Information Uncertainty and the Post-Earnings-Announcement Drift Anomaly: Insights from REITs

Also referred to in this paper as simply “the drift”.

First documented by Jegadeesh and Titman (1993), momentum is the phenomenon where stocks that have performed well (poorly) in recent months tend to continue to perform well (poorly).

The literature is replete with unsuccessful attempts to explain the drift. Researchers have not been able to attribute PEAD to various risk adjustments, poor research design, market frictions such as transactions costs and liquidity constraints, or the psychological biases of market participants.

Mahoney et al. (1996) show that while there are some differences in terms of geographic concentration and proportionate allocation to various property types (retail, multi-family, office/industrial, etc.) in the asset bases of publicly (NAREIT) and privately (NCREIF) held commercial real estate, there is nontrivial overlap as well. The overall location correlation is 0.79 and both NAREIT and NCREIF have roughly 85% of their holdings in the top 100 MSAs. For example, 1995 REIT allocations include 35% to retail, 24% to apartments, 12% to industrial/office, and 28% to ‘other’. 1995 NCREIF allocations are 36% in retail, 14% in apartments, 44% in industrial/office, and 6% to ‘other’.

See Bernard (1992) for a detailed review of possible explanations for the PEAD anomaly.

Pagliari et al. (2005) and Riddiough et al. (2005) both show that returns between the two parallel markets are not significantly different from one another after adjusting for characteristics such as leverage, liquidity, management structure, property type, and taxes. MacKinnon and Al Zaman (2009) provide additional evidence of linkages between the publicly and privately held commercial real estate markets.

The private market can only be considered to be “much larger” than the public market if property not tracked by NCREIF is taken into consideration in addition to NCREIF property. Otherwise, the two markets are close in size. NCREIF tracks just over 6,000 properties with an estimated combined market value of $238 billion as of the end of 2009. According to NAREIT, publicly traded REIT market capitalization as of the end of 2009 is roughly $248 billion.

http://​www.​ncreif.​com/​about/​

With respect to the price discovery processes, the direction of information flow is the subject of debate in the literature. Some contend that information flows from the private to the public real estate sector (Tuluca et al. 2000). Others argue that information flows from the public market to the private market (Barkham and Geltner 1995; Chiang 2009).

Internal Revenue Code §857(a)(1).

Equity REITs comprise the largest proportion of total publicly traded REITs and this proportion has increased over time. According to data available on the NAREIT website, Equity REITs made up 46%, 74%, and 81% of the total number of REITs in the 1982–1989, 1990–1999, and 2000–2008 time periods, respectively. Consequently, Mortgage (35%, 17%, 15%) and Hybrid (20%, 10%, 4%) REITs decreased as a percentage of total REITs during the same time periods. Although untabulated, the relative proportions based on market capitalization are even more pronounced in favor of Equity REITs.

Some of the PEAD literature, such as Bernard and Thomas (1989, 1990), restricts their samples to NYSE and AMEX firms. However, recent literature that investigates market anomalies in a REIT setting includes Nasdaq firms as well (Chui et al. 2003; Hung and Glascock 2008, 2010). In unreported analyses we find that our results are not sensitive to the inclusion of Nasdaq firms.

Except REITs, most of which are classified as SIC 6798.

This essentially represents the population of REITs and NonREITs during the 1982–2008 sample period for which sufficient data are available, consistent with prior drift studies (Foster et al. 1984; Bernard and Thomas 1989, 1990). However, in unreported analysis we also check our results using a size matched sample, a book-to-market equity matched sample, a volatility matched sample, and a random sample and find consistent results.

Both Chui et al. (2003) and Zhang (2006) use returns volatility as a measure of uncertainty. Greater volatility indicates greater uncertainty.

Although not shown, all volatility measurements are statistically different from one another at the 1% level.

We choose the stock price as of day t = −45 as the SURP scalar following Liang (2003). However, we find that our results are not sensitive to the choice of SURP scalar.

