Information Uncertainty and Behavioral e

Journal of International Money and Finance 31 (2012) 1930–1952
Contents lists available at SciVerse ScienceDirect
Journal of International Money

and Finance
journal homepage: www.elsevier.com/locate/jimf
Information, uncertainty, and behavioral effects: Evidence

from abnormal returns around real estate investment trust
earnings announcements
Frank Gyamfi-Yeboah a, David C. Ling b, Andy Naranjo b, *
a
Department of Land Economy, Kwame Nkrumah University of Science and Technology, Kumasi, Ghana
b
Department of Finance, Insurance, and Real Estate, Warrington College of Business Administration, University of Florida, P.O. Box
117168, Gainesville, FL 32611, USA
a b s t r a c t
Keywords:
In this study, we examine the influence of real estate market
Earnings announcements
sentiment, market-level uncertainty, and REIT-level uncertainty on
Real estate
REITs
cumulative abnormal earnings announcement returns over the
International 1995–2009 time period. We first document the relative coverage of
Behavioral effects analysts’ earnings forecasts on U.S. REITs, as well as REITs from
Uncertainty several countries (i.e., Australia, Belgium, Canada, France, Hong
Information processing Kong, Japan, the Netherlands, and UK). We show that coverage
outside of the U.S. is limited, and we consequently focus our
analysis on U.S. REITs. We find strong evidence that earnings
announcements contain pricing relevant information, with posi-
tive (negative) earnings surprises relative to analysts’ forecasts
resulting in significantly positive (negative) abnormal returns
around the announcement date. Consistent with the findings from
the broader equity market literature, we find limited evidence of
a pre-announcement drift in the cumulative abnormal returns.
However, in sharp contrast to the existing equity literature, we find
no evidence of a post-earnings announcement drift in our aggre-
gate sample or when the sample is restricted to the largest nega-
tive surprises. We find evidence of a post-earnings announcement
drift for only the largest positive earnings surprises. These results
are consistent with REIT returns more quickly impounding infor-
mation relative to the broader equity market, in part because of the
parallel private real estate market and because of the U.S. REIT
structure and information environment. Finally, in our conditional
* Corresponding author.
E-mail addresses: frank.gyamfi-yeboah@warrington.ufl.edu (F. Gyamfi-Yeboah), ling@ufl.edu (D.C. Ling), andy.naranjo@
warrington.ufl.edu (A. Naranjo).
0261-5606/$ – see front matter Ó 2012 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jimonfin.2012.05.013
F. Gyamfi-Yeboah et al. / Journal of International Money and Finance 31 (2012) 1930–1952 1931
regression analysis of cumulative abnormal returns, we find that

