Investors Opinion Divergence and Post-Earnings Announcement Drift in REITs

Transcription

1 Investors Opinion Divergence and Post-Earnings Announcement Drift in REITs Gow-Cheng Huang Department of International Finance International College I-Shou University Kaohsiung City Taiwan, R.O.C Phone: ext Kartono Liano Department of Finance and Economics Mississippi State University Mississippi State, MS Phone: Ming-Shiun Pan Department of Finance and Supply Chain Management Shippensburg University Shippensburg, PA Phone: January 2014

2 Investors Opinion Divergence and Post-Earnings Announcement Drift in REITs Abstract In this study, we examine whether investors opinion divergence has explanatory power for post-earnings announcement drift in REITs. We measure investors opinion divergence using abnormal trading turnover around the earnings announcement. We find a U-shaped pattern of abnormal trading turnover across earnings surprise quintiles. Our results show that while earnings surprise, initial market reaction to the announcement, and abnormal trading turnover all have incremental predictive power for the REIT post-earnings announcement drift, abnormal trading turnover has the most influence. Our results also show that earnings surprise, initial market reaction, and abnormal trading turnover can predict future stock returns separately in different time periods. Keywords: REITs, Post-Earnings Announcement Drift, Opinion Divergence JEL Classification: G14 1

3 Investors Opinion Divergence and Post-Earnings Announcement Drift in REITs 1. Introduction Prior studies (e.g., Ball and Brown (1968) and Bernard and Thomas (1989)) find that stock prices tend to move in the same direction as earnings surprises up to one year following earnings announcements. The finance literature suggests that the main cause of the post-earnings announcement drift is investors underreaction to earnings information contained in the announcement (Bernard and Thomas (1990)). These studies show that the post-earnings announcement drift is positively related to earnings surprise. Similar to industrial firms, Price, Gatzlaff, and Sirmans (2012) also find that real estate investment trusts (REITs) exhibit post-earnings announcement drift. REITs are required to pay out 90% before tax income as dividends to shareholders in order to avoid corporate taxes. 1 The payout requirement forces REITs to rely more on external capital than industrial firms. Consequently, REITs receive more capital market monitoring than their industrial counterparts. REITs also have greater institutional ownership. Institutional investors tend to prefer investing in stocks with high liquidity and large market value. In addition, commercial real estate assets are traded in the private market. These unique characteristics suggest that REITs are more transparent and have a lesser degree of information uncertainty when compared to industrial companies. Consequently, according to the information uncertainty hypothesis, REITs should exhibit smaller post-announcement drift than industrial firms. However, Price et al. find that REITs generate larger post-earnings announcement drift than non-reits. Price et al. (2012) also show that REIT post-earnings announcement returns are positively related to earnings surprises. In addition to earnings surprise, the literature shows that investors opinion divergence can explain post-earnings announcement drift (e.g., Garfinkel and Sokobin (2006)). In this study, we extend Price et al. s study and attempt to relate investor opinion divergence to the post-earnings announcement drift in the context of REIT stocks. Our paper is related to work that examines the effects of investor opinion divergence on post-event announcement drift. Studies have shown that differences of 1 The REIT Modernization Act of 1999, which goes into effect in 2001, reduces the payout ratio from 95% to 90%. 2

4 opinion affect post-event stock returns. These studies analyze events such as IPOs and SEOs (e.g., Loughran and Marietta-Westberg (2005)), earnings announcements (e.g., Garfinkel and Sokobin (2006) and Lerman, Livnat, and Mendenhall (2008)), and acquisition announcements (e.g., Alexandridia, Antoniou, and Petmezas (2007)), among others. We measure investor opinion divergence in various ways. Previous work mostly uses earnings surprise to measure signals contained in an earnings announcement. Earnings surprise likely capture only earnings-related information, but not other information. It is reasonable to argue that the market reaction to earnings announcements is affected by both earnings and non-earnings information released at the time of the announcement. For example, Liu and Thomas (2000) find that earnings-related information can only partially explain the market reaction to earnings announcements. Jegadeesh and Livnat (2006) show that the drift is associated with revenue surprise even after controlling for earnings surprise. Our investor opinion divergence measures include abnormal trading turnover around earnings announcements and initial market reaction at the time of the announcement. These two measures are calculated using market data and hence can capture non-earnings related information. Lerman, Livnat, and Mendenhall (2008) show that the earnings announcement period abnormal trading volume can predict subsequent returns. Brandt et al. (2008) show that the announcement period abnormal return is related to post-earnings returns. We examine 8,603 announcements of earnings by REITs over the period Consistent with the literature, we find that REIT stock prices move in the direction of the earnings surprise over the subsequent 60 trading days following the announcement. When we compare 60-day post-earnings announcement abnormal returns across extreme earnings surprise quintiles, we find a significant difference of about 2%, which is equivalent to an annual return of 8%. We find that abnormal trading turnover around the announcement follows a U-shaped pattern across earnings surprise quintiles. Our results show a significant positive relation between abnormal trading turnover and the post-earnings announcement returns. While other measures of information uncertainty time series earning forecast 3

