained the data from Kenneth French’s website: https://mba.tuck.dartmouth.edu/pages/faculty/ken.french/datalibrary.html

10 According to the NBER, the time frame of the recession was December 2007 to June 2009.

11 For example, see Equation 1, in Welch and Goyal (2008).

12 Essentially, Priestley (2019) cautions against the possibility that predictability results are driven by the economic shock during the Great Recession. Our analysis differs from Guidolin et al. (2023) on this dimension; they report predictability in real estate returns but do not control for the crisis period.

13 In addition to the Fama-French factors, we included the Pástor-Stambaugh aggregate market liquidity factor (see Pástor & Stambaugh, 2003, 2019) in our estimations, but this did not affect any of our conclusions on predictability. In the interest of brevity, we have not included those results in our tables.

14 There is nonstationarity with the returns series for privately held commercial property, which is eliminated with one lag (first difference). However, we fail to reject the null in an Engle-Granger test for cointegration. Thus, the returns series and macroeconomic uncertainty variables are not cointegrated. We also check our results using Newey-West standard errors to compute the t-statistics. This adjusts for heteroskedasticity and autocorrelation. In our regression results, we use both standard t-tests and Newey-West t-statistics. The results are similar using both techniques for all our tests. We thank an anonymous reviewer for suggesting this line of analysis.

15 Another concern associated with the delay in recording transaction prices is the possibility that the return series using those prices may manifest smoothing. To address this concern, in the spirit of Geltner (1991, 1993), Fisher et al. (1994), and Cho et al. (2003), we unsmooth the transaction price–based monthly return series for use in our estimations. Our inferences regarding predictability associated with macroeconomic uncertainty are robust to using these “unsmoothed” transaction price–based returns. We thank an anonymous reviewer for suggesting this robustness test.

16 Although not shown, in separate tests we excluded the NBER-defined crisis period and obtained consistent results.

17 Additionally, the coefficient on the interaction variable is positive and significant. This implies that while returns are negatively related to the macroeconomic uncertainty from two periods before, this effect is attenuated during the Great Recession period.

18 Oikarinen et al. (2011) and Pagliari et al. (2005) find that deleveraged publicly traded returns do not differ significantly from private market returns.

19 Ziman includes a monthly return series (usdretx) that excludes dividends.

20 We thank the Price Dynamics Platform at the MIT Center for Real Estate for providing the constant liquidity indices. See http://pricedynamicsplatform.mit.edu/team.html

21 Consistent with Bali et al. (2017), we find that contemporaneous macroeconomic risk is a priced factor in an asset pricing context.