real estate investor sentiment, market-wide uncertainty, and firm-
level uncertainty significantly affect the magnitude of abnormal
announcement returns and also influence the effect of unexpected
earnings on abnormal returns.
Ó 2012 Elsevier Ltd. All rights reserved.
1. Introduction
Accounting information plays a central role in financial markets by allowing capital providers to
evaluate the return potential of investment opportunities and to monitor the use of their capital
once committed (Beyer et al., 2010). It is therefore not surprising that financial market participants,
researchers, and policy makers often pay close attention to accounting information, with earnings
announcements receiving particular attention. Bernard (1992) argues that “no firm-specific
performance measure is more widely reported, followed, and analyzed than accounting earn-
ings,” and there is an extensive literature documenting the existence of unusual and significant
equity pricing activity around firm earnings announcements. In fact, after examining numerous
market anomalies, Fama (1998) points directly to the post-earnings announcement drift return
anomaly (PEAD) as one of two anomalies above suspicion in posing a challenge to the efficient
markets hypothesis.1 He states that it has survived numerous robustness checks, including exten-
sions to recent U.S. data and experiments using international data from equity markets around the
world.
In this paper, we extend the broader equity literature on earnings announcement effects by using
public real estate markets as a testing ground for the existence and sources of abnormal returns around
earnings announcement. The U.S. Real Estate Investment Trust (REIT) market provides a unique
experimental setting given the existence of a parallel private real estate market and a REIT structure
and information environment that mitigates some information uncertainty. Unlike industrial firms,
REITs are required to distribute at least 90% of their taxable earnings as dividends to maintain their
pass-through tax status.2 Consequently, REITs must return to the capital markets more frequently to
finance growth opportunities since they cannot retain substantial levels of earnings (e.g., Brown and
Riddiough, 2003 and Ott et al., 2005). These regulations and increased capital market interaction
provide additional incentives for REITs to remain transparent, including providing less noisy earnings
signals than industrial firms (Danielsen et al., 2009). At the same time, public REIT markets operate
alongside a parallel private commercial real estate market whose activity provides investors with an
additional layer of investment performance information.3 Overall, these characteristics suggest that
REIT earnings announcement effects, particularly the PEAD effect, should be more muted than the
effects documented in the broader equity literature.
Research assessing the information content of earnings is extensive and dates back to the pio-
neering work of Ball and Brown (1968) and Beaver (1968). Bernard (1992), Brown (1997), and
Richardson et al. (2010) provide a detailed survey of this literature. A key finding of this research, using
both U.S. and international data over many time periods, is that cumulative abnormal returns continue
to drift up (down) for firms reporting earnings greater (less) than expected. Because these findings
strike at the core of the efficient markets hypothesis, researchers have continued to investigate the
PEAD anomaly. Although no explanation has proven to be sufficient, the literature contains several that
can be categorized as either rational or behavioral/market inefficiency based. These explanations
include joint test problems associated with misspecified risk factors or the asset pricing models used to
measure abnormal returns, investor underreaction to earnings signals during the announcement
1
Jegadeesh and Titman’s (1993) momentum effect is the second anomaly Fama notes.
2
Prior to 2001, the distribution percentage was 95%.
3
Numerous organizations disseminate private real estate market transaction and performance information (e.g., NCREIF).
1932 F. Gyamfi-Yeboah et al. / Journal of International Money and Finance 31 (2012) 1930–1952
period with a subsequent correction from new information (overreaction), delayed information pro-
cessing biases, as well as other potential behavioral biases.
In our analysis, we first document the relative reporting of FFO (Funds from Operations) and EPS
(Earnings per Share) analysts’ forecasts on U.S. REITs, as well as REITs from several other countries (i.e.,
Australia, Belgium, Canada, France, Hong Kong, Japan, the Netherlands, and UK). The descriptive
characteristics we report show that REIT analyst coverage outside of the U.S. is currently very limited.
Consequently, we focus our analysis on U.S. REITs. To measure the significance of FFO and EPS
announcements on REIT returns, we estimate cumulative abnormal returns (CARs) around the cor-
responding announcements for various pre-, near-, and post-event windows. We then examine the
influence of the announcement surprise, real estate market sentiment, market-level uncertainty, and
REIT-level uncertainty on cumulative abnormal announcement returns over our 2000–2009 sample
period, while also controlling for fundamental factors. Our additional tests on the role of investor
sentiment and market-wide uncertainty provide a further unique contribution to the broader equity
literature on factors that influence earnings announcement effects.
The paper most similar to ours is Price et al. (2010) who also examine post-earnings announcement
drift in a REIT context. However, our experimental design, tests, and results differ significantly from
theirs, as do the additional questions that we address in our paper. In particular, we use REIT analysts’
earnings forecasts in a manner similar to the broader equity earnings announcement literature and also
use REIT-specific FFO announcements. In sharp contrast, Price et al. (2010) estimate earnings surprises
using a seasonal random walk model in which the surprise is measured as the difference between
earnings-per-share and earnings-per-share lagged four quarters, scaled by the stock price 45 days prior
to the earnings announcement.4 Moreover, Price et al. (2010) do not examine the influence of real
estate market sentiment, firm-level uncertainty, market-wide uncertainty, or a broader set of funda-
mental factors on the abnormal returns. They find that the initial response to earnings surprises is more
muted for REITs relative to ordinary common stocks, but that the magnitude of drift is significantly
larger for REITs. In contrast, we find significant earnings surprise effects and significantly muted post-
earnings announcement drifts. We also document that real estate investor sentiment, market-wide
uncertainty and firm-level uncertainty significantly affect the magnitude of abnormal announce-
ment returns; they also influence the effect of unexpected earnings on abnormal returns.
Focusing on the U.S. REIT market, we find that analyst forecast coverage is significantly greater for
FFO relative to EPS, consistent with the notion that FFO dominates EPS as the main valuation metric for
U.S. REITs.5 We also find strong evidence that U.S. REIT FFO announcements contain pricing relevant
information, with highly significant two- and three-day cumulative abnormal returns around FFO
announcements. Moreover, negative earnings surprises relative to analysts’ forecasts are associated
with CARs that are negative and significant at the one percent level; positive surprises are associated
with highly significant positive CARs around the announcement date.
In contrast to these near-announcement U.S. REIT CAR results, we find very limited evidence of
a pre-announcement CAR drift, which is consistent with the broader equity literature. However, in
sharp contrast to the results typically found in the broader equity literature, we find no evidence of
a post-earnings announcement drift (PEAD) in the aggregate sample or when the sample is restricted
to the largest negative surprises. Interestingly, the post-announcement CAR(1,60) and CAR(2,60)
associated with the largest positive surprises are positive and significant at the five percent level or
greater. That is, we find evidence of a post-earnings announcement drift for only the largest positive
FFO surprises. These results are consistent with REIT returns more quickly impounding information
4
We also only include equity REITs in our analysis, whereas Price et al. (2010) include equity, mortgage, and hybrid REITs in
their analysis. Our sample period also focuses on the modern REIT era from 1995 to 2009 that consists of a larger number of
equity REITs, whereas their 1982–2008 sample period includes the pre-modern REIT era with a significantly smaller number of
REITs.
5
We find that REIT analyst coverage is very limited outside of the U.S., but the coverage is increasing. The European Public
Real Estate Association (EPRA) documents 35 countries worldwide as having REIT or “REIT-like” legislation in place, but many of
these countries have only recently enabled the use of REIT-like structures. Eichholtz et al. (2011) show that increased global
transparency has leveled the playing field for foreign real estate investors, while Eichholtz et al. (2001) provide evidence on the
benefits of including public real estate securities in property investors’ portfolios.
relative to the broader equity markets examined in prior studies, in part because of the parallel private
real estate market and because of the U.S. REIT structure and information environment.
What explains our documented abnormal returns? To address this question, we examine the
surprise component of the announcement (SURPRISE), various fundamental variables, real estate
market sentiment, and both market-wide and firm-level information uncertainty. Controlling for time
fixed effects and potential clustering of standard errors by firm, we find that firm-level uncertainty,
market-wide uncertainty, and real estate investor sentiment influence the magnitude of abnormal
announcement returns as well as the effect of unexpected earnings on abnormal returns.
In particular, in regressions using three-day CARs [i.e., CAR(1,1)] as our dependent variable, we
show that the estimated coefficient on SURPRISE is significantly positive – consistent with the existing
earnings announcement literature. We also find that the level of real estate market sentiment and
market-wide uncertainty have a negative and significant effect on the abnormal returns. That is,
periods of high real estate investor sentiment and high forward looking market volatility predict lower
abnormal returns surrounding FFO announcements, all else equal. These effects are also economically
significant with a respective investor sentiment and market-wide uncertainty estimated impact of
11.7% and 12.7% on the abnormal returns from a one standard deviation change in these variables.
We also find that the level of firm uncertainty, as measured by each firm’s idiosyncratic volatility, has
a significantly negative impact on the abnormal returns; this impact is also economically significant
(8.84% from a one standard deviation change in idiosyncratic volatility).
To examine the extent to which real estate market sentiment, market-wide uncertainty, and firm-
level uncertainty affect the relation between the announcement surprises and abnormal returns, we
interact SURPRISE with each of these variables. We find that the estimated coefficient on the interaction
between real estate sentiment and SURPRISE is positive and significant (estimated economic impact of
9.48%), suggesting that the effect of an FFO surprise on abnormal returns is greater in periods of high
sentiment. Currently, no empirical evidence exists on whether investor sentiment amplifies or
dampens the effects of an earnings surprise on abnormal returns. Our finding is consistent with
a number of theoretical predictions, including the noise trader theory developed by De Long et al.
(1990). That is, in periods of high investor sentiment, asset prices are more likely to be determined
at the margin by optimistic noise traders among whom there is less dispersion in expectations about
expected future cash flows. In such a market, the price impact associated with earnings surprises is
magnified.
We also find that the estimated coefficient on firm-level uncertainty interacted with SURPRISE is
negative and significant (estimated economic impact of 11.6%). This result suggests that greater firm
uncertainty among analysts about a firm’s end-of-quarter FFO reduces the effect of an FFO surprise on
abnormal returns. This result is consistent with the findings in Imhoff and Lobo (1992) and Francis et al.
(2007). Furthermore, although we show that market-wide uncertainty is associated with lower
abnormal returns, we find that it does not alter the effect of FFO surprises on abnormal returns.
The remainder of the article proceeds as follows. The next section provides a brief discussion of the
background literature as a context for our research. Section 3 describes our methodology, while in
Section 4 we discuss our data. Section 5 provides our main empirical results, beginning with
descriptive statistics and then focusing on our conditional regression analysis. Our conclusions are
presented in the final section.
2. Background literature: research context
2.1. The case of REITs: Funds From Operations (FFO) versus EPS
The National Association of Real Estate Investment Trusts (NAREIT) introduced FFO in 1991 as
a supplementary performance measure for REITs. NAREIT and others argue that GAAP expense items,
especially tax depreciation, distort the reported performance of equity REITs. FFO therefore adds back
real estate depreciation and several other non-cash expenses. NAREIT has revised the definition of FFO
several times since 1991. The first major revision in 1995 called for the exclusion of non-recurring
items, such as gains from property sales. However, REIT managers still had discretion over which non-
recurring items to include or exclude. NAREIT subsequently established a Best Financial Practices
Council to provide recommendations on how best to improve the calculation of FFO. Effective January
2000, NAREIT adopted the recommendations of the Council to include both recurring and non-
recurring items in FFO.
In the spirit of the returns-earning research in the accounting and finance literature, Gore and Stott
(1998) examine the information content of FFO over the 1991–1996 and find that FFO is more closely
associated with stock returns than GAAP net income (NI), but FFO excludes value relevant investment
sales information. Graham and Knight (2000) also find that FFO is more informative than NI in pre-
dicting REIT stock returns over their 1989–1995 sample period. Using a sample period from 1994 to
1996, Vincent (1999) documents that both FFO and EPS consistently provide incremental information
content. More recently, Downs and Guner (2006) find that FFO forecasts are of a higher quality than EPS
and net income over their 1998–2001 sample period. Devos et al. (2007) find less analyst forecast bias
post-2000 and that analyst coverage increases REIT value. Baik et al. (2008) provide further evidence
suggesting that the information content of FFO has significantly increased since 2000 and that
investors perceive less manipulation and greater reliability. Gyamfi-Yeboah et al. (2010) also find that
FFO provides more useful information to investors than GAAP net income.
2.2. Uncertainty, sentiment, and announcement returns
The theoretical literature and empirical evidence provide mixed predictions and results on the
relation between firm-level uncertainty and PEAD. Theoretical research on the role of uncertainty and
asset pricing dates back several decades. For instance, Robichek and Myers (1966) and Epstein and
Turnbull (1980) provide a simple theoretical framework predicting that early resolution of uncer-
tainty should have a positive effect on stock prices. In contrast, Liang (2003) combines two strands of
the behavioral theoretical literature and hypothesizes that post-earnings announcement returns
should be positively related to divergences in analysts beliefs (i.e., positively associated with private
information) and inversely related to changes in uncertainty (i.e., positively related to the reliability of
information). These predictions are derived from information processing biases where overconfidence
in private information along with self-attribution biases (Daniel et al., 1998) results in a positive
relation between divergences in analysts’ beliefs and PEAD. The inverse relation between uncertainty
and PEAD arises from investors’ underreaction to reliable information (Griffin and Tversky, 1992).
Using this behavioral framework (deviations from Bayesian behavior), Liang (2003) provides empirical
evidence that 60-day PEADs are positively related to the heterogeneity of investor beliefs and nega-
tively related to changes in information uncertainty.
Zhang (2006) also develops a behavioral foundation for testing the relation between information
uncertainty and stock returns, positing that psychological biases are magnified when there is more
uncertainty (Hirshleifer, 2001; Daniel et al., 1998, 2001). Zhang (2006) argues that if the slow market
response to information is due to psychological biases such as overconfidence, the price response will
be slower because the biases will be larger when there is more information uncertainty. Similar to
Liang (2003), this framework suggests a positive relation between information uncertainty and drift.
Zhang (2006) examines price momentum and post-analyst forecast revision price drift (not PEAD) and
finds that greater information uncertainty leads to higher future returns following good news and
lower future returns following bad news, suggesting that uncertainty delays the flow of information
into stock prices.
In subsequent research using a rational Bayesian framework (Brav and Heaton, 2002), Francis et al.
(2007) argue that investors place less weight on signals characterized by greater information uncer-
tainty. As this uncertainty is resolved, investors increase their weights on the original information
signal, which results in subsequent movements in asset prices and a decrease in abnormal returns from
such price movements as uncertainty is resolved. Francis et al’s. (2007) theoretical framework suggests
that there should be a lower initial return response associated with firm-level uncertainty and longer-
run positive return relation with firm-level uncertainty.6 Consistent with their theoretical predictions,
6
Imhoff and Lobo (1992) also examine the effect of uncertainty on reaction to earnings surprises and find estimated earnings
response coefficients to be lower for firms with greater information uncertainty.
Francis et al. (2007) find initial lower response coefficients relating to unexpected earnings
announcement returns when the announcements are made by firms with higher information uncer-
tainty. Using a six-month drift period, they also find PEAD to be more pronounced for firms with higher
information uncertainty, in contrast to Liang (2003), although Liang examines 60-day PEADs.
While the literature to date has focused on firm-level uncertainty measures and provides conflicting
conclusions about the effect of this information uncertainty on drift, the potential additional effects of
broader market-level uncertainty have not been directly examined. Micro- and macro-level informa-
tion uncertainties are related, but they can have distinct effects and understanding their relative roles is
important. For example, a recent Wall Street Journal article (“Macro Forces in Market Confound Stock
Pickers,” 9/24/2010) states, “stock pickers say the market’s macro focus has meant that company
earnings no longer drive stock prices as they once did.” When macro-uncertainty dominates, firm-level
signals get lost or muted. This idea also ties in with Akerlof’s market for lemons where with greater
uncertainty, individual signals cannot be readily disentangled from the broader market signals. To
better understand the nature and differential effects of firm- versus market-level uncertainty, we also
examine the impact of broader market uncertainty on CARs. Given the conflicting theoretical predic-
tions on the effects of firm-level information uncertainty, the potential differential effects of market-
level uncertainty on CARs are an empirical question.
The issue of whether investor sentiment plays a role in the pricing of assets has long been debated
in the finance literature.7 Classical finance theorists argue that security prices should, in equilibrium,
depend on the expected cash flows and the actions of irrational investors are offset by rational arbi-
trageurs. However, De Long et al. (1990) and Shleifer and Vishny (1997) argue that the risk created by
the unpredictability of noise traders’ and the ineffectiveness of arbitrage in extreme circumstances
prevent arbitrageurs from driving prices back to their intrinsic values. Moreover, delays in information
transmission among investors may also cause prices to move away from intrinsic values. Daniel et al.
(1998) suggest that overreaction to private information and underreaction to public information by
informed investors tend to produce short-term continuation of investment returns, but long-term
reversals as public information is eventually incorporated into asset prices. Koedijk and Stork (1994)
also provide evidence suggesting that stock market levels are influenced by investor psychology.
Neal and Wheatley (1998) examine the ability of several popular measures of investor sentiment to
explain stock returns over short- and long-term horizons. The authors find that the closed-end fund
discount is related to expected returns. Brown and Cliff (2004, 2005) also find that investor sentiment
is highly correlated with contemporaneous stock returns and asset values. Yu and Yuan (2010) examine
the extent to which investor sentiment affects the well documented positive relation between the
market’s expected return and conditional variance. The authors find that the relation is positive in
periods of low investor sentiment but no such relation exists in periods of high investor sentiment.
Baker and Wurgler (2006) construct a sentiment index based on the common variation in six
sentiment proxies and find that investor sentiment has significant price effects. Hribar and McInnis
(2012) provide evidence suggesting that the relation between sentiment and the cross-section of
returns is partly explained by biased errors in expected future earnings. Although their findings do not
rule out rational explanations, they reinforce the view that sentiment may cause investors to misprice
assets. Seybert and Wang (2009) and Safdar (2009) also provide evidence that analysts’ forecasts and
recommendations are influenced by investor sentiment. Finally, Bergman and Roychowdhury (2008)
find that management disclosures during periods of high and low sentiment reveal efforts to benefit
from sentiment-induced biases.
In summary, there is a growing body of empirical evidence that investor sentiment plays a role in
asset pricing. However, studies of how market reactions to corporate announcements are affected by
investor sentiment are limited. A recent exception is Livant and Petrovits (2009) who find abnormal
returns associated with extreme good news to be higher during periods of low investor sentiment. The
authors, however, do not examine whether the level of investor sentiment affects how investors
respond to earnings surprises.
7
Brounen and Ben-Hamo (2009) also find the presence of pricing anomalies in international real estate markets.
3. Empirical methodology
We use standard event study methodology to derive abnormal returns around FFO and EPS
announcements. Our reported results use a single-factor market model to derive abnormal returns,
though our results are robust to the use of multi-factor models that include the Fama–French model
and extensions with momentum and liquidity. We calculate daily abnormal returns around event dates
using daily return data from the Center for Research in Security Prices (CRSP) and a market model
estimation period for beta of 250 days, ending 60 days prior to the event.
In our analysis, we accumulate daily market model abnormal returns (CARs) around the earnings
announcement dates for various event windows to capture pre-, near- and post-announcement
abnormal returns. To eliminate the effects of extreme outliers, we delete FFO (EPS) surprises that
are outside of the 1st and 99th percentile, although our results are robust to their inclusion. The pre-
event windows CAR(60,2) and CAR(60,1) are, respectively, cumulative abnormal returns over
the 59 and 58 days prior to the earnings announcement. CAR(1,1) is the three-day cumulative
abnormal return centered around the announcement date, while CAR(1,0) and CAR(0,1) are two-day
cumulative abnormal returns. Finally, the post-event windows CAR(1,60) and CAR(2,60) are the
cumulative returns over the 59 and 58 days, respectively, following the earnings announcement.
To examine the conditional effects of earnings surprises, firm and market uncertainty, and investor
sentiment on CARs, we estimate several versions of the following OLS regression, controlling for time
fixed effects and potential clustering of standard errors by firm:
CARðs1 ; s2 Þ ¼ b0 þ b1 SURPRISE þ b2 Firm UNCERTAINTY þ b3 SURPRISE Firm UNCERTAINTY

