![Sample our Built Environment journals, sign in here to start your access, Latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5926/TOC-Subject-Sample-2021-Built-Environment.jpg"})
Abstract
This study examines the liquidity risk of real estate investment trusts (REITs) as measured by their return sensitivity to marketwide liquidity shocks. Due to their unique dividend payout rules and associated high cash payouts, REITs should benefit investors by reducing their reliance on the stock market to satisfy liquidity needs. Using a sample of 440 equity REITs from 1980 through 2015, we find empirical evidence consistent with this paradigm along four key dimensions. First, unlike non-REIT real estate firms, REITs exhibit a negative sensitivity to marketwide liquidity shocks. More specifically, when marketwide liquidity declines, REIT prices tend to increase relative to the broader stock market. Second, our findings are not property type specific, but rather are evident across broad classifications of property type sectors. Third, consistent with the importance of cash flow stability, smaller REITs provide protection against liquidity risk only when their dividend frequency is relatively high. Finally, examining only those firms changing their REIT status within the sample period, we find marketwide liquidity risk is lower when these firms operate as REITs than when they operate as non-REITs. Taken together, these findings provide support for the notion that investors view dividend payouts as a source of enhanced liquidity, and further, that REITs, as a security class with relatively high regulatory mandated payout requirements, provide investors with an important benefit in the form of reduced liquidity risk.
Keywords:
Acknowledgments
For helpful discussions and comments, we would like to thank Honghui Chen, Bill Hardin (the Editor), Charles Schnitzlein, Ajai Singh, Geoffrey Turnbull, Qinghai Wang, and seminar participants at the University of Central Florida, University of Southern Maine, and the American Real Estate Society 2017 annual meetings. The insightful comments of three anonymous referees have greatly improved the paper. We remain responsible for any errors.
Notes
1 The REIT Modernization Act of 1999 reduced the mandatory minimum dividend distribution requirement to retain tax exempt status for federal income tax purposes from 95% of ordinary taxable income to its current level of 90%. See 26 U.S.C §857 (2015) for the U.S. Internal Revenue code pertaining to REITs.
2 See, for example, Wang et al. (Citation1993).
3 The limitations on earnings retention should effectively reduce a REIT’s ability to accumulate free cash flows, thereby placing limits on managerial options and their ability to direct shareholder resources toward personal utility maximizing initiatives. Likewise, the lack of profit retention capabilities forces many REITs to become frequent participants within the capital markets, where transparency is rewarded with easier access to capital and/or more attractive lending terms.
4 Consistent with this framework, Pástor and Stambaugh (Citation2003) provide evidence that the common component in liquidity is a state variable which is both material and value relevant with respect to asset prices.
5 By “high cash payouts” we mean that ex ante, investors expect most of their total return to come from cash dividends rather than capital gains (see also Case et al., Citation2012; Hardin et al., Citation2002; Lee & Kau, 1987).
6 Alternative measures of a stock’s financial market liquidity employ bid-ask spread metrics. While we do not dispute the importance of such metrics in measuring information asymmetry, the focus of our analysis is on the relative value of underlying equity positions in response to marketwide innovations in liquidity as opposed to potential changes in the transactions costs of trading associated with such events. See for example Ametefe et al. (2016), Bertin et al. (2005), Blau et al. (2015), Cannon and Cole (2011), Clayton and MacKinnon (2000), and Danielsen and Harrison (2000, 2007) for further discussions of market microstructure based metrics of REIT financial market liquidity.
7 A disambiguation here is necessary, as other studies (e.g., Chordia et al., 2001) use this term to denote the variability of a security’s stock market liquidity over time.
8 Within the context of multi-factor models, returns are measured in relation to the benchmark returns captured by additional factors. In other words, a security with a negative liquidity beta may in fact produce a positive return when marketwide liquidity improves, yet its return is negative relative to the benchmark. For example, in untabulated tests we find that the univariate relation between returns and marketwide liquidity shocks is positive both for REITs and for non-REITs; REITs exhibit a sensitivity of around 0.057 while non-REITs exhibit a significantly higher sensitivity of 0.228.
9 Implicit in our argument is the well-established (both on theoretical and empirical grounds) positive relation between stock market liquidity and asset prices. Amihud et al. (2005) provide a review of this extensive literature.
10 In contrast to Banerjee et al. (Citation2007), who completely exclude REITs from their estimation sample, the current investigation focuses extensively on these firms and exploits the legal environment of REITs to mitigate this endogeneity concern. Since REITs are required by law to distribute 90% of their taxable income as dividends (95% prior to 2000), the decision to pay dividends is driven in large part by exogenous factors, such as firm profits, cash flows, or both.