5 errors and abnormal returns around the announcement all have incremental explanatory power for the post-earnings announcement drift, abnormal trading turnover appears to have the greatest influence. Our results suggest that earning surprise, announcement period abnormal return, and abnormal trading turnover may reflect different information signals. We then test whether the impacts of these three variables on the post-earnings announcement returns change over time considering that the REIT sector underwent structural changes in the 1990s. Our subperiod analysis shows that the positive relation between abnormal trading turnover and subsequent returns prevails only in the data after The rest of the proposal is organized as follows. Section 2 reviews related studies and discusses the hypothesis. Section 3 defines our variables. Section 4 describes our sample. Section 5 presents our empirical results and the final section provides concluding remarks. 2. Related Studies and Hypothesis Development One prominent explanation for post-earnings announcement drift is the market s delayed reaction to the information contained in the earnings announcement (Bernard and Thomas (1990)). The delayed reaction hypothesis suggests that investors take time to process the earnings announcement news and hence the market reaction will extend over a long period of time. The delayed reaction hypothesis implies that the market does not fully incorporate the information conveyed in an earnings announcement. Hirshleifer, Lim, and Teoh (2011) suggest that the delayed reaction to earnings surprise is due to investors limited attention. They provide a model showing that investors fail to immediately incorporate earnings into valuation because of attentional constraints. The limited investor attention induces postearnings announcement drift. Other than the delayed reaction hypothesis, another possible explanation for the post-earnings announcement drift is differences of investors opinions in valuing a stock. Harris and Raviv (1993) and Chordia, Subramanyam, and Anshuman (2001) show that disagreement among investors over the relation between public announcements and announcing firms future performance is a major factor of trading. 4

6 Miller (1977) and Mayshar (1983) argue that in the presence of divergent opinion in valuing a stock, the holders of a stock tend to be more optimistic about its prospects. Earnings announcements will increase a stock s visibility and likely leads to an increase in the pool of potential buyers, but not potential sellers because of short-selling restrictions and other constraints. Consequently, the price of a stock will go up (down) upon a positive (negative) shock (Mendenhall (2004)). Gervais, Kaniel, and Mingelgrin (2001) argue that shocks in the trading activity of a stock through news affect the stock s visibility and its future price. They conjecture that the news-induced increase in the investor base of a stock should subsequently lead to an appreciation in the stock s price. Using trading volume to measure a stock s visibility, Gervais et al. find that trading volume over periods of a day or a week positively correlates with returns in the following month. This result is consistent with Harris and Raviv (1993) who show that trading volume and future returns are positively related due to disagreement among investors in interpreting the news. From these studies, we hypothesize that investors divergent opinion in deciphering the news on earnings announcements leads to post-earnings performance in REITs. Specifically, we hypothesize that: Hypothesis. Investors opinion divergence at the time of the earnings announcement is positively related to the REIT post-earnings announcement drift. In summary, the literature suggests a link between divergence of investors opinions and future stock returns. In this study, we use abnormal trading turnover around earnings announcements as a proxy for investors opinion divergence. We then assess whether divergence of opinions among investors can explain the REIT post-earnings announcement drift. 3. Variable Construction 3.1. Measure of Abnormal Trading Turnover Proxy for Opinion Divergence 5

7 Prior studies have used various proxies for investors opinion divergence, including dispersion in analysts earnings forecasts 2, trading volume (Bamber, Barron, and Stober (1999)), bid-ask spread (e.g., Handa, Schwartz, Tiwari (2003)), and stock return volatility (Shalen (1993)). In this study, we use abnormal trading turnover at the earnings announcement as a proxy for divergence of investors opinions or heterogeneous investor beliefs. Garfinkel and Sokobin (2006) and Lerman, Livnat, and Mendenhall (2008) find a positive relation between abnormal trading volume around earnings announcements and subsequent returns. Garfinkel (2009) shows that abnormal trading turnover is a better proxy for investors opinion divergence than bid-ask spread, stock return volatility, and analysts forecast dispersion. In practice, investors follow certain order submission strategies in order to obtain the best possible price in their transactions. Accordingly, trading orders are a good indicator of investors valuations on securities. While limit orders contain information about a trader s private valuation, market orders contain information about the trader s reservation price. Given that there is a strong link between order type, order price, and investors reservation price, Garfinkel (2009) proposes using limit order and market order data as a proxy for investors varying opinions on the valuations of securities. Specifically, Garfinkel uses the difference between the requested price of an order and the most recent trade price preceding the order to measure an investor s opinion divergence. Investors opinion divergence is then measured using the standard deviation of the price differences of all orders for a day. Garfinkel finds that among the commonly used proxies for investor opinion divergence, trading turnover correlates the most with his measure of investor opinion divergence. We measure abnormal trading turnover (ATO) using a REIT sector-adjusted trading turnover over a 2-day earnings announcement window [ 1, 0] as follows: 2 See Barron, Stanford, and Yu (2009) and the references therein for using analysts forecast variation to proxy for investors opinion divergence. However, Doukas, Kim, and Pantzalis (2006) find that the dispersion in analysts earnings forecasts is a poor proxy for investors opinion divergence. 6

