Introduction

REITs as an asset class have become an increasingly important part of well diversified portfolios for both individual and institutional investors. Although REITs are very different from non-REIT equities in many ways, it has been shown that REIT and non-REIT equity returns have striking similarities. For example, both Chui et al. (2003a, b) and Derwall et al. (2009) use factor models to show that REIT returns are dominated by a momentum effect. The momentum effect in non-REIT equity returns has been described as an anomaly that evades rational explanations by asset pricing models. Such factor-model findings suggest that REIT returns and equity returns are similar with respect to the momentum effect. There exists, however, an alternative empirical framework with which to examine cross-sectional determinants, that being the “characteristics” framework of Daniel and Titman (1997, 1998).

In this study, we reexamine the influence of momentum and other characteristics on REIT returns using the Daniel and Titman framework and find clear distinctions between our evidence for REIT returns and prior studies’ findings for non-REIT equity returns.Footnote 1 One of the starkest contrasts is found when comparing characteristic sorted portfolios across different interest rate environments

Motivation and Empirical Framework

The ubiquitous Fama and French (1993) and Carhart (1997) factor models used to explain patterns in the cross-section of stock and REIT returns rely on a risk, or covariance story. Specifically, the return premium for value stocks over growth stocks is due to the observation that value stocks covary similarly among themselves and with the underlying factor that proxies for risk. Though ubiquitous, factor models are not without controversy. Among the criticisms is that they are empirically motivated and have no theoretical linkage to the underlying fundamental drivers of returns. While there is much empirical support for factor models, Lewellen et al. (2010), among others, are highly critical of many of the tests and demonstrate that the results from them can be misleading.Footnote 2

Daniel and Titman (1997, 1998) provide an alternative framework for explaining cross-sectional stock return patterns. They show that it is the underlying characteristics of the stocks (e.g., low book-to-market versus high book-to-market ratios), rather than the return covariance with a risk factor, that are responsible for cross-sectional return differences. Thus, Daniel and Titman offer a characteristics story, rather than strictly a covariance story. Like factor models, Daniel and Titman’s characteristics model also has detractors (see, for example, Davis et al. 2000, and the subsequent “rebuttal” by Daniel et al. 2001).

Derwall et al. (2009), as well as earlier work by Chui et al. (2003a, b), extend non-REIT examinations to an examination of the cross section of REIT returns. Perhaps the most notable and consistent finding in these studies is that REIT momentum emerges as the dominant determinant of the cross-section of REIT returns. The competing variables examined in their tests (and in earlier studies they build upon) include size, book-to-market, momentum, volume, turnover, and analyst coverage, as well as the use of the Fama and French (1993) and Carhart (1997) models. Interestingly, however, an extension of Daniel and Titman’s (1997, 1998) approach to this issue is missing, providing a key motivation for our study.

In this study, we follow Derwall et al. (2009), Chui et al. (2003a, b), and earlier studies on the cross-section of REIT returns, and reexamine whether and how REIT characteristics are related to REIT returns, while controlling for the importance of momentum. Our interest is not in testing an asset pricing model in the sense of looking for “alpha.” While Derwall et al. (2009) use factor models to provide important insights regarding momentum’s influence on REIT fund performance measurement and comparison, the use of factor models in this study would not have furthered our understanding of the drivers of REIT returns. Instead, we seek to explore the drivers of REIT returns in Daniel and Titman’s characteristics framework while accounting for the now well-established momentum effect. We are not aware of any studies that do this. Our extension updates Chui et al. (2003a) data by nearly a decade and introduces institutional ownership, a variable known to influence the existence of a value premium,Footnote 3 as a new variable not widely examined in REITs. We also contribute to the literature by examining the inter-temporal influences of interest rate levels, driven by Federal Reserve policy rates, on cross-sectional REIT returns and characteristics. While past studies have found the Fed’s policy stance to be related to the level of REIT returns, we are unaware of studies that examine the influence of changes in policy rates on the relations between firm characteristics and REIT returns in a cross-sectional setting.

The Characteristics

Beyond the studies mentioned above, a vast literature examines the cross-sectional patterns of expected stock and REIT returns. Among the variables found to be most strongly related empirically to (non-REIT) stock returns are size and book-to-market equity (Fama and French 1992), momentum (Jegadeesh and Titman 1993), liquidity (Brennan et al. 1998; Liu 2006; and Korajczyk and Sadka 2008), institutional ownership (Nagel 2005), analyst coverage (Hong et al. 2000; Diether et al. 2002), and monetary policy (Jensen et al. 1996; and Jensen and Mercer 2002).