þ b4 Market UNCERTAINTY þ b5 SURPRISE Market UNCERTAINTY þ b6 RESENT
þ b7 SURPRISE RESENT þ b8 Z þ ε
(1)
SURPRISE is the difference between actual FFO (EPS) and forecasted FFO (EPS) announced in quarter
t, scaled by the REIT’s share price at the end-of-quarter t 1. Firm_UNCERTAINTY is a proxy for firm-
level information uncertainty, Market_UNCERTAINTY is a proxy for market-wide information uncer-
tainty, RESENT is a proxy for real estate market sentiment, and Z is a vector of fundamental control
variables including firm size, the firm’s book-to-market value ratio, and the Fama–French factors
augmented with momentum and liquidity factors.
4. Data
4.1. FFO and EPS data sources and definitions
We use the Institutional Brokers’ Estimate System (IBES) database to identify equity REITs with FFO
and EPS announcements during the period 1995Q1–2009Q3. The IBES database provides information
on analysts’ forecasts and actual FFO and EPS for REITs listed in the U.S. and, to a much lesser extent, for
REITs listed in several other countries (i.e., Australia, Belgium, Canada, France, Hong Kong, Japan, the
Netherlands, and UK). We use the mean of the analysts’ FFO (EPS) forecasts closest to the end of the
quarter t 1 as our proxy for expected FFO (EPS) in quarter t.
4.2. Firm-level and market-wide uncertainty proxies
As proxies for firm-level information uncertainty, we use the number of revisions and dispersion of
analysts’ forecasts in the quarter prior to FFO (EPS) announcements. REVISION is the number of times
analysts covering the firm revise FFO (EPS) estimates in a quarter scaled by the number of analysts.
DISPERSION is the standard deviation of analysts’ forecasts prior to the earnings announcement scaled
by the mean absolute forecast. As a robustness check, we also use idiosyncratic volatility, IVOL, as
a proxy for firm-level uncertainty. Following Ang et al. (2006), we define IVOL as the standard deviation
of the residuals that result from regressing daily firm-level excess returns on the Fama–French factors
using a one-month window. We use IVOL in the month preceding earnings announcements in our CAR
regression models.
We obtain our market-wide uncertainty proxy, the daily CBOE Volatility Index (VIX), from the
Chicago Board of Options Exchange. VIX estimates market uncertainty by averaging the weighted prices
of two sets (first and second contract months) of near-term S&P 500 (SPX) call and put options to
bracket a 30-day calendar period. The level of the VIX on the day of the FFO (EPS) announcement is used
in our CAR regression models.
4.3. Real estate investor sentiment
Following Baker and Wurgler (2006, 2007), we use principal component analysis to construct
a quarterly real estate sentiment index based on the common variation in seven underlying proxies of
investor sentiment in commercial real estate markets: (i) the average REIT stock price premium to net
asset value (NAV), (ii) the percentage of properties sold each quarter from the National Council of Real
Estate Investment Fiduciaries (NCREIF) Property Index (NPI), (iii) the number of REIT IPOs, (iv) the
average first-day returns on REIT IPOs, (v) the share of net REIT equity issues relative to total net REIT
equity and debt issues, (vi) net commercial mortgage flows as a percentage of GDP, and (vii) net capital
flows to dedicated REIT mutual funds.8
Utilizing quarterly data from 1995 to 2009, we generate a composite real estate sentiment index,
RESENT, based on the first principal component of the contemporaneous levels of each of the seven
sentiment proxies.9 RESENT is standardized to have a mean of zero and unit variance. Fig. 1 provides
a plot of our real estate investor sentiment measure and Baker and Wurgler’s (2006, 2007) stock market
sentiment measure over time. Although there are periods during which the two sentiment measures
follow a similar pattern, there are significant divergences in sentiment across the two asset markets.
4.4. Control variables
To control for other potential sources of variation in cumulative abnormal returns, we include year
dummies, firm size, the firm’s book-to-market value ratio, and the Fama–French factors augmented
with momentum and liquidity in our regression specifications. Firm-level equity market capitalizations
and book value of equity data are obtained from Compustat. We define firm size, SIZE, as the natural log
of market capitalization and BM as the ratio of the firm’s book value of equity to its market value of
equity. Both variables are measured as of the end of the quarter prior to the earnings announcement.
We use the Fama–French, momentum, and Pastor and Stambaugh (2003) liquidity factors on the day
prior to the earnings announcement in our abnormal return regressions.
5. Results
5.1. Descriptive statistics
5.1.1. FFO and EPS announcements

Table 1 reports descriptive statistics on the coverage of quarterly analysts’ earnings forecasts for
REITs in the IBES database across countries over the 2000–2009 sample period. Australia, Belgium,
8
The average quarterly U.S. REIT price premium to NAV is obtained from Green Street Advisors, a prominent buy-side REIT
advisory firm (see www.greenstreetadvisors.com). We use the percentage of properties sold from the NCREIF NPI each quarter
as a proxy for aggregate liquidity in the private commercial real estate market. The NPI is a benchmark index of returns on
a large portfolio of institutional quality properties (see www.ncreif.org). The number of REIT IPOs and average first-day IPO
returns are constructed using data provided by the National Association of Real Estate Investment Trusts (NAREIT). The share of
REIT equity issues relative to total REIT equity and debt offerings is obtained from the Federal Reserve Flow of Funds Accounts
(see Federal Reserve of the U.S. Flow of Funds Accounts: www.federalreserve.gov). Quarterly commercial mortgage flows are
also obtained from the Federal Reserve Flow of Funds Accounts and then scaled to be a percentage of GDP. Finally, the quarterly
flow of investment capital into, and out of, dedicated REIT mutual funds is obtained from AMG Data Services.
9
We detrend the commercial mortgage flow series using the prior 2-year rolling average before inclusion in the principal
component analysis (e.g., Ling et al., 2012).
Fig. 1. Real estate and general stock market investor sentiment. This figure plots commercial real estate market and stock market
investor sentiment for the period 1992Q2–2008Q4. Sentiment indices are generated through the use of principal component
analysis. Similar to Baker and Wurgler (2006, 2007), our indirect measure of real estate sentiment, RESENT (i.e., INDRES), is the first
principal component extracted from seven underlying proxies of investor sentiment in commercial real estate markets: (i) the REIT
price premium to net asset value, (ii) the percentage of properties sold from the NCREIF Property Index, (iii) the number of REIT IPOs,
(iv) the average first-day returns on REIT IPOs, (v) the share of net REIT equity issues in total net REIT equity and debt issues, (vi)
commercial mortgage flows as a percentage of GDP, and (vii) capital flows to dedicated REIT mutual funds. Our indirect measure of
stock market sentiment, INDSMS, is the first principal component extracted from six underlying proxies of investor sentiment in the
stock market: (i) the average difference between the net asset values of closed-end fund stock shares and their market prices, (ii)
share turnover on the NYSE, (iii) the number of IPOs, (iv) the average first-day returns on IPOs, (v) the share of equity issues in total
equity and debt issues, and (vi) the dividend premium. Each index is standardized to have a mean of zero and unit variance for the
period over which it was generated.
Canada, France, Hong Kong, Japan, the Netherlands, UK and the U.S. are represented in the sample,
showing the spread of REIT structures across countries and analysts coverage of those REITs. However,
only the U.S. has a sufficient sample of REITs and associated forecasts to do meaningful, inferential
analyses with the available data. Given the current REIT analysts coverage limitations outside of the
Table 1
Descriptive statistics on REIT analyst earnings forecast coverage across countries from IBES. This table reports descriptive
statistics on REIT analyst quarterly earnings forecast coverage across countries from IBES over the 2000–2009 sample period. The
descriptive statistics include the available earnings forecast measure (EPS and FFO), number of available REITs, number of REITs
covered by analysts reported in IBES, average number of analysts, total available observations with at least one analyst over the
sample, total available observations with at least two analysts over the sample, and the forecast accuracy of the analyst’s
forecasts. Forecast accuracy is defined as the absolute value of actual EPS minus. mean analysts’ forecast all divided by actual EPS.
Country Available earnings Number of Number of Average number Total # of Total # of Forecast
forecast measures REITs REITs in IBES of analysts observations observations accuracy
with at least with at least
one analyst two analysts
Australia EPS 49 0 – – – –
Belgium EPS 17 2 1.38 16 6 0.066
Canada EPS 29 21 1.22 94 18 0.606
FFO 29 23 4.16 420 363 0.148
France EPS 46 1 1.00 1 0 0.065
Hong Kong EPS 9 0 – – – –
Japan EPS 34 0 – – – –
Netherlands EPS 5 4 1.12 33 5 0.158
UK EPS 45 1 1.70 10 7 0.102
U.S EPS 238 155 2.26 2545 1423 0.492
FFO 238 170 7.48 4033 3759 0.107
Table 2
Descriptive statistics on U.S. REIT FFO and EPS announcements. This table reports descriptive statistics on U.S. FFO and EPS equity
REIT announcements from IBES over the 1995–2009 sample period. The descriptive statistics include the number of REITs,
average and median market capitalization, and the number of observations for different levels of filtering on the number of
analysts and FFO versus EPS announcements.
Number of Average market Median market Observations

REITs capitalization capitalization
(in billion US $) (in billion US $)
Panel A: sample selection – FFO
All Equity REITs 320 1.29 0.58 10,261
Equity REITs with at least one analyst forecast 186 1.65 0.85 5647
Equity REITs with at least two analysts 178 1.74 0.93 5273
Panel B: sample selection – EPS
All Equity REITs 320 1.29 0.58 10,261
Panel C: firms announcing both FFO and EPS in the same quarter
Panel D: firms announcing only FFO in a given quarter
Equity REITs with at least two analysts forecast 175 1.22 0.72 3938
Panel E: firms announcing only EPS in a given quarter
Equity REITs with at least two analysts forecast 46 2.78 1.28 181
U.S., our analysis focuses on U.S. REITs – though future related research analyzing other markets would
be interesting as the analyst coverage of REITs increases across countries.
Focusing on U.S. REITs, Table 2 reports further descriptive statistics on our broader U.S. sample of
REIT FFO and EPS announcements. Over the 1995Q1–2009Q3 sample period, 320 unique equity REITs
are identified in the CRSP data. These REITs have a respective mean and median equity market capi-
talization of $1.29 billion and $0.58 billion, and are associated with 10,261 firm-quarters. If FFO
announcements not associated with at least one analyst forecast are eliminated (Panel A), the sample is
reduced to 186 firms and 5647 firm-quarters. The sample is further reduced to 178 firms (5273 firm-
quarters) if we require FFO announcements to be associated with at least two analyst forecasts. As
expected, REITs followed by an analyst(s) have larger market capitalizations, on average, than those not
covered by analysts.
If we eliminate EPS announcements not associated with at least one analyst forecast (Panel B), our
EPS sample is reduced from 320 to 178 firms and 2751 firm-quarters. Requiring EPS announcements to
be paired with at least two analyst forecasts reduces the number of REITs to 138 (1516 firm-quarters),
which is a more significant reduction than the imposition of the two analyst threshold on our sample of
FFO announcements. We again observe that the average market capitalizations of REITs with an analyst
following are significantly larger than REITs without an analyst following.
The number of REITs that provide both FFO and EPS announcements in the same month and that
have at least one analyst providing a forecast of each is 162 as shown in Panel C of Table 2. Requiring at
least two analyst forecasts of both FFO and EPS further reduces the sample to 121 REITs. Finally, Panel D
reveals there are 176 REITs in our sample (3149 firm-quarters) that have only FFO forecasts with at least
one analyst. The corresponding sample of REITs that have only EPS forecast is 86 with just 253 firm-
quarters as shown in Panel E.
Overall, the descriptive statistics reported in Table 2 reveal that FFO is more widely estimated by
REIT analysts, which is consistent with the results reported for Canada in Table 1. Our results are also
consistent with Downs and Guner (2006) who report similar analyst coverage patterns and find that
the quality of FFO forecasts exceeds that of EPS forecasts over their 1998–2001 sample period, which is
consistent with the results reported in Table 1 where the forecast accuracy is better in both the U.S. and
Canada for FFO versus EPS. Baik et al. (2008) also find that REIT analysts are more likely to provide FFO-
only forecasts. We also document significant time variation in EPS analysts’ forecasts for REITs, with
relative analyst forecast reporting as low as 5 percent of the FFO sample in the mid- to later-1990s.
5.1.2. Cumulative abnormal returns