8 0 Vol i, t Volt ATO 2, (1) t 1, Shsi t REIT Shst All REITs where Vol i,t is the ith REIT s trading volume on day t (t = 0 is the earnings announcement date) and Shs i,t is the ith REIT s shares outstanding on day t. The second term inside the bracket is the average trading turnover of an equally-weighted REIT portfolio on day t. All REITs with data available in CRSP are included in the equally-weighted REIT portfolio. We adjust for a REIT sector-wide trading turnover because the entire REIT sector might have more or less trading on some days due to news related to the REIT sector. ATO measures an abnormal trading turnover at the time of an earnings announcement and is used as a proxy for investors opinion divergence Measuring Abnormal Return and Drift Following the earnings drift literature, we calculate abnormal return (AR) as the daily return difference between a sample REIT i on day t and a benchmark portfolio: AR i, t = R i, t R benchmark, t, (2) where R i, t is the return for REIT i on day t and R benchmark, t is the mean return of a benchmark portfolio. We use two benchmark portfolios in the analysis. Since REITs are from a homogeneous sector, our first benchmark is a value-weighted REIT portfolio. This approach allows us to better measure a REIT s stock performance relative to its peers. Similar to the calculation of ATO, all REITs with return data available in CRSP are included in the value-weighted REIT portfolio. The second benchmark is an equallyweighted portfolio of all NYSE/AMEX/NASDAQ firms that fall in the same size decile as REIT i 45 days prior to the earnings announcement. 3 The post-earnings announcement drift is the cumulative abnormal return (CAR) over a 60-day post-earnings announcement window [1, 60] Measuring Earnings Surprise 3 We use the size portfolio as our second benchmark following Price et al. (2012). 7

9 As in Price et al. (2012), we use a seasonal random walk earnings surprise to measure earnings surprise. Earnings surprise is calculated as the difference between earnings per share excluding extraordinary items for the quarter prior to the earnings announcement and earnings per share lagged four quarters, 4 scaled by share price 45 days prior to the earnings announcement. As pointed out by Price et al., the random walk model has several advantages over analysts earnings forecasts. First, it avoids the biases that are commonly associated with analyst forecasts. Second, analysts usually cover larger firms and REITs tend to be small. Lastly, Bradshaw et al. (2012) find that simple random walk time-series models provide more accurate earnings estimates than analysts forecasts for smaller or younger firms. 4. Sample Data We collect all quarterly earnings announcements for REITs announced during the period We identify REITs from Compustat with SIC code 6798 and NAICS codes , , , , and Earnings announcement dates and quarterly earnings are retrieved from Compustat. Daily stock prices, stock returns, trading volume, and shares outstanding are retrieved from CRSP. The final sample contains 8,603 earnings announcements from 1980 to Table 1 reports earnings surprise (SURP), cumulative abnormal stock returns (CAR), and trading statistics for the entire sample as well as for positive and negative earnings surprise sub-samples. Of the 8,603 earnings announcements, there are 4,726 positive or zero earnings surprises and 3,877 negative surprises. The mean earnings surprise is The mean earnings surprise for the positive surprise sample is 0.026, which is significantly higher than that of the negative surprise sample ( 0.045). The market reacts favorably to the earnings announcement when the daily abnormal return is calculated as the return difference between a sample REIT and a value-weighted REIT portfolio consisting of all REITs with SIC code 6798, share code 18 (ordinary common shares, REITs), or share code 48 (shares of beneficial interest, REITs) in CRSP. The mean cumulative abnormal stock return over the two-day announcement window [ 1, 0] is 0.003%. REITs with positive surprises have significantly higher market 4 Earnings per share are diluted earnings per share and exclude extraordinary items. 8