For comparability between REITs and NonREITs we follow the literature and use EPS as the earnings measure in the surprise calculation. The question remains regarding the most appropriate earnings surprise calculation in a REIT setting given the widespread use of Funds From Operation (FFO) as an industry specific earnings measure. However, we leave a complete exploration of this issue to further study.

Bernard and Thomas (1989) find that PEAD is not a result of risk mismeasurement and, thus, is unaffected by risk adjustment.

See Fig. 5 in Bernard and Thomas (1989)

There are large differences in the amount of total drift (pre- and post-earnings announcement) for REITs, NonREITs, and Small NonREITs as shown in Figs. 1, 2, and 3. Upon closer visual inspection of Figs. 1 and 2, it appears that much of this is attributable to differences in decile spreading during the anticipation window. While the primary focus of this study is on stock returns behavior during the post-announcement window, with a secondary focus on stock returns during the initial reaction window, we acknowledge the related nature of the anticipation window returns as well. As such, later analysis includes tabulated results for the total drift period [CAR(−60,60)], although we do not attempt to render a detailed interpretation. We conclude that REIT returns are less able to anticipate future earnings surprises than NonREITs (Bernard and Thomas 1989; Campbell et al. 2009). However, there is little literature to build upon in deciphering the exact meaning of the anticipation window differences and we leave a more in depth look at this component to future study.

We also observe that PEAD, as plotted in Fig. 4, seems to be counter-cyclical, particularly for REITs relative to NonREITs. Lending support to the idea that the drift is related to uncertainty, REIT PEAD increases dramatically in times of general economic uncertainty such as the recessions in the early 1990s, early 2000s, and late 2000s.

We also use alternative drift windows of 30 [CAR(1,30)] and 45 [CAR(1,45)] days and find consistent results (unreported). However, the focus of our analysis is on the sixty day [CAR(1,60)] post-earnings-announcement drift window, consistent with the literature (Foster et al. 1984; Bernard and Thomas 1989; Liang 2003; Campbell et al. 2009). Additionally, following the suggestion of a referee, we also winsorize cumulative abnormal returns to check for potential outlier effects and our inferences remain unchanged (unreported).

We find the regression results to be robust to alternative regression methods which are much less powerful. We run Fama and MacBeth (1973) cross sectional regressions where the first stage consists of 109 quarterly cross sectional regressions and the second stage effectively evaluates the results of the cross sectional regressions using the time series standard errors. The resulting coefficient estimates are highly significant with signs of the same direction and magnitudes that are nearly identical to those in Table 2.

At the suggestion of a referee we check for outlier effects by winsorizing the cumulative abnormal returns and find results that are not materially different (unreported).

In unreported results we also regress CAR(1,60) and CAR(−1,0) on an operating partnership indicator variable in a manner similar to Eqs. 6, 7, and 8, only we limit the sample to all REIT observations and drop the variables that differentiate between REIT and NonREIT observations. The regression results are consistent with those in Table 2, Panels A and B, and suggest that the greater potential uncertainty associated with the complex operating partnership ownership structure is related to greater PEAD.

We acknowledge that the increased proportion of Equity REITs may be driving this result. As a percentage of total REIT market capitalization, Equity REITs made up 46% in 1982–1989, 74% in 1990–1999, and 81% in 2000–2008. However, most of the increase can be seen when comparing the first two sub-periods. Thus, if the increase in Equity REITs is driving the result, it would seem likely that we would also find significant differences between the 1982–1989 and 1990–1999 sub-periods as well. However, when the 1990s are included by arbitrarily cutting the sample in the middle of the decade, the differences become more difficult to detect. In this scenario (not reported), the earlier and latter sub-periods are not significantly different from one another.

Neither NonREITs nor Small NonREITs show this same upward trajectory across the three sub-periods. We find all NonREIT and Small NonREIT differences to be statistically insignificant.

We thank a reviewer for suggesting that potential profitability may be tempered by short side implementation issues. Over the full sample period, roughly half of the REIT long-short zero investment portfolio returns can be attributed to the short position.