Among the variables found to be most strongly related empirically to REIT returns are size and book-to-market equity (Peterson and Hsieh 1997), momentum (Chui et al. 2003b; Hung and Glascock 2008, 2010), liquidity (Clayton and MacKinnon 2000; Cannon and Cole 2011), institutional ownership (Wang et al. 1995), analyst coverage (Chui et al. 2003a; Devos et al. 2007), and monetary policy (Chen et al. 2011; Chang et al. 2011). Each of these studies provide motivation for the particular variables they investigate, but in general the REIT literature has followed the non-REIT asset pricing literature to similarly motivate the variables. Given the findings of these studies, we examine the role of each of these variables.Footnote 4

Our sample consists of all equity REITS from the CRSP database listed by PERMNOs in Feng et al. (2011) over the period 1993–2009.Footnote 5 We calculate monthly observations of each of the characteristics for every REIT in our sample, and define them as follows. Size (MV) is the month-end product of common shares outstanding and price per share from CRSP. Book-to-market (B/M) is calculated as the book value of common equity from Compustat in year t-1, where the year starts in June and ends in May, divided by the firm’s market capitalization in December of year t-1. Momentum (MOM) is the compounded 6-month stock return from month t-7 to month t-2. Illiquidity (ILL) is the standardized turnover-adjusted number of zero daily trading volumes over the prior 12 months as defined in Liu (2006).Footnote 6 Institutional ownership (IO) is measured as the percent of shares outstanding held by institutions in the previous quarter as reported in the Thomson Reuters Institutional Holdings database. Analyst Coverage (COV) is defined as a binary indicator variable taking on the value of one if the firm had one or more analyst estimates in IBES, and zero otherwise.

A key contribution of this study is our use of the characteristic-sort procedure of Daniel and Titman (1997, 1998). This procedure, unlike a factor model approach, allows us to easily examine the influence of interest rates, a second key contribution presented later in the study. At the end of every month we rank all REITs on each of the characteristics and assign each REIT to one of three portfolios. The “High” portfolio contains the 1/3 of REITs with the highest value of the particular characteristic and the “Low” portfolio contains the 1/3 of REITs with the lowest value of the particular characteristic.Footnote 7 Following Derwall et al. (2009), we then obtain the next month’s return on every REIT and calculate equally-weighted returns for the High, Medium, and Low portfolios. This procedure is repeated for every month in the sample. By focusing on one-month holding periods, we minimize the impact of momentum on the return measurementFootnote 8 (thereby allowing us to measure the effect of the other characteristics), but still allow for momentum in the construction of the portfolios.

Table 1 provides summary statistics for each of the characteristic variables. The sample contains 311 REITs from 1993 to 2009, resulting in 35,198 firm-year-month observations. Consistent with previous studies of REIT markets, the typical firm in our sample has a market capitalization of approximately $1.5 billion with an average Book-to-Market ratio of 0.855. Not surprisingly, we also observe a dramatic increase in REIT market values preceding the financial crisis and a stunning reversal in Book-to-Market values after its onset. Finally, the summary statistics in Table 1 readily illustrate the dramatic increase in liquidity, Institutional Ownership, and Analyst Coverage of REIT markets during our sample period.

Table 1 Summary statistics
Full size table

Table 2 further delineates the attributes of our sample and presents the correlation coefficients between each pair of firm characteristics. While the vast majority of observed correlations are statistically significant, only four characteristic pairs exhibit coefficient estimates greater than 0.25 in absolute value. Specifically, Market Value appears to be positively correlated with both Institutional Ownership (ρ = 0.413) and Analyst Coverage (ρ = 0.386), while Institutional Ownership is inversely related to Illiquidity (ρ =−0.383) and directly related to Analyst Coverage (ρ = 0.522).