10 reactions with a mean return of 0.256% compared to 0.306% for the negative-surprise REITs. The mean post-earnings announcement drift up to 60 days following the announcement (CAR[1, 60]) for the entire sample is 0.475%. However, the positive-surprise REITs yield a positive post-earnings announcement drift of 0.327%, which is significantly higher than the negative-surprise REITs 1.452%. When abnormal returns are calculated using size-adjusted returns, the cumulative abnormal returns appear to be greater than those calculated using REIT portfolio-adjusted returns. The mean two-day cumulative return and post-earnings announcement drift are 0.087% and 1.567%, respectively. Still, the positive-surprise REITs have significantly higher market reaction and post-announcement drift than the negative-surprise REITs. The average market capitalization (SIZE) on day 45 is $1.424 billion. The firm sizes do not differ significantly between the two earnings-surprise groups. REITs with positive surprises have significantly lower book-to-market (BM) ratios with an average of compared to for the negative-surprise REITs. Table 1 also shows the mean abnormal trading turnover (ATO) our proxy for investor opinion divergence. The mean abnormal trading turnover over the window [ 1, 0] is 0.034% for the positivesurprise REITs and is 0.045% for the negative-surprise REITs. While the negative-surprise REITs exhibit wider investor opinion divergence around the earnings announcement, the difference in abnormal trading turnover between the two sub-samples is not statistically significant. In Table 2, we report summary statistics of the same data variables across earnings surprise quintiles. Earnings surprises in quintile one (Q1) represent the largest negative earnings surprises, while earnings surprises in quintile five (Q5) represent the largest positive earnings surprises. The mean twoday cumulative abnormal return is negative in quintile 1 and positive in quintile 5, regardless of whether abnormal return is REIT portfolio-adjusted or size-adjusted. Moreover, the increases across earnings surprise quintiles are monotonic. The mean post-earnings announcement drift in quintile 5 is significantly higher than that in quintile 1 for both REIT portfolio-adjusted and size-adjusted returns. REITs with extreme earnings surprises tend to be smaller in size and have higher book-to-market ratio than the other 9

11 quartiles. Similarly, the abnormal trading turnover is also higher for REITs in the extreme earnings surprise quintiles. Specifically, the abnormal trading turnover decreases from Q1 to Q2, reaches the lowest level at Q3, and increases from Q4 to Q5, showing a U-shaped pattern. In other words, extreme earnings surprises generate more divergent opinions. Figure 1, Panel A shows the mean cumulative REIT portfolio-adjusted abnormal returns from day 1 to day 60 following the announcement. The drifts are monotonically increasing across quintiles except at the most extreme negative surprise quintile (Q1). The cumulative abnormal returns in quintile 5 drift down initially and then begin to drift up around day 14 following the announcement. Figure 1, Panel B plots the drifts that are calculated using size-adjusted returns. While the drifts also monotonically increase across quintiles except Q1, all cumulative abnormal returns appear to drift up around day 14 following the announcement. As a result, all cumulative abnormal returns on day 60 are positive. Although the post-earnings announcement drifts are different between the two measures of abnormal return, our analysis in what follows yields results that are qualitatively the same. Therefore, we only report those based on REIT portfolio-adjusted abnormal returns to save space. Table 3 presents the correlation coefficients among the variables. As shown, SURP is positively correlated with CAR[ 1, 0] and CAR[1, 60], suggesting that positive (negative) earnings surprises are associated with positive (negative) market reactions and also positive (negative) long-term abnormal returns following the announcement. ATO is strongly correlated with CAR[1, 60], indicating that REITs with wider investor opinion divergence tend to have larger post-earnings announcement drifts. ATO is also strongly correlated with firm size and book-to-market ratio. However, ATO is not correlated with CAR[ 1, 0] and SURP, suggesting that abnormal trading turnover around the announcement captures some non-earnings information not reflected in the market reaction as well as the earnings surprise. Overall, the univariate results support the notion that the greater the opinion divergence among investors, the larger the post-earnings announcement drift. 5. Regression Results 10

12 5.1. Investor Opinion Divergence and the Post-Earnings Announcement Drift To examine the relation between investor opinion divergence and the post-earnings announcement drift, we regress post-announcement cumulative abnormal return on abnormal trading turnover (ATO) and various control variables. The cross-sectional regression model is: CAR[1, 60] i = SURP i + 2 CAR[ 1, 0] i + 3 ATO i + 4 SIZE i + 5 BM i + i. (3) CAR[1, 60] is the cumulative abnormal return over a 60-day window [1, 60] following the announcement reported in Table 1. SURP is the earnings surprise reported in Table 1. We include SURP as a control variable because prior studies (e.g., Price et al. (2012)) show that earnings surprise is related to the drift. CAR[ 1, 0] is the cumulative abnormal return over a 2-day announcement window [ 1, 0] reported above. Liu and Thomas (2000) find that information other than earnings surprise can predict postearnings announcement returns. Jegadeesh and Livnat (2006) show that earnings announcements contain non-earnings related information such as sales and profit margins. Since the market reaction to an earnings announcement may capture both earnings and non-earnings related news, we therefore include CAR[ 1, 0] in the regression model. ATO is the abnormal trading turnover reported in Table 1. A significant coefficient for ATO would indicate that investors opinion divergence can help explain the REIT post-earnings announcement drift. SIZE is included to capture the information asymmetry. Financial analysts tend to provide less coverage for smaller firms and hence smaller firms are mispriced more frequently than larger firms. Book-to-market (BM) ratio is used to control for valuation effect, which suggests that high book-to-market firms tend to be more undervalued. In addition, the regression model includes quarter dummies to control for time-effects in the earnings announcement (not reported). The regression model is estimated using heteroscedasticity consistent standard errors adjusted for clustering at the firm level (Petersen (2009)). Table 4 reports the regression results of equation (3). As in Price et al. (2012), we first conduct the regression analysis without the inclusion of CAR[ 1, 0] and ATO. The results (model 1) show that the 11