Table 3 reports cumulative abnormal returns (CARs) around FFO and EPS announcements over our
1995–2009 sample. A positive surprise occurs when actual FFO (EPS) exceeds the mean analyst fore-
cast, while a negative surprise occurs when actual FFO (EPS) is lower than the mean forecast. Deciles 1
and 10 are the largest negative and positive surprise decile portfolios, respectively.10
Turning first to the FFO CAR results in Panel A of Table 3, we find little evidence of a pre-
announcement drift. Consistent with the literature, CAR(60,2) and CAR(60,1) are not signifi-
cantly different from zero. This result also holds if the sample is restricted to the largest negative
(Decile 1) and largest positive (Decile 10) earnings surprises when we include announcements with
one analyst forecast. However, when at least two analyst forecasts are required, the largest positive
(negative) surprises are associated with significantly higher (lower) abnormal returns over both the
59- and 58-day windows prior to the FFO announcement.
We next examine the two- and three-day cumulative abnormal returns around the FFO
announcement [i.e., CAR(1,0), CAR(0,1) and CAR(1,1)]. Over all three event windows, negative
earnings surprises are associated with negative CARs that are significant at the one percent level. In
contrast, positive surprises are associated with positive and highly significant CARs around the
announcement date. Clearly, FFO announcements contain new information.11
In sharp contrast to the highly significant positive and negative post-earnings announcement drift
(PEAD) typically found in the literature (Kim and Kim, 2003, is an exception), we find no evidence of
a post-FFO earnings announcement drift in our aggregate sample or when the sample is restricted to
the largest negative surprises. However, we find strong evidence of a PEAD among the largest positive
FFO surprises.
Overall, our reported weaker evidence of a pre-announcement drift for the largest decile surprises,
along with our stronger evidence of a positive post-announcement earnings drift for the largest
positive decile of FFO surprises, is consistent with some use of earnings management by REIT managers
for the largest decile surprise portfolios, although the guidance is incomplete given the size of the
surprise and post-drift. Our positive post-drift results are also consistent with Bartov et al. (2002) and
Lopez and Rees (2002) who provide evidence that the rewards for firms that beat analyst forecast is
higher than the penalty for those that fail to meet expectations. Basu (1997) also notes that conser-
vatism encourages investors to anticipate bad news and to regard it as a one-time earnings shock. In
contrast, good news is recognized gradually over an extended period of time. As a result, the market’s
reaction will be stronger for positive than negative earnings surprises.
In Panel B of Table 3 we report the corresponding CARs for EPS surprises. Similar to our FFO results,
we find no evidence of a statistically significant pre-announcement drift in CARs. For CAR(1,1), we
find a somewhat similar pattern to the FFO results, although the magnitude and significance of the EPS
results vary. Negative (positive) surprises are associated with significantly negative (positive) CARs
over the three-day window. However, the one analyst forecast aggregate result is not significantly
different from zero and its two analyst forecast counterpart result is only marginally significant at the
10 percent level.
10
An issue that has received less attention is whether the market responds equally to positive and negative earnings
surprises. Shipper (1991) summarizes evidence from several empirical studies and suggests that analysts, on average, tend to be
optimistic in their forecasts. Freeman et al. (1992) hypothesize that if investors are aware of the bias there should be stronger
market responses to positive earnings surprises than to negative surprises of equivalent magnitudes. However, they find no
empirical support for their hypothesis. In a more recent study, Brown (2001) finds that the median earnings surprise was
negative in the eighties, zero in the early nineties and positive in the mid to late nineties. This suggests that if the bias in
analysts’ forecasts is the sole explanation for an asymmetric response to earnings surprises, the pessimistic bias in analysts’
estimates in the late nineties should lead to stronger market reaction to negative surprises. Conrad et al. (2002) find that the
market’s response to a negative surprise is stronger than the response to a positive surprise, while Skinner and Sloan (2002)
find a similar result but show that the asymmetrical response holds only for growth stocks.
11
The CARs decomposed by property types are similar to each other for U.S. REITs. We also find that the Canadian REIT CARs
are similar to the U.S. CAR results. While the signs are similar to the U.S. CAR results for the CARs calculated using the other
country’s available data in Table 1, the cross-country CAR results are largely not different from zero given the large
measurement errors from the very small sample sizes of available REITs in each country.
Table 3
Cumulative abnormal announcement returns. This table reports the cumulative abnormal returns around the announcement of
FFO and EPS by Equity REITs listed on the NYSE and AMEX between 1995 and 2009. The cumulative abnormal returns are
estimated based on the market model using the value-weighted daily return data from CRSP and an estimation period of 250
days ending 60 days prior to the event. CAR(60,2), (60,1) are the cumulative abnormal returns 59 and 58 days prior to the
announcement while CAR(1,1) and CAR(0,1) are the 3 day and 2 day cumulative abnormal returns around the announcement
respectively. CAR(1,60) and CAR(2,60) are the cumulative abnormal returns for the 59 and 58 days following the announcement
respectively. SURPRISE is defined as the difference between actual FFO (EPS) and expected FFO (EPS) scaled by price at the end of
the quarter. Deciles 1 and 10 are the largest negative and positive surprise decile portfolios respectively.
CAR(60,2) CAR(60,1) CAR(1,1) CAR(1,0) CAR(0,1) CAR(1,60) CAR(2,60)

Panel A: FFO surprises
At least 1 analyst forecast
Negative surprise 0.37% 0.52% 0.94%*** 0.54%*** 0.79%*** 0.12% 0.29%
Positive surprise 0.10% 0.01% 0.41%*** 0.34%*** 0.33%*** 0.21% 0.06%
At least 2 analyst forecasts
Positive surprise 0.02% 0.11% 0.42%*** 0.33%*** 0.32%*** 0.08% 0.07%
Decile 1 0.74% 0.78% 0.96%*** 0.52%*** 0.92%*** 0.62% 1.05%
Decile 10 0.35% 0.54% 0.77%*** 0.55%*** 0.58%*** 1.97%*** 1.74%***
Decile 1 0.43%* 0.52%** 1.18%*** 0.67%*** 1.08%*** 0.38% 0.89%
Decile 10 0.90%** 1.12%** 0.78%*** 0.57%*** 0.57%*** 1.29%** 1.08%**
Panel B: EPS surprises
Positive surprise 0.47% 0.52% 0.17% 0.19% 0.22%** 0.19% 0.16%
Positive surprise 0.97% 1.02% 0.28%* 0.11% 0.33%*** 0.19% 0.58%
Decile 1 0.76% 0.69% 1.37%*** 0.58% 1.44%*** 1.76%* 2.55%**
Decile 10 1.15% 1.14% 0.80%*** 0.01% 0.80%*** 0.83% 1.64%**
Decile 1 1.20% 1.09% 3.01%*** 1.32%*** 3.12%** 3.46%* 5.15%***
Decile 10 2.41% 2.39% 1.26%*** 0.13% 1.24%*** 0.03% 1.10%
***, ** and * indicates statistical significance at the 1%, 5% and 10% levels respectively.
Fig. 2 plots the average cross-sectional FFO and EPS CAR(1,1) in each quarter over the 1995–2009
sample period assuming at least two analysts. Although CAR(1,1) return patterns are similar across
the FFO and EPS announcements, there are significant differences. The EPS abnormal returns display
more volatility over time, although both converge in the latter part of the sample. This is not surprising
since the EPS sample size is approximately 75 percent of the FFO sample size in the late 2000’s, while it
is only 5 percent of the FFO sample in the mid- to later-1990s, 9 percent in the early 2000’s and 46
percent in the mid 2000’s. The increasing number of EPS forecasts over time is consistent with several
Wall Street firms including EPS estimates in their REIT research reports, along with FFO, beginning in
the early 2000s.
Given sample size differences, it is not surprising that the CAR results for the two-day EPS
announcement windows are also noticeably different than the corresponding FFO results. We find
positive CARs surrounding a positive surprise only in the post-announcement window (0,1). Even
when we restrict the sample to the largest positive surprises (Decile 10), CAR(1,0) cannot be
distinguished from zero.
Similar to our FFO results, we find no evidence of a statistically significant PEAD in our aggregate
EPS data. However, when we restrict the sample to the largest negative surprises, we detect evidence of
a positive PEAD, especially for the CAR(2,60) window. Thus, although negative EPS surprises are
associated with negative CARs around the announcement date, the largest negative EPS surprises
produce significantly positive CARs over the post-announcement period. This anomalous finding is
likely a manifestation of the limited and incomplete EPS forecast coverage. Moreover, in contrast to our
FFO results, the largest positive surprises are not associated with significant CARs in the post-
Fig. 2. Abnormal REIT returns: FFO vs. EPS. This figure plots the average 3-day cumulative abnormal REIT announcement returns
(CAR(1,1)) for the time period 1995Q1–2009Q3. The cumulative abnormal returns are measured around the announcement of FFO
and EPS by Equity REITs listed on the NYSE and AMEX with at least two analyst forecasts in each period. The cumulative abnormal
returns are estimated based on the market model using value-weighted daily return data from CRSP and an estimation period of 250
days ending 60 days prior to the event.
announcement period. In fact, in one case [CAR(2,60) with at least two analyst forecasts] the cumu-
lative abnormal return associated with the largest positive surprise is negative and significant at the 5
percent level. These mixed results are not too surprising given the relatively limited EPS forecast
coverage. Given the limited and incomplete EPS forecast coverage, we recommend caution in drawing
inferences from the EPS results.
Table 4
Descriptive statistics and correlations. This table reports descriptive statistics and correlations for our measures. The sample
period is between 2000 and 2009. Panel A reports the mean, median, standard deviation, minimum and maximum, while panel
B presents the correlations among the variables. CAR(1,1) is the three-day cumulative abnormal returns around U.S. REIT FFO
announcements (with at least two analyst forecasts in each period), SURPRISE is defined as the difference between actual FFO and
expected FFO scaled by price at the end of the quarter, SIZE is the natural log of market capitalization at the end of the quarter,
and BM is the ratio of the book value of equity to the market value of equity. RESENT is an indirect measure of real estate market
sentiment estimated using the first principal component extracted from seven underlying proxies of investor sentiment in
commercial real estate markets similar to Baker and Wurgler (2006, 2007). VIX is the Chicago Board Options Exchange (CBOE)
S&P 500 volatility index, and IVOL is idiosyncratic volatility measured as the standard deviation of the residuals from regressing
daily firm-level excess returns on the Fama–French factors using a 1-month window. REVISION is the number of times analysts
revise FFO estimates in a quarter scaled by the number of analysts, and DISPERSION is the standard deviation of analysts’ forecast
scaled by absolute mean forecast.
Variable CAR(1,1) SURPRISE SIZE BM RESENT VIX IVOL REVISION DISPERSION