Table 2 Correlations
Full size table

The characteristic sorted REIT returns for the High, Medium, and Low portfolios formed on each of the characteristics are provided in Table 3. We present mean monthly returns, standard deviation of monthly returns, median returns, and p-values from t-tests of whether the REIT returns in the High versus Low portfolios are different. Consistent with the existing literature, the descriptive analysis of our results focuses on differences in mean returns across portfolio characteristics. Given the relatively small portfolio groupings in some years, particularly when sorted on multiple characteristics, we acknowledge the potential for a small group of outliers to materially influence individual results. As such, throughout our analyses we also report median portfolio returns. In general, the results from the two alternative metrics are qualitatively similar.Footnote 9

Table 3 Characteristic sorted REIT returns
Full size table

The t-tests reveal significantly different mean returns on High versus Low portfolios when formed on Momentum, Institutional Ownership, and Illiquidity. As such, we implicitly control for Momentum in our subsequent analysis through a double sorting procedure. As expected, low Institutional Ownership portfolio mean returns are shown to be higher than high Institutional Ownership portfolio returns. Also as expected, mean returns from portfolios of illiquid REITs are significantly higher than mean returns from portfolios of liquid REITs. Market Value, Book-to-Market, and Analyst Coverage do not appear to be significantly related to average portfolio returns in this univariate setting. Interestingly, results obtained from examining median returns are somewhat different. Specifically, Market Value, Illiquidity, and Analyst Coverage all appear to be positively related to median returns, while Momentum and Institutional Ownership lose their previously reported significance levels.

Federal Reserve Monetary Policy and Policy Rates

Jensen and Mercer (2002) and Becher et al. (2008) provide evidence supporting a link between cross-sectional stock returns and changes in the Federal Reserve’s policy rates. Specifically, they show that return premiums associated with cross-sectional characteristics (such as book-to-market) are generally higher in periods of falling policy rates which, if the policy stance is counter-cyclical, will more typically coincide with periods of economic slowdown and periods of lower consumption. Jensen and Moorman (2010) find inter-temporal variation in the return premium to illiquid stocks that is directly related to the same Federal Reserve policy rates. Buetow and Johnson (2001) examine the level (rather than the cross section) of REIT returns and Fed policy rates and find that REIT returns are significantly higher in periods of falling policy rates versus periods of rising policy rates, while Darrat and Glascock (1989) conclude that a measure of monetary policy (Granger) causes changes in real estate returns.

In contrast, Mueller and Pauley (1995) find little relation between interest rates and REIT returns, while Chen et al. (2011) full sample results fail to identify significant linkages between EREIT returns and monetary policy changes. Interestingly, Chen et al. (2011) also demonstrate that the influence of monetary policy on EREIT returns may well be contingent on market states (bull vs. bear markets) and expected returns. Specifically, using quantile regression, they conclude monetary policy changes only significantly influence EREIT returns during bull markets when investors possess relatively low expectations for future real estate price increases.

From a traditional real estate valuation perspective, required returns and capitalization rates in commercial property markets are determined by a variety of factors including income growth expectations, capital market risk premiums, and equilibrium market interest rates.Footnote 10 Based on this evidence, we control for Fed policy rates in the analysis via the creation of characteristic-sorted portfolios over periods of falling policy rates (expansive monetary policy) versus periods of rising policy rates (restrictive monetary policy).

Following Jensen et al. (1996), we classify every month as either an expansive policy month or a restrictive policy month. The rule is: expansive policy phases of the cycle begin when a discount rate decrease follows a discount rate increase, and ends with the next discount rate increase; restrictive policy phases of the cycle begin when a discount rate increase follows a discount rate decrease, and ends with the next discount rate decrease. This approach is followed by numerous researchers; for a recent example see Buetow et al. (2009).

Table 4 provides summary statistics for the High, Medium, and Low portfolios formed on each of the characteristics in both restrictive and expansive monetary policy environments. We present mean and median monthly returns, and p-values from t-tests of whether the REIT returns in the High versus Low characteristic sorted portfolios are different, for both policy environments. With respect to mean returns, the policy environment does not appear to alter the relation of Illiquidity or Analyst Coverage sorted portfolio returns. That is, the relations look similar in both restrictive and expansive environments, with Illiquidity increasing returns and Analyst Coverage being unrelated to mean returns.

Table 4 Characteristic sorted REIT returns conditioned on monetary policy
Full size table

Regarding size, we see that large REITs significantly outperform small REITs in a restrictive environment, yet when rates are falling in an expansive environment this return difference is statistically insignificant. This result is inconsistent with the evidence for non-REIT common stocks, where premiums are larger during periods of loosening monetary policy. Footnote 11

Next, we observe that high Book-to-Market REITS (marginally) outperform low Book-to-Market REITs during periods of expansive credit, but this relationship is non-existent during restrictive periods. Given the marginal significance of this relationship, we refrain from drawing strong conclusions as to the consistency with broader finance literature findings. Turning to Momentum, we find winner portfolios outperform loser portfolios during restrictive periods, but no differences are observed during expansionary policy periods. This result is again inconsistent with the findings for non-REITs, where Jensen and Mercer (2002) and Becher et al. (2008) find return premiums are more significant during periods of monetary expansion.