13 coefficient on SURP is positive and statistically significant at the 10% level. Consistent with the literature, this finding suggests that the post-earnings announcement drift is caused by under-reactions to the information contained in earnings announcement. The book-to-market (BM) variable is significantly positive. REITs with high book-to-market ratios exhibit greater post-earnings announcement drifts. Next, we include the announcement period abnormal return (CAR[ 1, 0]) in the regression model (model 2). The coefficient on SURP remains positive with the significance level increases to 5%, while the coefficient on CAR[ 1, 0] is significantly negative. Larger initial market responses to the earnings announcement are associated with smaller drifts, suggesting that the initial market reaction might not be a good proxy for investor opinion divergence. The book-to-market variable continues to be significantly positive. Model 3 in Table 4 examines whether investors opinion divergence is a significant determinant of the post-earnings announcement drift. The results show that abnormal trading turnover (ATO) is positively related to the drift at the 1% level. This finding is consistent with our hypothesis outlined in Section 2. SURP, CAR[ 1, 0], and BM remain statistically significant Control for Risk Prior studies (e.g., Shalen (1993)) find that greater divergence of opinion is related to higher return volatility, suggesting that investor opinion divergence could simply be a risk proxy. To examine the incremental influence of our proxy of opinion divergence on the post-earnings announcement drift, we also include change in risk in our regressions. We measure risk using beta. The pre-earnings announcement beta is calculated using daily returns over the window [ 60, 1], while the post-earnings announcement beta is calculated over the window [+1, +60]. Beta is estimated by regressing daily stock returns on contemporaneous returns of the value-weighted REIT portfolio. The regression results with the inclusion of the change in beta are also provided in Table 4 (model 4). Similar to model 3, we observe a significant positive effect of ATO on post-earnings announcement returns. The coefficient estimate on BETA is positive and significant at the 5% level. 12

14 This result suggests that in addition to investor opinion divergence, risk is an important determinant of the post-earnings announcement drift Subperiod Analysis REITs experienced significant structural changes beginning in the early 1990s. The structural changes include allowing REITs to be internally managed and internally advised, removing the preferential tax treatment provided to limited partnerships, allowing the Umbrella Partnership REIT organization structure, and relaxing restriction in institutional ownership. 5 After these structural changes, REIT becomes more uncertain in valuation (Ling and Ryngaert (1997)) and larger in size and more liquid (Clayton and MacKinnon (2000)). REITs also experience an increase in institutional holdings (Chan, Leung, and Wang (1998)) and more insider ownership (Capozza and Seguin (2003b)). These structural changes likely will affect information uncertainty associated with REITs earnings announcements. To examine whether REITs structural changes affect the relation between investor opinion divergence and the drift, following Price et al. (2012), we conduct the analysis on three subperiods, , , and Table 5 reports that the means and medians of earnings surprises, earnings announcement returns, and abnormal trading turnovers for the three subperiods. The results show that the means (median) of the two-day market reaction are 0.099% (0.127%), 0.176% (0.148%), and 0.066% ( 0.075%) for , , and , respectively. These mean and median returns are statistically different. The mean (median) drifts are 0.079% ( 0.603%), 0.798% ( 0.758%), and 0.417% ( 0.515%) for , , and , respectively. However, the mean and median drifts for the three subperiods are not statistically different. The mean earnings surprises for the three subperiods are also not statistically different, while the period exhibits a higher median surprise. The average (median) ATO increases from 0.215% ( 0.082%) in to 5 See Capozza and Seguin (2003a) on changes in the REIT regulatory environment that lead to these structural changes. 13

15 0.028% ( 0.040%) in and to 0.068% ( 0.040%) in The results indicate that investor opinion divergence (abnormal trading turnover) becomes wider over time. Table 6 reports the regression results for the three subperiods. Interestingly, the results are very much different between the three periods. Drift is significantly negatively related to the two-day market reaction to the announcement in the period , but not in the other two periods. We find a significantly positive relation between earnings surprise and drift in the period , but no such relation was detected in the other two periods. Similarly, abnormal trading turnover is significantly positively associated with drift only in the period In summary, our subperiod analysis indicates that the impact of investor opinion divergence on the REIT post-earnings announcement drift is more pronounced in the most recent period. 6. Conclusions Similar to industrial firms, Price et al. (2012) find that REITs exhibit abnormal returns over the 60 days following the earnings announcement. While REITs have less information uncertainty compared to industrial firms, Price et al. find that REITs yield larger post-earnings announcement drift than industrial firms. They also find that earnings surprise can explain the post-announcement drift. In this study, we examine whether investor opinion divergence contributes significantly to the REIT postearnings announcement drift. Using a sample of 8,603 earnings announcements from 1980 to 2012, we find that positive (negative) earnings surprises generate positive (negative) post-earnings announcement drift in the 60 days following earnings announcements. We construct a REIT sector-adjusted trading turnover as a proxy for investor opinion divergence. We find that extreme earnings surprises are associated with more opinion divergence. Our results show that abnormal trading turnover is positively related to the REIT drift. Our results also show that the significant relation between abnormal trading turnover and the REIT drift holds only in the time period after 2000 a period with more information 14