Panel A: descriptive statistics
Mean 0.002 0.000 7.037 0.646 0.181 21.805 0.013 0.854 0.037
Median 0.001 0.000 7.076 0.568 0.214 20.190 0.010 0.000 0.022
Std. Dev. 0.042 0.003 1.116 0.440 0.909 9.914 0.009 1.566 0.104
Min 0.508 0.026 2.271 0.000 1.884 9.970 0.001 0.000 0.000
Max 0.562 0.011 10.121 7.515 1.639 80.060 0.199 16.000 4.500
Panel B: correlations
CAR(1,1) 1.000
SURPRISE 0.112*** 1.000
SIZE 0.019 0.115*** 1.000
BM 0.080*** 0.087*** 0.397*** 1.000
RESENT 0.031** 0.044*** 0.070*** 0.249*** 1.000
VIX 0.026 0.033** 0.108*** 0.240*** 0.623*** 1.000
IVOL 0.053*** 0.122*** 0.150*** 0.383*** 0.413*** 0.494*** 1.000
REVISION 0.025 0.006 0.223*** 0.053*** 0.077*** 0.115*** 0.135*** 1.000
DISPERSION 0.004 0.110*** 0.119*** 0.223*** 0.033 0.019 0.119*** 0.104*** 1.000
Fig. 3. Real estate investor sentiment, uncertainty, and abnormal REIT returns. This figure plots real estate investor sentiment,
market-wide uncertainty, and firm-level uncertainty against average 3-day cumulative abnormal REIT announcement returns
(CAR(1,1)) for the time period 1995Q1–2009Q3 in Panels A, B, and C respectively. Panel D plots the relation between VIX and
idiosyncratic volatility, IVOL. The cumulative abnormal returns are measured around the announcement of FFO by Equity REITs listed
on the NYSE and AMEX with at least two analyst forecasts in each period. The cumulative abnormal returns are estimated based on
the market model using value-weighted daily return data from CRSP and an estimation period of 250 days ending 60 days prior to
the event. Real estate sentiment, RESENT, is the first principal component extracted from seven underlying proxies of investor
sentiment in commercial real estate markets similar to Baker and Wurgler (2006, 2007). Market-wide uncertainty, VIX, is the
Chicago Board Options Exchange (CBOE) S&P 500 volatility index. Firm-level, uncertainty, REVISION, is the number of times analysts
revise FFO estimates in a quarter scaled by the number of analysts. IVOL, another measure of firm-level uncertainty, is idiosyncratic
volatility measured as the standard deviation of the residuals from regressing daily firm-level excess returns on the Fama–French
factors using a 1-month window. Panel A: cumulative abnormal returns and real estate investor sentiment, Panel B: cumulative
abnormal returns and VIX, Panel C: cumulative abnormal returns and revision, Panel D: VIX and idiosyncratic volatility (IVOL).
Table 4 reports descriptive statistics and correlations for FFO CAR(1,1), FFO surprises, investor
sentiment, market-wide uncertainty, firm-level uncertainty, and our key control variables. Panel A
reports means, medians, standard deviations, minimums and maximums, while Panel B presents the
contemporaneous correlations among the variables. CAR(1,1) is positively related to SURPRISE and BM
but negatively related to RESENT and IVOL. REVISION and DISPERSION are not unconditionally related to
CAR(1,1). Sentiment is positively correlated with the magnitude of surprise. In contrast, VIX and IVOL
are negatively related to SURPRISE. That is, more uncertainty is unconditionally correlated with smaller
Fig. 3. (Continued).
earnings surprises. The correlation between VIX and RESENT is negative and very high (r ¼ 0.623),
suggesting that periods of high market uncertainty are associated with low investor sentiment. There is
also a strong positive correlation between VIX and IVOL, indicating that REIT-level uncertainty and
market-wide uncertainty have a common component.
Fig. 3 provides plots of our key variables over the sample period. In Panel A, we display the average
CAR(1,1) in each quarter for FFO announcements along with real estate investor sentiment. The
persistent negative relation between CAR(1,1) and real estate investor sentiment is clearly visible. In
the latter 1990s, when real estate investor sentiment is relatively high, the CARs are below zero. During
the early 2000’s as real estate investor sentiment drops, the CARs follow a similar pattern. However, the
two series begin to move inversely again by 2003 and continue to do so throughout the rest of the 2000’s.
Interestingly, although the correlation between VIX and CAR(1,1) is not statistically significant, the
plot of these two series in Panel B of Fig. 3 clearly reveals periods of a persistent inverse relation as well
as periods during which market-wide uncertainty precedes the CARs. Our firm-level uncertainty proxy,
REVISION, has a similar relation with CARs, as shown in Panel C of Fig. 3. Our subsequent conditional
analysis provides a clearer picture of the influence of firm-level and market-wide uncertainty on CARs.
Finally, in Panel D of Fig. 3 we plot VIX and IVOL. There is a significant positive relation between these
two series over time as also noted in the descriptive statistics (r ¼ 0.494). To better disentangle the
Table 5
Regression of cumulative U.S. REIT abnormal returns on FFO and EPS surprises. This table presents regression results of the three-
day cumulative abnormal returns (CAR(1,1)) on FFO surprises (panel A) and EPS surprises (panel B) for five sub-periods: 1995–
1999, 2000–2003, 2004–2006, 2007–2009, 2000–2009; and the full sample period: 1995–2009. The three-day cumulative
abnormal returns is estimated using the market model and SURPRISE is defined as the difference between actual FFO (EPS) and
expected FFO (EPS) scaled by the share price at the end of the prior quarter (with at least two analyst forecasts in each period). All
the regressions include year dummies and standard errors are adjusted for clustering by firm following Petersen (2009). P-values
are in parenthesis.
1995–1999 2000–2003 2004–2006 2007–2009 2000–2009 1995–2009

Panel A: FFO surprises
CONSTANT 0.004** (0.014) 0.005*** (0.009) 0.003** (0.039) 0.003 (0.153) 0.005*** (0.010) 0.003* (0.055)
SURPRISE 0.045 (0.926) 1.600** (0.049) 3.586*** (0.000) 1.181* (0.076) 1.672*** (0.000) 1.361*** (0.000)
Adjusted R2 0.006 0.017 0.065 0.009 0.017 0.016
N 1416 1561 1242 956 3759 5175
Panel B: EPS surprises
CONSTANT 0.008 (0.170) 0.117 (0.285) 0.007** (0.015) 0.005** (0.044) 0.002 (0.835) 0.014*** (0.000)
SURPRISE 0.415 (0.267) 1.093*** (0.002) 0.330 (0.184) 0.695*** (0.006) 0.666*** (0.000) 0.466*** (0.000)
2
Adjusted R 0.010 0.062 0.007 0.043 0.054 0.051
N 71 142 568 713 1423 1494
market-wide and firm-level uncertainty effects associated with this proxy, we also orthogonalize IVOL
with respect to VIX before including IVOL in our CAR regression models.
5.2. Regression results
5.2.1. Cumulative abnormal returns around FFO announcements

Table 5 provides results from our regressions of three-day cumulative abnormal returns on FFO
surprises (Panel A) and EPS surprises (Panel B) for the full sample period (1995–2009) and five sub-
periods (1995–1999, 2000–2003, 2004–2006, 2007–2009, and 2000–2009). The sub-periods also
allow us to observe the information content of FFO and EPS over time, particularly after the changes
NAREIT made to FFO in 2000 and SEC’s rules changes in 2003.12 The three-day CARs are estimated
using the market model. All regressions include year dummies; standard errors are adjusted for
clustering by firm (Petersen, 2009). P-values are reported in parentheses.
Looking first at the last column in Table 5, we find that the estimated coefficient on SURPRISE is
positive and highly significant over the 1995–2009 sample period in both the FFO and EPS specifica-
tions. This result is consistent with the earnings announcement literature. There is also significant time
variation in the number of EPS forecasts, with relative analyst forecast reporting as low as 5 percent of
the FFO sample in the mid- to later-1990s, 9 percent in the early 2000’s, 46 percent in the mid 2000’s,
and reaching 75 percent by the late 2000’s.
To better understand the potential time variation in the relation between earnings surprises and
CARs, we provide sub-period estimates in the first five columns. For the 1995–1999 period, the esti-
mated coefficient on SURPRISE in both the FFO and EPS regressions cannot be distinguished from zero.
These results are consistent with Vincent (1999) who finds unexpected FFO (EPS) based on analysts’
forecasts to be unrelated to announcement period abnormal returns over 1994–1996 sample period.
However, the coefficient on SURPRISE in the FFO regressions is also positive and significant in the
remaining four sub-periods. The estimated coefficient on SURPRISE in the EPS regressions is positive
and highly significant in 2000–2003 and 2007–2009, but cannot be distinguished from zero during the
2004–2006 sub-period. Interestingly, this is the sub-period during which the magnitude and signifi-
cance of the SURPRISE coefficient in the FFO regressions is at its maximum.
Overall, the results in Table 5 suggest that both FFO and EPS announcements contain valuable
information, except during the 1995–1999 window. Devos et al. (2007) also find less analyst forecast
12
In 2003, the SEC introduced Regulation G, which requires firms to show a reconciliation of each non-GAAP financial
measure to the most directly comparable GAAP measure.
Table 6
Firm-level uncertainty, market-wide uncertainty and real estate sentiment effects on cumulative abnormal returns around U.S.
REIT FFO Announcements. This table presents the results of regressions of the three-day cumulative abnormal returns
(CAR(1,1)) on real estate market sentiment (RESENT), market-wide uncertainty (VIX), and firm-level uncertainty measures
(REVISION and DISPERSION) for the period 2000–2009. The three-day cumulative abnormal return is estimated using the market
model. SURPRISE is defined as the difference between actual FFO and expected FFO scaled by the share price at the end of the
prior quarter (with at least two analyst forecasts in each period). SIZE is the natural log of the REIT’s market capitalization at the
end of the quarter, and BM is REIT’s ratio of the book value of equity to the market value of equity. RESENT is an indirect measure
of real estate sentiment estimated using the first principal component extracted from seven underlying proxies of investor
sentiment in commercial real estate markets similar to Baker and Wurgler (2006, 2007). VIX is the Chicago Board Options
Exchange (CBOE) S&P 500 volatility index. REVISION is the number of times analysts revise their FFO estimates in the quarter
prior to the announcement scaled by the number of analysts, and DISPERSION is the standard deviation of analysts’ forecast
scaled by absolute mean forecasts. REVISION*SURPRISE, VIX*SURPRISE and RESENT*SURPRISE are the interactions of REVISION and
SURPRISE, VIX and SURPRISE and RESENT and SURPRISE respectively. All the regressions include year dummies and standard errors
are adjusted for clustering by firm following Petersen (2009). P-values are in parenthesis.
(1) (2) (3) (4) (5)