Finally, the relationship between Institutional Ownership portfolio returns is also impacted by the monetary policy environment. Specifically, low Institutional Ownership portfolios provide higher portfolio returns when rates are falling in expansive environments, while no significant return differences are found during restrictive monetary policy environments. This latter result is consistent with the finding for non-REIT common stocks that higher returns are required on stocks less-held by institutional investors when the Fed is loosening.

To summarize, in periods of falling policy rates under an expansive monetary policy, when cross-sectional risk premiums are typically different (higher), we find significantly higher mean returns on REITs with 1) higher Book-to-Market ratios, 2) lower Institutional Ownership and 3) higher Illiquidity. In periods of rising rates under a restrictive monetary policy, when cross-sectional risk premiums are typically not different (lower), we find significantly higher mean returns on REITs with 1) larger Market Value, 2) higher Momentum, and 3) greater Illiquidity. Thus, like the evidence found for non-REIT common stocks, there is clear support for the hypothesis that REIT characteristics are related to returns differently depending on macroeconomic conditions, in this instance being proxied by the Fed policy-rate cycle. However, for several of these characteristics, the relation runs in the opposite direction from that found for non-REIT common stocks.

Double Sorts

Like Chui et al. (2003a), we next present the results from a double-sort procedure. They form characteristic-based portfolios (first on size, then separately on analyst coverage, turnover, and finally book-to-market), and then sub-divide these portfolios into two momentum portfolios (past winners versus past losers) to report the effect of each of the characteristics on the momentum profits. Because it is clear that momentum continues to be a dominant characteristic (for example, see Fama and French’s 2011 recent evidence), we instead reverse the order of the double-sorts and explore the influence of the characteristics after controlling for momentum. This is also consistent with our empirical approach in which we examine only the next-month return after portfolio formation. That is, we do not measure the impact of momentum, which plays out over several more months than our return horizon.

Table 5 presents mean monthly returns, median monthly returns and p-values from t-tests for return differences across the portfolios. In examining mean returns, the first entries indicate that in both loser and winner portfolios, Market Value does not appear to be related to returns. This is also the case for the previously reported univariate results before controlling for the monetary policy environment. Similarly, in both the loser and winner portfolios, the performance of low (no) Analyst Coverage REITs is indistinguishable from that of high (covered) Analyst Coverage REITs. This is again consistent with the univariate results. Thus, controlling for Momentum does not appear to alter the return relationships for Market Value and Analyst Coverage, which again is inconsistent with the evidence for non-REIT common stocks. In loser portfolios, the high Book-to-Market portfolio provides significantly higher returns than the low Book-to-Market portfolio. However, there is no significant return difference observable in winner portfolios. Similarly, in loser portfolios the low Institutional Ownership portfolio again provides significantly higher returns than the high Institutional Ownership portfolio, while there is no significant return difference in winner portfolios. Regarding liquidity, we see that the less-liquid portfolio outpaces the more-liquid portfolio in loser portfolios, but once again there is no significant return difference in winner portfolios. Thus, in the portfolios of REITs with the lowest momentum, there is a mean return premium associated with higher Book-to-Market values, less Institutional Ownership, and greater Illiquidity. However, these return premiums disappear in portfolios of REITs with the highest Momentum.

Table 5 Momentum and characteristic sorted REIT returns
Full size table

Double Sorts with Fed Policy Rates

We next present the results from a double-sort procedure incorporating monetary policy. Similar to our results presented in Table 5, we form characteristic-based portfolios (sorting first on momentum, then separately on each of the remaining characteristics). We then compare the portfolios formed in restrictive versus expansive monetary policy periods to report the effect of each of the characteristics after incorporating monetary policy.