16 uncertainty than earlier time periods. The structural changes in the 1990s appear to impact the level of information uncertainty in the REIT sector. 15

17 References Alexandridis, G., A. Antoniou, and D. Petmezas, 2007, Divergence of Opinion and Post-Acquisition Performance, Journal of Business Finance & Accounting 34, Ball, R. and P. Brown, 1968, An Empirical Evaluation of Accounting Income Numbers, Journal of Accounting Research 6, Bamber, L., O. Barron, and T. Stober, 1999, Differential Interpretations and Trading Volume, Journal of Financial and Quantitative Analysis 34, Barron, O., M. Stanford, and Y. Yu, 2009, Further Evidence on the Relation between Analysts Forecast Dispersion and Stock Returns, Contemporary Accounting Research 26, Bernard, V. L. and J. K. Thomas, 1989, Post-Earnings Announcement Drift: Delayed Price Response or Risk Premium?, Journal of Accounting Research 27, Supplement, Bernard, V. L. and J. K. Thomas, 1990, Evidence that Stock Prices Do Not Fully Reflect the Implications of Current Earnings for Future Returns, Journal of Accounting and Economics 13, Bradshaw, M. T., M.S. Drake, J. N. Myers, and L.A. Myers, 2012, A Re-Examination of Analysts Superiority over the Time-Series Forecasts of Annual Earnings, Review of Accounting Studies 17, Brandt, M. W., R. Kishore, P. Santa-Clara, M. Venkatachalam, 2008, Earnings Announcements are Full of Surprises, Duke University working paper. Capozza, D. and P. Seguin, 2003a, Special Issue: Real Estate Investment Trusts Foreword from the Guest Editors, Real Estate Economics 31, Capozza, D. and P. Seguin, 2003b, Inside Ownership, Risk Sharing and Tobin s q-ratios: Evidence from REITs, Real Estate Economics 31, Chan, S. H., W. K. Leung, and K. Wang, 1998, Institutional investment in REITs: Evidence and Implications, Journal of Real Estate Research 16, Chordia, R., A. Subramanyam, and V. R. Anshuman, 2001, Trading Activity and Expected Stock Returns, Journal of Finance 59,

18 Clayton, J. and G. MacKinnon, 2000, Measuring and Explaining Changes in REIT Liquidity: Moving Beyond the Bid/Ask Spread, Real Estate Economics 28, Doukas, J. A., C. Kim, and C. Pantzalis, 2006, Divergence of Opinion and Equity Returns, Journal of Financial and Quantitative Analysis 41, Garfinkel, J. A. and J. Sokobin, 2006, Volume, Opinion Divergence, and Returns: A Study of Post- Earnings Announcement Drift, Journal of Accounting Research 44, Garfinkel, J. A., 2009, Measuring Investors Opinion Divergence, Journal of Accounting Research 47, Gervais, S., R. Kaniel, and D. H. Mingelgrin, 2001, The High-Volume Return Premium, Journal of Finance 56, Handa, P., R. Schwartz, and A. Tiwari, 2003, Quote Setting and Price Formation in an Order Driven Market, Journal of Financial Markets 6, Harris, M. and A. Raviv, 1993, Differences of Opinion Make a Horse Race, Review of Financial Studies 6, Hirshleifer, D., S. S. Lim, and S. H Teoh, 2011, Limited Investor Attention and Stock Market Misreactions to Accounting Information, Review of Asset Pricing Studies 1, Jegadeesh, N. and J. Livnat, 2006, Post-Earnings-Announcement Drift: The Role of Revenue Surprises, Financial Analyst Journal 62, Lerman, A., J. Livnat, and R. R. Mendenhall, 2008, The High-Volume Return Premium and Post- Earnings Announcement Drift, New York University working paper. Ling, D. C. and M. D. Ryngaert, 1997, Valuation Uncertainty, Institutional Involvement, and the Underpricing of IPOs: The Case of REITs, Journal of Financial Economics 43, Liu, J. and J. Thomas, 2000, Stock Returns and Accounting Earnings, Journal of Accounting Research 38, Loughran, T. and J. Marietta-Westberg, 2005, Divergence of Opinion Surrounding Extreme Events, European Financial Management 11,

19 Mayshar, J., 1983, On Divergence of Opinion and Imperfections in Capital Markets, American Economic Review 73, Mendenhall, R., 2004, Arbitrage Risk and Post-Earnings-Announcement Drift, Journal of Business 77, Miller, E. M., 1977, Risk, Uncertainty, and Divergence of Opinion, Journal of Finance 32, Petersen, M. A., 2009, Estimating Standard Errors in Finance Panel Data Sets: Comparing Approaches, Review of Financial Studies 22, Price, S. M., D. H. Gatzlaff, and C. F. Sirmans, 2012, Information Uncertainty and the Post-Earnings- Announcement Drift Anomaly: Insights from REITs, Journal of Real Estate Finance and Economics 44, Shalen, C., 1993, Volume, Volatility, and the Dispersion of Beliefs, Review of Financial Studies 6,