CONSTANT 0.002 (0.812) 0.003 (0.703) 0.002 (0.788) 0.002 (0.786) 0.003 (0.709)
SURPRISE 1.736*** (0.000) 2.612*** (0.000) 2.829* (0.057) 2.540*** (0.000) 1.983 (0.224)
SIZE 0.001 (0.402) 0.001 (0.509) 0.001 (0.432) 0.001 (0.430) 0.001 (0.515)
BM 0.008* (0.092) 0.008* (0.079) 0.008* (0.096) 0.008* (0.095) 0.008* (0.085)
REVISION 0.001 (0.311) 0.001 (0.405) 0.001 (0.303) 0.001 (0.314) 0.001 (0.416)
DISPERSION 0.002 (0.901) 0.001 (0.940) 0.002 (0.894) 0.001 (0.972) 0.001 (0.963)
VIXa 0.004* (0.065) 0.004* (0.059) 0.004* (0.058) 0.004* (0.064) 0.004* (0.063)
RESENT 0.005*** (0.000) 0.005*** (0.000) 0.005*** (0.000) 0.005*** (0.000) 0.005*** (0.000)
REVISION*SURPRISE 0.779** (0.013) 0.736** (0.017)
VIX*SURPRISE 0.041 (0.500) 0.058 (0.456)
RESENT*SURPRISE 0.979** (0.041) 1.189* (0.072)
Adjusted R2 0.027 0.034 0.028 0.031 0.041
a
The coefficient for VIX has been multiplied by a factor of 10.
bias post-2000 and significant increases in analysts coverage in the 2000’s, peaking at 73 percent at the
end of their sample in 2004. Because of the limited observations in the 1995–1999 sample period and
changes in FFO definitions and rules after this period, we focus the remainder of our regression analysis
on the 2000–2009 sample period. We also focus our analysis on FFO announcements given the much
more extensive analysts’ coverage for FFO. However, our key findings are robust to using the larger
sample period as well as using EPS for the 2000–2009 sample period.
5.2.2. Uncertainty, sentiment, and CARs

We build on the analysis of FFO announcements and abnormal returns presented in Table 5 by
adding several control variables to our CAR regressions, along with proxies for firm uncertainty, market
uncertainty, and investor sentiment. The results obtained from estimating these augmented abnormal
return specifications over the 2000–2009 time period are presented in Table 6. We use CARs over
a three-day window [CAR(1,1)] as the dependent variable. All regression specifications include year
dummies and standard errors are adjusted for clustering by firm following Petersen (2009).
The estimated coefficient on SURPRISE in Model (1) is positive and highly significant. Firm size is not
associated with three-day CARs, although we detect a weak positive relation between a firm’s book-to-
market ratio and CARs. The estimated coefficients on both REVISION and DISPERSION are consistently
insignificant. Thus, we find no conditional evidence that uncertainty about firm-level FFO projections is
associated with CARs, controlling for the influence of market-wide uncertainty and investor sentiment.
However, the estimated coefficient on VIX is negative and significant (p-value ¼ 0.065), suggesting
abnormal returns surrounding an FFO announcement are inversely related to market-wide uncer-
tainty.13 Finally, the estimated coefficient on RESENT is negative and highly significant. That is, periods
of high investor sentiment predict lower abnormal returns surrounding an FFO announcement, all else
13
The estimated coefficient on VIX has been multiplied by a factor of 10.
equal. These effects are also economically significant with a respective RESENT and VIX estimated
impact of 11.7% and 12.7% on the abnormal returns from a one standard deviation change in these
variables using the standardized coefficient estimates.
The first regression specification reported in Table 6 provides evidence on the relation between
abnormal announcement returns and our three variables of primary interest: firm uncertainty, market
uncertainty, and sentiment. However, we are also interested in the extent to which these variables
affect the relation between SURPRISE and abnormal returns. We therefore interact SURPRISE with each
of our three uncertainty variables and then separately add each interaction variable to our abnormal
return regression [Models (2)–(4)].
The estimated coefficient on REVISION*SURPRISE [Model (2)] is negative and significant (p-
value ¼ 0.013), with an estimated economic impact of 11.6% from a one standard deviation change.
This suggests greater firm uncertainty among analysts about a firm’s end-of-quarter FFO reduces the
effect of an FFO surprise on abnormal returns. This result is consistent with the findings in Imhoff and
Lobo (1992) and Francis et al. (2007) who find that estimated earnings response coefficients are lower
for firms with greater information uncertainty. Note that the inclusion of the interaction term
strengthens the estimated coefficient on SURPRISE. The coefficient on VIX*SURPRISE [Model (3)] cannot
be distinguished from zero; thus, although market uncertainty is associated with lower abnormal
returns it does not alter the effect of an FFO surprise on abnormal returns.
Evidence exists in the literature that investor sentiment can push prices away from fundamental
values (e.g., De Long et al., 1990; Shleifer and Vishny, 1997 and Daniel et al., 1998). However, periods of
high investor sentiment are often followed by decreased abnormal returns as prices revert toward their
fundamental values. Currently, no evidence exists in the literature on whether investor sentiment
amplifies or dampens the effects of an earnings surprise on abnormal returns. The estimated coefficient
on RESENT*SURPRISE [Model (4)] is positive and significant (p-value ¼ 0.041, estimated economic
impact of 9.48%), indicating that the effect of an FFO surprise on abnormal returns is greater in periods
of high sentiment. This result is consistent with a number of theoretical predictions such as those by De
Long et al. (1990). That is, in periods of high investor sentiment, asset prices are more likely to be
determined at the margin by optimistic noise traders among whom there is less dispersion in
expectations about expected future cash flows. In such a market, the price impact associated with
earnings surprises is magnified.
In the last specification in Table 6, we include all three interaction terms in a single regression
specification as an additional robustness check. The estimated coefficient on REVISION*SURPRISE
remains negative and significant, the coefficient on RESENT*SURPRISE remains positive and significant
(although with slightly reduced statistical significance), and the VIX*SURPRISE interaction remains
indistinguishable from zero. However, controlling simultaneously for these three surprise interactions
reduces the magnitude of the SURPRISE coefficient and negates its statistical significance.
5.2.3. Robustness checks

The results reported in Table 6 are estimated using REVISION and DISPERSION to capture firm-level
uncertainty. However, these variables may be endogenous to our SURPRISE variable and can therefore
be problematic in drawing firm-level uncertainty inferences from them. As a robustness check, we
replace REVISION and DISPERSION in our specifications with IVOL, our additional proxy for firm-level
uncertainty.14 The results from estimating these revised specifications are reported in Table 7.
It is important to note that the coefficient estimates on SURPRISE are little changed by the substi-
tution of IVOL for REVISION and DISPERSION; FFO surprises remain positively and significantly related to
CARs around FFO announcement dates. Firm size remains insignificant, but the magnitude and
significance of BM is increased. The estimated coefficient on IVOL is consistently negative and signif-
icant across the five specifications reported in Table 7. Moreover, the use of IVOL increases the
magnitude and significance of the negative coefficient on VIX. The coefficient on RESENT remains
negative and highly significant; the estimated coefficient on RESENT*SURPRISE remains positive and
14
To avoid the collinearity problem resulting from the high positive correlation between IVOL and VIX, we use the residuals
from a regression of IVOL on VIX.
Table 7
Idiosyncratic volatility and cumulative abnormal returns around FFO announcements. This table presents the results of
regressions of the three-day cumulative abnormal returns (CAR(1,1)) on market sentiment (RESENT), market-wide uncertainty
(VIX), and firm-level uncertainty using idiosyncratic volatility (IVOL) for the period 2000–2009. The 3-day cumulative abnormal
return is estimated using the market model. SURPRISE is defined as the difference between actual FFO and expected FFO scaled by
the share price at the end of the prior quarter (with at least two analyst forecasts in each period). SIZE is the natural log of the
REIT’s market capitalization at the end of the quarter, and BM is REIT’s ratio of the book value of equity to the market value of
equity. RESENT is an indirect measure of real estate sentiment estimated using the first principal component extracted from
seven underlying proxies of investor sentiment in commercial real estate markets similar to Baker and Wurgler (2006, 2007). VIX
is the Chicago Board Options Exchange (CBOE) S&P 500 volatility index, and IVOL is idiosyncratic volatility measured as the
standard deviation of the residuals from regressing daily firm-level excess returns on the Fama–French factors using a 1-month
window. IVOL*SURPRISE, VIX*SURPRISE and RESENT*SURPRISE are the interactions of IVOL and SURPRISE, VIX and SURPRISE and
RESENT and SURPRISE respectively. All the regressions include year dummies and standard errors are adjusted for clustering by
firm following Petersen (2009). P-values are in parenthesis.
(1) (2) (3) (4) (5)

CONSTANT 0.000 (0.983) 0.001 (0.954) 0.001 (0.943) 0.001 (0.933) 0.000 (0.986)
SURPRISE 1.701*** (0.000) 1.900*** (0.001) 2.511* (0.057) 2.472*** (0.000) 1.247 (0.428)
SIZE 0.000 (0.621) 0.000 (0.633) 0.000 (0.639) 0.000 (0.656) 0.000 (0.633)
BM 0.011** (0.036) 0.011** (0.033) 0.011** (0.036) 0.011** (0.040) 0.011** (0.042)
IVOL 0.392* (0.068) 0.440* (0.053) 0.402* (0.062) 0.413* (0.056) 0.409* (0.073)
VIXa 0.005** (0.039) 0.005** (0.038) 0.005** (0.034) 0.005** (0.038) 0.005** (0.042)
RESENT 0.005*** (0.000) 0.005*** (0.000) 0.005*** (0.000) 0.005*** (0.000) 0.005*** (0.000)
IVOL*SURPRISE 18.737 (0.579) 2.639 (0.932)
VIX*SURPRISE 0.030 (0.610) 0.061 (0.408)
RESENT*SURPRISE 1.005** (0.048) 1.477** (0.016)
Adjusted R2 0.034 0.034 0.034 0.038 0.039
a
The coefficient for VIX has been multiplied by a factor of 10.
significant. The coefficient on IVOL*SURPRISE is not statistically significant. Overall, the results reported
in Tables 6 and 7 suggest that firm-level uncertainty, market uncertainty, and investor sentiment
influence the magnitude of abnormal announcement returns as well as the effect of unexpected
earnings on abnormal returns.
The market model used to calculate cumulative abnormal returns is a single-factor asset pricing
model. As such, the CARs may be biased by the exclusion of other potential asset pricing factors. To
address this concern, we rerun the five model specifications reported in Tables 6 and 7 after adding
SMB, HML, momentum, and liquidity to each specification. Although not separately tabulated, the
magnitude and significance of the estimated coefficients on SURPRISE are little altered by the
inclusion of these additional factors: positive FFO surprises are still associated with significantly
higher CARs. Similarly, the coefficient estimate on VIX remains negative and marginally significant.
Although the coefficients on SMB, HML, and liquidity cannot be distinguished from zero in any of the
five specifications, the estimated coefficient on momentum is uniformly positive and often signifi-
cant. That is, general stock market momentum is positively associated with CARs around FFO
announcements. This is consistent with the findings of Naranjo and Porter (2010) who report
persistent momentum in portfolio returns using 17,000 firms across 40 countries. Overall, the
augmented CAR specifications do not materially alter the estimated coefficients on REVISION,
DISPERSION, IVOL, VIX, or RESENT, or the coefficients on the interaction of these three variables with
SURPRISE.
5.2.4. The influence of uncertainty and sentiment on pre- and post-earnings announcement drift
As reported in Table 3, we find no pre-announcement or post-announcement drift in excess
returns in our aggregate sample. However, we do find some weak evidence of a positive and
negative pre-announcement drift for the largest decile surprises as well as stronger evidence of
a positive post-announcement earnings drift for the largest positive decile of FFO surprises. To
further investigate the influence of uncertainty and sentiment on cumulative abnormal returns,
we re-estimate our CAR regression using the most negative (Decile 1) and most positive (Decile
10) FFO surprises. The results using REVISION and DISPERSION to capture firm-level uncertainty are
Table 8
The influence of sentiment and uncertainty on pre- and post-earnings announcement drift. This table presents the results of
regressions of the 58-day pre- and post-announcement cumulative abnormal returns for deciles 1 and 10 on real estate market
sentiment (RESENT), market-wide uncertainty (VIX) and firm-level uncertainty measures (REVISION and DISPERSION) for the
period 2000–2009. Deciles 1 and 10 are the largest negative and positive surprise decile portfolios respectively. CAR(60,2)
and CAR(2,60) are the 58-day cumulative abnormal returns respectively and are estimated using the market model. SURPRISE is
defined as the difference between actual FFO and expected FFO scaled by the share price at the end of the prior quarter (with at
least two analyst forecasts in each period). SIZE is the natural log of the REIT’s market capitalization at the end of the quarter, and
BM is REIT’s ratio of the book value of equity to the market value of equity. RESENT is an indirect measure of real estate sentiment
estimated using the first principal component extracted from seven underlying proxies of investor sentiment in commercial real
estate markets similar to Baker and Wurgler (2006, 2007). VIX is the Chicago Board Options Exchange (CBOE) S&P 500 volatility
index. REVISION is the number of times analysts revise FFO estimate in a quarter scaled by the number of analysts, and
DISPERSION is the standard deviation of analysts’ forecast scaled by absolute mean forecasts. All the regressions include year
dummies and standard errors are adjusted for clustering by firm following Petersen (2009). P-values are in parenthesis.
Decile 1 Decile 10
CAR(60,2) CAR(2,60) CAR(60,2) CAR(2,60)