11 Income and expense accruals, as well as capital raising activities, may also support consistently high payout thresholds. Moreover, Hardin et al. (2009) show that the cash holdings of REITs are orders of magnitude lower than those of the average public firm.
12 Consistent with this notion, additional empirical evidence confirms increased dividend payouts are observed when taxation policy favors dividends. For example, Chetty and Saez (2005, 2006) find an increase in dividends after the Jobs and Growth Tax Relief Reconciliation Act of 2003 that reduced the maximum tax rate on dividends from 38% to 15%. Conversely, Maris and Elayan (1991) fail to find evidence of tax induced clientele effects for REITs surrounding the Tax Reform Act of 1986.
13 See, for example, Amihud et al. (2005), Lang and Maffett (2011), and the references therein.
14 The general finance literature consistently demonstrates firm size is highly correlated with a firm’s level of financial constraint. See for example Kaplan and Zingales (2000).
15 Appendix A describes the main variables used in this study.
18 ARIMA models are often used to decompose marketwide variables into predictable and unpredictable components (e.g., French et al., 1987, among others). The unpredictable component is then interpreted as an innovation (or a shock) to the variable of interest. We use the smallest canonical correlation method to select the AR and MA degrees. Our findings are similar using different degrees for the AR or MA processes. Following Brennan et al. (2013), we also compute both dollar volume- and turnover-based versions of the Amihud measure. The presented findings are based on the turnover version, primarily because the turnover version adjusts for firm size and for the aggregate size of the market. Nonetheless, we have verified that the dollar volume measure leads to similar estimates and conclusions.
19 On the surface, including firms with negative earnings that pay dividends leads to negative observed payout ratios. While obviously problematic from a conceptual perspective, dividend payout ratios do not enter directly into our estimation of the factor models, and thus do not pose an econometric concern. Rather, we contend the omission of firms with negative current period earnings would pose a greater threat to the generalizability of our findings.
20 Unlike SIC codes, GICS codes explicitly define firms operating in the real estate sector of the economy, permitting a distinction between REITs and non-REIT real estate firms.
21 Examining the annual reports of several firms operating within GICS codes 40401020, 40403010, 40403020, 40403030, or 40403040, there are cases where firms discuss plans to become REITs. There are also cases where firms explicitly consider REIT status yet choose not to operate as REITs due to the stricter rules REITs need to follow.
22 We obtain this data through www.snl.com. These 48 property companies are not a perfect subset of the GICS-based 138 non-REIT real estate firms, as somewhat surprisingly only 17 firms enter both comparison groups.
23 All results are available upon request from the authors.
24 See for example Boudry et al. (2010), Feng et al. (2007), or Harrison et al. (2011).
25 The exceptionally high aggregate dividend of non-REIT stocks in 2006 is due to the cash acquisition of Realogy Corporation by Apollo Management, L.P.
26 While Danielsen and Harrison (2000) find differences in liquidity between REITs that trade on organized exchanges compared to those that trade over-the-counter, their use of a microstructure liquidity measure differs greatly from our operationalization of liquidity.
27 To account for possible correlations over time, standard errors are clustered at the firm level (Petersen, 2009).
28 The firm-level regressions in Table 4 test for differences between REITs and non-REITs using seemingly unrelated estimation of the model for the two samples. To confirm the robustness of our findings with respect to a different specification, Table B.2, Panel A of Appendix B reports estimates from a model pooling REITs with non-REITs and using interaction effects. The estimates from this alternative approach are nearly identical to the ones reported in Table 4, Panel A.
29 To allow for serial correlation in the time series of portfolio returns, we use Newey-West estimation with twelve lags (Newey & West, 1987). Using one, three, six, or 24 lags does not materially change our findings.
30 Using equally-weighted portfolio estimates, the liquidity beta of REITs is 0.116 lower than the liquidity beta of non-REIT real estate firms. Thus, a one standard deviation decline in marketwide liquidity leads to an annualized increase in relative REIT price of approximately 8.23%.
31 Note that the tables report the adjusted R2, whereas we use the unadjusted R2 to assess the explanatory power of marketwide liquidity.
32 For the equally-weighted portfolio, liquidity incrementally explains 3.56% (or around one third) of a total R2 of 10.61%. The F-statistic across this alternative specification equals 16.97, which is again significant at the 0.01 level. As a separate point, the portfolio regressions estimated separately for the REIT and non-REIT samples produce R2 values that are generally high, typically ranging from 60% to 70%. High R2 values are to be expected in these portfolio regressions. For example, Fama and French (Citation1993) and Chen et al. (Citation2012) report an R2 of nearly 70% (or even higher) for portfolios based on non-REIT common stocks and on REIT common stocks, respectively.