20 CAR CAR Panel A: REIT Portfolio-Adjusted CARs 3.00% 2.00% 1.00% 0.00% -1.00% % -3.00% Q1 Q2 Q3 Q4 Q5 Panel B: Size-Adjusted CARs 4.00% 3.00% 2.00% 1.00% 0.00% % -2.00% Q1 Q2 Q3 Q4 Q5 Figure 1: Post-Earnings Announcement Drifts by Earnings Surprise Quintiles This figure shows post-earnings announcement drifts by earnings surprise quintiles. The post-earnings announcement drift is based on the cumulative abnormal return (CAR) from day 1 to day 60. Daily abnormal returns are REIT portfolio-adjusted (Panel A) and size-adjusted (Panel B). Earnings surprise is calculated as (EPS a EPS a 4 ) scaled by the stock price on day 45. Earnings per share exclude extraordinary items. 19

21 Table 1: Summary Statistics by Earnings Surprise Type This table reports means of the REIT earnings announcement sample by earnings surprise type. The statistics are based on 8,603 REIT earnings announcements over the period Earnings surprise is calculated as (EPS a EPS a 4 ) scaled by the stock price on day 45, where subscript a represents the announcement quarter. Earnings per share exclude extraordinary items. CAR1[t1, t2] and CAR2[t1, t2] are the REIT portfolio-adjusted and sizeadjusted cumulative abnormal returns from day t1 to day t2, respectively. Day 0 is the earnings announcement date. SIZE is the market value of equity, in billions, on day 45. BM is the book-to-market ratio between the book value of equity lagged two quarters and the market value of equity on day 45. ATO is the average REIT sector-adjusted trading turnover in a two-day window [ 1, 0]. CAR and ATO are in percentage. *** denotes significantly different from zero using a t-test at the 1% level. Positive or Zero Negative Full Sample Surprise Surprise Positive (N = 8,603) (N = 4,726) (N = 3,877) Negative SURP *** CAR1[ 1, 0] *** CAR1[1, 60] *** CAR2[ 1, 0] *** CAR2[1, 60] *** SIZE BM *** ATO

22 Table 2: Summary Statistics by Earnings Surprise Quintiles This table reports means of the REIT earnings announcement sample by earnings surprise quintiles. The statistics are based on 8,603 REIT earnings announcements over the period Earnings surprise (SURP) is calculated as (EPS a EPS a 4 ) scaled by the stock price on day 45, where subscript a represents the announcement quarter. Earnings per share exclude extraordinary items. CAR1[t1, t2] and CAR2[t1, t2] are the REIT portfolioadjusted and size-adjusted cumulative abnormal returns from day t1 to day t2, respectively. Day 0 is the earnings announcement date. SIZE is the market value of equity, in billions, on day 45. BM is the book-to-market ratio between the book value of equity lagged two quarters and the market value of equity on day 45. ATO is the average REIT sector-adjusted trading turnover in a two-day window [ 1, 0]. CAR and ATO are in percentage. Tests of equality in means are an F-test. *** denotes significantly different at the 1% level. Q1 Q2 Q3 Q4 Q5 Q5 Q1 Equal Means SURP *** 81.66*** CAR1[ 1, 0] *** 24.75*** CAR1[1, 60] *** 10.86*** CAR2[ 1, 0] *** 22.47*** CAR2[1, 60] *** 6.62*** SIZE *** BM *** *** ATO *** 21

23 Table 3: Correlation Coefficients among Variables This table reports the correlation coefficients among the variables. The sample includes 8,603 REIT earnings announcements over the period CAR[t1, t2] is the REIT portfolio-adjusted cumulative abnormal return from day t1 to day t2. Day 0 is the earnings announcement date. Earnings surprise (SURP) is calculated as (EPS a EPS a 4 ) scaled by the stock price on day 45, where subscript a represents the announcement quarter. Earnings per share exclude extraordinary items. SIZE is the market value of equity, in billions, on day 45. BM is the book-tomarket ratio between the book value of equity lagged two quarters and the market value of equity on day 45. ATO is the average REIT sector-adjusted trading turnover in a two-day window [ 1, 0]. *** and ** denote significantly different from zero at the 1% and 5% levels, respectively. CAR[ 1, 0] CAR[1, 60] SURP SIZE BM ATO CAR[ 1, 0] CAR[1, 60] 0.060*** SURP 0.091*** 0.028** SIZE ** BM *** 0.269*** 0.160*** ATO *** *** 0.041***