CONSTANT 0.022 (0.760) 0.000 (0.975) 0.211** (0.018) 0.104 (0.119)
SURPRISE 5.216** (0.038) 13.822** (0.011)
SIZE 0.008 (0.312) 0.020*** (0.003) 0.022** (0.017) 0.005 (0.418)
BM 0.036 (0.127) 0.043** (0.036) 0.143*** (0.000) 0.055*** (0.006)
REVISION 0.021*** (0.004) 0.000 (0.994) 0.011 (0.446) 0.002 (0.699)
DISPERSION 0.004 (0.836) 0.050** (0.023) 0.075 (0.462) 0.106** (0.029)
RESENT 0.044*** (0.008) 0.080*** (0.000) 0.040** (0.026) 0.058*** (0.000)
VIXa 0.020 (0.116) 0.010 (0.728) 0.002 (0.866) 0.030** (0.031)
Adjusted R2 0.076 0.147 0.274 0.333
a
The coefficient estimate for VIX has been multiplied by a factor of 10.
reported in Table 8; the corresponding results using IVOL as a proxy for firm-level uncertainty are
reported in Table 9.
The estimated coefficient on SUPRISE in the Decile 10 CAR(2,60) regression (column 3) is positive and
highly significant. That is, among the largest positive surprises, variation in the magnitude of the surprise
is positively associated with abnormal returns during the post-announcement period. In contrast,
Table 9
Idiosyncratic volatility and pre- and post-earnings announcement drift. This table presents the results of regressions of the 58-
day pre- and post-announcement cumulative abnormal returns for deciles 1 and 10 on real estate market sentiment (RESENT),
market uncertainty (VIX) and firm-level uncertainty measures (IVOL) for the period 2000–2009. Deciles 1 and 10 are the largest
negative and positive surprise decile portfolios respectively. CAR(60,2) and CAR(2,60) are the 58-day cumulative abnormal
returns respectively and are estimated using the market model. SURPRISE is defined as the difference between actual FFO and
expected FFO scaled by the share price at the end of the prior quarter (with at least two analyst forecasts in each period). SIZE is
the natural log of the REIT’s market capitalization at the end of the quarter, and BM is REIT’s ratio of the book value of equity to
the market value of equity. RESENT is an indirect measure of real estate sentiment estimated using the first principal component
extracted from seven underlying proxies of investor sentiment in commercial real estate markets similar to Baker and Wurgler
(2006, 2007). VIX is the Chicago Board Options Exchange (CBOE) S&P 500 volatility index. IVOL is idiosyncratic volatility
measured as the standard deviation of the residuals from regressing daily firm-level excess returns on the Fama–French factors
using a 1-month window. All the regressions include year dummies and standard errors are adjusted for clustering by firm
following Petersen (2009). P-values are in parenthesis.
Decile 1 Decile 10
CAR(60,2) CAR(2,60) CAR(60,2) CAR(2,60)

CONSTANT 0.070 (0.487) 0.443*** (0.000) 0.430*** (0.000) 0.358*** (0.000)
SURPRISE 3.991 (0.110) 14.156*** (0.009)
SIZE 0.002 (0.765) 0.020*** (0.002) 0.014** (0.038) 0.004 (0.438)
BM 0.020 (0.382) 0.037* (0.067) 0.119** (0.012) 0.082*** (0.000)
VIXa 0.030* (0.096) 0.010 (0.603) 0.010 (0.494) 0.030*** (0.002)
IVOL 0.097 (0.956) 2.569** (0.027) 2.856** (0.038) 3.876*** (0.000)
RESENT 0.038** (0.024) 0.075*** (0.000) 0.049*** (0.002) 0.062*** (0.000)
Adjusted R2 0.047 0.164 0.311 0.334
a
The coefficient estimate for VIX has been multiplied by a factor of 10.
variation in the magnitude of the surprise is negatively associated with abnormal returns during the
post-announcement period for the Decile 1. The estimated coefficient on REVISION is negative and highly
significant during the pre-announcement period among the largest negative surprises, and DISPERION is
statistically significant in the post-announcement period for both deciles. Market-wide uncertainty,
however, is only significant in the post-announcement period among the largest positive surprises. The
estimated coefficient on IVOL is also significant in three of the four specifications reported in Table 9.
Overall, firm-level uncertainty appears to explain a significant portion of pre- and post-abnormal returns
among the largest positive and negative surprises in the sample.
The most consistent and, arguably, most important result reported in Tables 8 and 9 is the
consistently negative and highly significant coefficient on RESENT. That is, higher levels of investor
sentiment consistently predict lower abnormal returns over both the pre- and post-announcement
windows.
6. Summary and conclusion
Research assessing the information content of earnings announcements is extensive. A key