Table 6 presents mean monthly returns and p-values from t-tests for average REIT return differences across the portfolios. With respect to Market Value, placing our findings in the context of Table 4, the significant relationship for Market Value in the restrictive monetary environment appears to be robust across Momentum portfolios. Continuing, in both the loser and winner portfolios, the performance of low Book-to-Market REITs is indistinguishable from that of high Book-to-Market REITs during restrictive environments. On the other hand, during expansionary periods and within loser portfolios, high Book-to-Market REITs marginally outperform their low Book-to-Market counterparts. This is again consistent with the results in Tables 3 and 4. Thus, controlling for Momentum does not appear to significantly alter the return relationships for Market Value and Book-to-Market. In winner portfolios in restrictive monetary policy periods, the high Institutional Ownership portfolio provides significantly higher returns than the low Institutional Ownership portfolio. During expansionary times, high Intuitional Ownership firms exhibit significantly lower returns than their low Institutional Ownership counterparts. This latter result is robust across both winner and loser portfolios, and again clearly demonstrates the potential impact of monetary policy on REIT return relationships. Regarding Illiquidity, as in Tables 4 and 5, we see that the less-liquid portfolio outperforms the more-liquid counterpart in loser portfolios across both the restrictive and expansive monetary policy environments. In winner portfolios, no significant return differences are found during either monetary policy environment. Finally, portfolio returns sorted on Analyst Coverage appear invariant to our Momentum and monetary policy environment groupings.

Table 6 Momentum and characteristic sorted REIT returns conditioned on monetary policy
Full size table

To summarize, in the portfolios of REITs with the lowest Momentum, there is a return premium associated with higher Market Values during restrictive policy environments, higher Book-to-Market and lower Institutional Ownership during expansionary environments, and higher Illiquidity regardless of a restrictive or expansive monetary policy environment. In portfolios of REITs with the highest momentum, there are mean return premiums associated with larger Market Values during periods of restrictive monetary policy, while high Institutional Ownership portfolios exhibit significantly higher returns during restrictive periods of monetary policy and significantly lower mean returns during expansionary periods.

Conclusion

In this paper, we examine the patterns of cross-sectional REIT returns based upon both the underlying characteristics of the individual firms, and the underlying Federal Reserve policy rate environment in which they operate. Specifically, using a sample of equity REITs over the period January 1993 through December 2009, we find significant return differences across portfolios of firms sorted on measures of Momentum, Institutional Ownership, and Illiquidity. Consistent with the results of the previous literature on non-REIT firms, organizations characterized by either lower levels of Institutional Ownership, or greater Illiquidity, exhibit higher portfolio returns.

More intensive analysis of these relationships, in which we condition our portfolio construction upon the (restrictive or expansive) monetary policy and policy rate environment facing the firms, reveals similar results for our Illiquidity characteristics. Interestingly, while the unconditioned relationship between firm size and portfolio returns is statistically insignificant, when conditioned upon a restrictive monetary policy environment we find larger REITs significantly outperform their smaller counterparts. Once again, this result is inconsistent with the existing evidence for non-REIT stocks, where Market Value premiums are typically larger during periods of expansive monetary policy, and may be due to the unique regulatory environment in which REIT firms operate. Similarly, both our Book-to-Market and Institutional Ownership results appear to be influenced by Fed policy rates, as our characteristic sorted portfolio returns along these dimensions exhibit significant differences only during falling rate periods.

Finally, employing a double-sort procedure, where firms are first sorted into portfolios based upon Momentum terciles and then separately on their individual characteristics, we find evidence that the observed relationships between higher returns and 1) higher Book-to-Market ratios, 2) lower levels of Institutional Ownership, and 3) greater Illiquidity are driven by our low Momentum (i.e., “loser”) portfolios. While these Illiquidity results appear consistent across alternative policy rate environments, the Book-to-Market and Institutional Ownership metrics do not. Specifically, evidence of significant mean return premiums accruing to high Book-to-Market REITs is confined to lower momentum portfolios during periods of expansive monetary policy. Turning to our Institutional Ownership results, consistent with expectations, during periods of falling policy rates firms with low Institutional Ownership outperform their high Institutional Ownership counterparts. This result holds regardless of Momentum tercile. On the other hand, during periods of rising policy rates, Institutional Ownership appears unrelated to characteristic sorted returns for portfolios of both low- and medium-Momentum REITs, while our high-Momentum portfolios actually exhibit an unexpected positive relationship between realized returns and Institutional Ownership. In sum, we view these results as strongly supportive of the notion that the determinants of characteristic-sorted REIT returns diverge markedly from those of non-REIT common stocks, and further, that such relationships are related to the Federal Reserve policy rate environment faced by these firms.