24 Table 4: Post-Earnings Announcement Drift and Abnormal Trading Turnover This table reports the regression coefficients from regressing post-earnings announcement drift (CAR[1, 60]) on abnormal trading turnover (ATO) and various control variables. Earnings surprise (SURP) is calculated as (EPS a EPS a 4 ) scaled by the stock price on day 45. Earnings per share exclude extraordinary items. CAR[ 1, 0] is the REIT portfolio-adjusted cumulative abnormal return from day 1 to day 0. CAR[1, 60] is the REIT portfolioadjusted cumulative abnormal return from day 1 to day 60. ATO is the average sector-adjusted trading turnover in a two-day window [ 1, 0]. ΔBETA is the difference in beta between the windows [ 60, 1] and [1, 60]. SIZE is the market value of equity, in billions, on day 45. BM is the book-to-market ratio between the book value of equity lagged two quarters and the market value of equity on day 45. CAR[1, 60], SURP, CAR[ 1, 0], ATO, and ΔBETA have been winsorized at the 0.5th and 99.5th percentiles. Inside the parentheses are t-statistics, based on heteroscedasticity consistent standard errors adjusted for clustering by firm and quarter. ***, **, and * denote significantly different from zero at the 1%, 5%, and 10% levels, respectively. Model 1 Model 2 Model 3 Model 4 Intercept 0.011*** 0.011*** 0.010*** 0.011*** ( 3.05) ( 3.06) ( 3.04) ( 3.05) SURP 0.081* 0.088** 0.088** 0.088** (1.83) (1.99) (1.98) (1.99) CAR[ 1, 0] 0.150** 0.154*** 0.153*** ( 2.54) ( 2.61) ( 2.60) ATO 0.776*** 0.766*** (2.66) (2.63) SIZE ( 0.65) ( 0.64) ( 1.02) ( 1.02) BM 0.010*** 0.010*** 0.010*** 0.010*** (3.09) (3.05) (2.97) (2.97) ΔBETA 0.009** (2.52) Adjusted R

25 Table 5: Earnings Announcement Cumulative Abnormal Returns by Decades This table reports means (medians) of earnings surprises (SURP), cumulative abnormal returns (CAR), and abnormal trading turnover (ATO) for the three subperiods, , , and Earnings surprise is calculated as (EPS a EPS a 4 ) scaled by the stock price on day 45, where subscript a represents the announcement quarter. Earnings per share exclude extraordinary items. CAR[-1, 0] is the REIT portfolio-adjusted cumulative abnormal return from day 1 to day 0, where day 0 is the earnings announcement date. CAR[1, 60] is the REIT portfolio-adjusted cumulative abnormal return from day 1 to day 60. ATO is the average sector-adjusted trading turnover in a two-day window [ 1, 0]. All data are in percentage. Tests of equality in means and in medians are an F-test and a x 2 -test, respectively. *** and ** denote significantly different at the 1% and 5% levels, respectively Test of Equality in (N=571) (N=2,050) (N=5,982) Means (Medians) SURP (0.000) (0.089) (0.000) (88.30)*** CAR[ 1, 0] ** (0.127) (0.148) ( 0.075) (18.70)*** CAR[1, 60] ( 0.603) ( 0.758) ( 0.515) (0.59) ATO *** ( 0.082) ( 0.040) ( 0.040) (29.60)*** 24

26 Table 6: Post-Earnings Announcement Drift and Abnormal Trading Turnover by Decades This table reports the regression coefficients from regressing post-earnings announcement drift (CAR[1, 60]) on abnormal trading turnover (ATO) and various control variables for the three sample periods, , , and CAR[1, 60] is the REIT portfolio-adjusted cumulative abnormal return from day 1 to day 60. Earnings surprise (SURP) is calculated as (EPS a EPS a 4 ) scaled by the stock price on day 45. Earnings per share exclude extraordinary items. CAR[ 1, 0] is the REIT portfolio-adjusted cumulative abnormal return from day 1 to day 0. ATO is the average sector-adjusted trading turnover in a two-day window [ 1, 0]. ΔBETA is the difference in beta between the windows [ 60, 1] and [1, 60]. SIZE is the market value of equity, in billions, on day 45. BM is the book-to-market ratio between the book value of equity lagged two quarters and the market value of equity on day 45. CAR[1, 60], SURP, CAR[ 1, 0], ATO, and ΔBETA have been winsorized at the 0.5th and 99.5th percentiles. Inside the parentheses are t-statistics, based on heteroscedasticity consistent standard errors adjusted for clustering by firm and quarter. ***, **, and * denote significantly different from zero at the 1%, 5%, and 10% levels, respectively Intercept ** 0.010** (1.01) ( 2.23) ( 2.17) SURP *** (1.58) (2.67) (1.20) CAR[ 1, 0] 0.424** ( 2.53) ( 1.54) ( 1.41) ATO *** (0.25) (0.75) (2.74) SIZE 0.007** ( 2.58) ( 1.15) ( 0.82) BM *** ( 0.04) (1.04) (2.68) ΔBETA * (1.36) (1.90) (1.53) Adjusted R