33 To further elucidate the question of how the underlying real estate assets are related to marketwide liquidity, in separate tests (untabulated) we examine how real estate property prices correlate with marketwide liquidity. We expect the concern of omitted risk factors to be especially problematic if we also find real estate assets to be negatively related to marketwide liquidity. Fortunately, however, we do not find such a result. In fact, the value of real estate assets is positively but insignificantly related to marketwide liquidity shocks.
34 In Table B.2, Panel C of Appendix B, we extend the eight-factor model even further by including additional control variables such as each firm’s market capitalization, return volatility, ROA, leverage, market-to-book ratio, and share turnover. Once again, our findings are not materially affected by the addition of these control variables. We thank an anonymous referee for suggesting these tests.
35 We examine the individual unclassified REITs and cannot identify a consistent pattern to allow for their classification into one of the other groups. Of note, many represent specialized facilities which are neither well diversified, nor fit cleanly into the existing CRSP/Ziman property type categories.
36 Using the CRSP/Ziman constructed indices by property type leads to similar conclusions. The number of monthly observations varies across property focus due to data availability. Nevertheless, portfolio return series for most property types cover all months from 1980 through 2015.
37 We note two exceptions to this general pattern. Specifically, the `ι>`ι > βLIQ`/ι>`/ι> estimates for REITs investing in either lodging/resorts and/or residential properties are higher than those found for diversified REITs. As such, we urge caution when drawing definitive conclusions regarding the relative magnitude of liquidity betas across these sectors.
38 In addition to altering mandatory dividend requirements, the Act dramatically broadened the array of services REITs could directly provide to their tenants. We make no a priori assumptions as to the anticipated effects of these additional changes on REIT liquidity betas.
39 While the REIT Improvement Act of 2003 was initially introduced in Congress by James “Jim” McCrery III (R–Louisiana) on April 30th, 2003, it was not formally signed into law until October 2004 by President George W. Bush.
40 A chi-squared test shows that the decline is not significantly different from zero (p-value of 0.483).
41 Readers may reasonably wonder whether the hypothesized relations are altered by the rise of institutional investors which accompanied the onset of the modern REIT era in the early to mid-1990s. We explore the relevance of institutional investor ownership for liquidity risk. In addition, untabulated analyses conducted exclusively on both sample observations occurring post-1993, as well as on organizations incorporated subsequent to 1993, produce qualitatively similar results to those found throughout the paper. These results should not be particularly surprising, as while institutional investors have clearly changed the operating landscape of REIT markets, the dividend requirements which we argue are driving the mitigation of liquidity risk within this market sector are fundamentally independent of the presence of institutional investors in the marketplace. For additional insight on REIT market changes during the modern REIT era, see Ott et al. (Citation2005).
42 Here we do not further divide the firms into high dividend payers and low dividend payers, mainly because of the relatively lower dividend frequency for the non-REIT sample and the relatively lower number of observations.
43 Based on the much larger sample of non-REIT common stocks, we also perform additional analyses suggested by a referee. Specifically, we compare the liquidity risk of the 100 largest firms in the S&P 500 with the liquidity risk of non-S&P 500 firms. After controlling for the other risk factors, we find the 100 largest firms in the S&P 500 exhibit a slightly lower liquidity risk than non-S&P 500 firms. However, the difference in liquidity betas is very small at -0.02, and is statistically indistinguishable from 0 in most specifications.
44 All estimates are available from the authors upon request.
45 In additional tests, we re-evaluate our findings while focusing only on REITs with positive base rent revenue (S&P Global Market Intelligence item 132516) and no development revenue (item 132522) or property development pipeline (item 131588). Investors in REITs with no rental revenue or an active property development pipeline may rationally expect lower dividend payments, and hence, we need to ensure our findings are not sensitive to the inclusion of such REITs in our main sample. Indeed, our findings remain similar after excluding such REITs. These estimates are reported in Table B.6 of Appendix B.
46 Institutional ownership data are obtained from Thomson Reuters’ 13f filings.
47 In another test, we examine whether REIT liquidity risk varies conditional on the existence of a dividend reinvestment plan (DRIP). Bond et al. (2018) find that REITs with DRIPs tend to pay a high dividend and maintain a stable payout policy. Moreover, such programs may provide enhanced flexibility to investors with irregular transactions needs, effectively providing them access to periodic cash flows when necessary, while simultaneously decreasing transactions costs of re-investment when access to such cash flows is not required. The estimates show liquidity risk is generally similar between REITs with and without DRIPs.
48 54 firms were excluded due to this restriction. The inclusion of these firms leads to qualitatively similar results.