finding is that cumulative abnormal returns continue to drift up (down) after firms announce
earnings that are greater (less) than forecasted. This post-earnings announcement drift (PEAD)
finding is inconsistent with the efficient markets hypothesis. While no explanation for PEAD has
proven to be complete or sufficient, the literature has provided several explanations that can be
categorized as either rational or behavioral/market inefficiency based. These explanations include
the use of misspecified asset pricing models in the calculation of abnormal returns, investor
underreaction to earnings signals during the announcement period with a subsequent correction
from new information (overreaction), delayed information processing biases, as well as and other
behavioral biases.
We extend the literature by using public real estate markets (REITs) to test for the existence and
sources of abnormal returns around earnings announcements. The public REIT market provides
a unique testing ground given the existence of a parallel private real estate market in which properties
very similar to those held by REITs are transacted and an ownership structure and information envi-
ronment that potentially reduce information uncertainty, including the wide use of Funds from
Operations (FFO) as a common valuation metric. Our tests on the role of investor sentiment and
market-wide uncertainty also provide a unique contribution to the broader literature on earnings
announcement effects.
We first document the relative reporting of FFO and EPS analysts’ forecasts on U.S. REITs, as well as
REITs from several other countries (i.e., Australia, Belgium, Canada, France, Hong Kong, Japan, the
Netherlands, and UK). We provide some descriptive characteristics associated with these international
samples and show that coverage outside of the U.S. is very limited. Consequently, we focus our analysis
on U.S. REITs.
To measure the significance of FFO and EPS announcements on REIT returns, we estimate the
cumulative abnormal returns (CARs) around the corresponding announcements for various pre-, near-
and post-event windows. We also condition on the sign and magnitude of the announcement relative
to analysts forecasts. We then examine the influence of the magnitude of the earnings surprise, real
estate market sentiment, market-level uncertainty, and firm-level uncertainty on CARs over the 1995–
2009 time period.
We find strong evidence that FFO announcements contain pricing relevant information, with highly
significant two- and three-day CARs around FFO announcements. More specifically, negative earnings
surprises relative to analysts’ forecasts are associated with negative CARs that are significant at the one
percent level, while positive surprises are associated with highly significant positive CARs around the
announcement date.
In contrast to these near-announcement CAR results, we find very limited evidence of a pre-
announcement drift in CARs. Moreover, in sharp contrast to the results typically found in the litera-
ture, we find no evidence of a post-FFO earnings announcement drift in our aggregate sample. This
result, however, is consistent with REIT prices and returns more quickly impounding information
relative to the broad equity market examined in prior studies.
The remainder of the paper contains an analysis of the determinants of firm-level CARs. We regress
CARs on earnings surprises, various fundamental variables, real estate market sentiment, and both
market-wide and firm-level information uncertainty. Controlling for time fixed effects and the
potential clustering of standard errors by firm, we find that the estimated coefficient on unexpected
earnings is positive and highly significant across the full sample period and most sub-periods. This
result is consistent with the existing broader equity earnings announcement literature. However, we
also find strong evidence that firm-level uncertainty, market-wide uncertainty, and investor sentiment
influence the magnitude of CARs. More specifically, periods of high real estate investor sentiment and
high forward looking market volatility predict lower abnormal returns surrounding FFO announce-
ments, all else equal. These findings add to the evidence in the existing literature. We also find that the
level of firm uncertainty, as measured by each firm’s idiosyncratic return volatility, has a significantly
negative impact on the abnormal returns.
Our finding that investor sentiment amplifies the effects of an earnings surprise on abnormal
returns is consistent with the noise trader theory developed by De Long et al. (1990). That is, in periods
of high investor sentiment, asset prices are more likely to be determined at the margin by optimistic
noise traders among whom there is less dispersion in expectations about expected future cash flows. In
such a market, the price impact of an earnings surprise is magnified. Our finding that greater firm-level
uncertainty among analysts about a firm’s end-of-quarter FFO reduces the effect of an FFO surprise on
abnormal returns is consistent with the findings of Imhoff and Lobo (1992) and Francis et al. (2007).
According to the European Public Real Estate Association (EPRA), there are 35 countries worldwide
that have REIT or “REIT-like” legislation in place. Many of these countries have only recently enabled
the use of REIT-like structures. As REITs and REIT analyst coverage increase across countries, future
related research analyzing other markets would be interesting given the capital market, macroeco-
nomic, legal, and institutional differences across countries. Furthermore, as FFO-like earnings metrics
come to dominate analysts’ forecasts and the reporting and analysis provided by real estate firms, as
they have in the U.S., the findings of this paper should prove useful in providing evidence on the
relative informativeness of FFO and EPS forecasts and announcements as well as the role played by
fundamentals, uncertainty, and sentiment in predicting abnormal earnings announcement returns.
References
Ang, A., Holdrick, R.J., Xing, Y., Zhang, X., 2006. The cross section of volatility and expected returns. Journal of Finance 61 (1),
259–299.
Baik, B., Billings, K.B., Morton, M.R., 2008. Reliability and transparency of non-GAAP disclosures by real estate investment trusts
(REITs). Accounting Review 83 (2), 271–301.
Baker, M., Wurgler, J., 2006. Investor sentiment and the cross-section of stock returns. Journal of Finance 61 (4), 1645–1680.
Baker, M., Wurgler, J., 2007. Investor sentiment in the stock market. The Journal of Economic Perspectives 21 (2), 129–151.
Ball, R., Brown, P., 1968. An empirical evaluation of accounting income numbers. Journal of Accounting Research 6 (2), 159–178.
Bartov, E., Givoly, D., Hayn, C., 2002. The rewards to meeting or beating earnings expectations. Journal of Accounting and
Economics 33, 173–204.
Basu, S., 1997. The conservatism principle and the asymmetric timeliness of earnings. Journal of Accounting and Economics 24,
3–37.
Beaver, W.H., 1968. The information content of annual earnings announcements. Journal of Accounting Research, 67–92.
Bergman, N.K., Roychowdhury, S., 2008. Investor sentiment and corporate disclosures. Journal of Accounting Research 46 (5),
1057–1083.
Bernard, V., 1992. Stock Price Reactions to Earnings Announcements, Advances in Behavioral Finance. Russell Sage Foundation,
New York, pp. 303–340.
Beyer, A., Cohen, D., Lys, T., Walther, B., 2010. The financial reporting environment: review of the recent literature. Journal of
Accounting and Economics 50, 296–343.
Brav, A., Heaton, J.B., 2002. Competing theories of financial anomalies. Review of Financial Studies 15 (2), 575–606.
Brounen, D., Ben-Hamo, Y., 2009. Calendar anomalies: the case of international property shares. Journal of Real Estate Finance
and Economics 38, 115–136.
Brown, G.W., Cliff, M.T., 2004. Investor sentiment and the near term stock market. Journal of Empirical Finance 11, 1–27.
Brown, G.W., Cliff, M.T., 2005. Investor sentiment and asset valuation. Journal of Business 78 (2), 405–440.
Brown, D.T., Riddiough, T.J., 2003. Financing choice and liability structure of real estate investment trusts. Real Estate Economics
31 (3), 313–346.
Brown, D.L., March/April 1997. Earnings surprise research: synthesis and perspectives. Financial Analysts Journal, 13–19.
Brown, D.L., 2001. A temporal analysis of earnings surprises: profits versus losses. Journal of Accounting Research 39 (2), 221–241.
Conrad, J., Cornell, B., Landsman, W.R., 2002. When is bad news really bad news? Journal of Finance 57 (6), 2507–2532.
Daniel, K., Hirshleifer, D., Subrahmanyam, A., 1998. Investor psychology and security market under and overreactions. Journal of
Finance 53, 1839–1886.
Daniel, K., Hirshleifer, D., Subrahmanyam, A., 2001. Overconfidence, arbitrage, and equilibrium asset pricing. Journal of Finance
56, 921–965.
Danielsen, B.R., Harrison, D.M., Van Ness, R.A., Warr, R.S., 2009. REIT auditor fees and financial market transparency. Real Estate
Economics 37 (3), 515–557.
De Long, B.J., Shleifer, A., Summers, L.H., Waldman, R.J., 1990. Noise trader risk in financial markets. Journal of Political Economy
98, 703–738.
Devos, E., Ong, S.E., Spieler, A.C., 2007. Analyst activity and firm value: evidence from the REIT sector. Journal of Real Estate
Finance and Economics 35 (3), 333–356.
Downs, D.H., Guner, Z.N., 2006. On the quality of FFO forecasts. Journal of Real Estate Research 28 (3), 259–274.
Eichholtz, P.M.A., Koedijk, K., Schweitzer, M., 2001. Global property investment and the costs of international diversification.
Journal of International Money and Finance 20, 349–366.
Eichholtz, P.M.A., Gugler, N., Kok, N., 2011. Transparency, integration and the cost of international real estate investments.
Journal of Real Estate Finance and Economics 43, 152–173.
Epstein, L.G., Turnbull, S.M., 1980. Capital asset prices and the temporal resolution of uncertainty. Journal of Finance 35 (3), 627–
643.
Fama, E., 1998. Market efficiency, long-term returns, and behavioral finance. Journal of Financial Economics 49, 283–306.
Francis, J., Ryan, L., Olson, P., Shipper, K., 2007. Information uncertainty and post-earnings announcement drift. Journal of
Business Finance and Accounting 34 (3–4), 403–433.
Freeman, Robert, Senyo, Tse., 1992. A nonlinear model of security price responses to unexpected earnings. Journal of Accounting
Research 30 (2) (Autumn), 185–209.
Gore, R., Stott, M.D., 1998. Toward a more informative measure of operating performance in the REIT industry: net income vs.
funds from operations. Accounting Horizons 12 (4), 323–339.
Graham, C.M., Knight, J.R., 2000. Cash flows vs. earnings in the valuation of equity REITs. Journal of Real Estate Portfolio
Management 6 (1), 17–25.
Griffin, D., Tversky, A., 1992. The weighting evidence and determinants of confidence. Cognitive Psychology 24, 411–435.
Gyamfi-Yeboah, F., Ziobrowski, A.J., Lambert, L.S., 2010. REITs’ price reaction to unexpected FFO announcements. Journal of Real
Estate Finance and Economics.
Hirshleifer, D., 2001. Investor psychology and asset pricing. Journal of Finance 56, 1533–1596.
Hribar, P., McInnis, J., 2012. Investor sentiment and analysts’ earnings forecast errors. Management Science 58, 293–307.
Imhoff, E.A., Lobo, G.J., 1992. The effect of ex ante earnings uncertainty on earnings response coefficients. The Accounting
Review 67 (2), 427–439.
Jegadeesh, N., Titman, S., 1993. Returns to buying winners and selling losers: implications for stock market efficiency. Journal of
Finance 48, 65–91.
Kim, D., Kim, M., 2003. A multifactor explanation of post-earnings announcement drift. Journal of Financial and Quantitative
Analysis 38 (2), 383–398.
Koedijk, K.G., Stork, P.A., 1994. Should we care? Psychological barriers in stock markets. Economics Letters 44, 427–432.
Liang, L., 2003. Post-earnings announcement drift and market participants’ information processing biases. Review of
Accounting Studies 8, 321–345.
Ling, D., Naranjo, A., Scheick, B., 2012. Investor Sentiment, Limits to Arbitrage, and Private Market Returns. University of Florida
Working Paper. University of Florida Working Paper.
Livant, J., Petrovits, C., 2009. Investor Sentiment, Post Earnings Announcement Drift and Accruals. Working Paper. New York
University, New York.
Lopez, J.T., Rees, L., 2002. The effect of beating and missing analysts’ forecasts on the information content of unexpected
earnings. Journal of Accounting, Auditing and Finance 17 (2), 155–184.
Naranjo, A., Porter, B., 2010. Risk factor and industry effects in the cross-country comovement of momentum returns. Journal of
International Money and Finance 29 (2), 275–299.
Neal, R., Wheatley, S.M., 1998. Do measures of investor sentiment predict returns? Journal of Financial and Quantitative Analysis
33, 523–547.
Ott, S.H., Riddiough, T.J., Yi, H., 2005. Finance, investment and investment performance: evidence from the REIT sector. Real
Estate Economics 33 (1), 203–235.
Pastor, L., Stambaugh, R.F., 2003. Liquidity risk and expected stock returns. Journal of Political Economy 111 (3), 642–685.
Petersen, M.A., 2009. Estimating standard errors in finance panel data sets: comparing approaches. Review of Financial Studies
22 (1), 435–480.
Price, M.S., Gatzlaff, D.H., Sirmans, C.F., 2010. Information uncertainty and the post-earnings announcement drift. Journal of Real
Estate Finance and Economics. doi:10.1007/s11146-010-9275-y.
Richardson, S., Tuna, I., Wysocki, P., 2010. Accounting anomalies and fundamental analysis: a review of recent advances. Journal
of Accounting and Economics 50, 410–454.
Robichek, A.A., Myers, S.C., 1966. Valuation of the firm: effects of uncertainty in a market context. Journal of Finance 21 (2), 215–
227.
Safdar, I., 2009. A Joint Test of Accounting Information and Investor Sentiment Theory. SSRN Working Paper #: 1483396.
Seybert, N., Wang, S., 2009. False Prophets: How Analysts Lead Short Sellers Astray during Periods of High Investor Sentiment.
University of Texas, Austin. Working Paper.
Shipper, K., 1991. Analysts forecasts. Accounting Horizons 5, 105–121.
Shleifer, A., Vishny, R., 1997. The limits of arbitrage. Journal of Finance 52, 35–55.
Skinner, D.J., Sloan, R.G., 2002. Earnings surprises, growth expectations, and stock returns or don’t let an earnings torpedo sink
your portfolio. Review of Accounting Studies 7, 289–312.
Vincent, L., 1999. The information content of funds from operations (FFO) for real estate investment trusts (REITs). Journal of
Accounting and Economics 26 (1–3), 69–104.
Yu, J., Yuan, Y., 2010. Investor sentiment and the mean-variance relation. Journal of Financial Economics 100 (2), 367–381.
Zhang, F., 2006. Information uncertainty and stock returns. Journal of Finance 61 (1), 105–137.

Information Uncertainty and Behavioral e

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Information Uncertainty and Behavioral e

Uploaded by

Copyright:

Available Formats

Journal of International Money and Finance 31 (2012) 1930–1952

Contents lists available at SciVerse ScienceDirect

Journal of International Money

Information, uncertainty, and behavioral effects: Evidence

regression analysis of cumulative abnormal returns, we ﬁnd that

2. Background literature: research context

2.2. Uncertainty, sentiment, and announcement returns

CARðs1 ; s2 Þ ¼ b0 þ b1 SURPRISE þ b2 Firm UNCERTAINTY þ b3 SURPRISE Firm UNCERTAINTY

4.1. FFO and EPS data sources and deﬁnitions

4.2. Firm-level and market-wide uncertainty proxies

4.3. Real estate investor sentiment

4.4. Control variables

5.1. Descriptive statistics

5.1.1. FFO and EPS announcements

Number of Average market Median market Observations

5.1.2. Cumulative abnormal returns

CAR(60,2) CAR(60,1) CAR(1,1) CAR(1,0) CAR(0,1) CAR(1,60) CAR(2,60)

Variable CAR(1,1) SURPRISE SIZE BM RESENT VIX IVOL REVISION DISPERSION

1995–1999 2000–2003 2004–2006 2007–2009 2000–2009 1995–2009

5.2. Regression results

5.2.1. Cumulative abnormal returns around FFO announcements

(1) (2) (3) (4) (5)

5.2.2. Uncertainty, sentiment, and CARs

5.2.3. Robustness checks

(1) (2) (3) (4) (5)

CAR(60,2) CAR(2,60) CAR(60,2) CAR(2,60)

CAR(60,2) CAR(2,60) CAR(60,2) CAR(2,60)

6. Summary and conclusion

Research assessing the information content of earnings announcements is extensive. A key

You might also like