The Post Forecast Revision Drift and Underreaction to Industry-Wide and/or Firm-Specific Earnings

Transcription

1 The Post Forecast Revision Drift and Underreaction to Industry-Wide and/or Firm-Specific Earnings Kai Wai Hui Department of Accounting Hong Kong University of Science and Technology Clear Water Bay, Kowloon, Hong Kong Eric Yeung * J.M. Tull School of Accounting University of Georgia Athens, GA yeung@terry.uga.edu January 2011 Abstract We examine whether the post forecast revision drift is attributable to investors underreaction to industry-wide and/or firm-specific earnings news in analysts forecast revisions. We find a large drift associated with industry-wide earnings news but, on average, no drift associated with firm-specific earnings news. Consistent with the functional fixation hypothesis, we show empirically that the drift associated with industry-wide earnings news is driven by investors underreaction to the greater persistence of industry-wide earnings news. We also find that the magnitude of the drift associated with firm-specific earnings news varies with insider trading intensity, consistent with the information availability hypothesis. * We thank valuable comments from Dan Dhaliwal, Gilles Hilary, Bin Ke, Yong Yu, Guochang Zhang, and workshop participants at the Chinese University of Hong Kong, Hong Kong University of Science and Technology, INSEAD, McMaster University, and Nanyang Technology University of Singapore.

2 The Post Forecast Revision Drift and Underreaction to Industry-Wide and/or Firm-Specific Earnings 1. Introduction Individual financial analysts forecast revisions play a significant role in the dissemination of information about corporate earnings. Prior research has consistently documented a large drift following analysts earnings forecasts, indicating that investors are slow in processing earnings news in analysts forecasts. 1 Given the robust evidence of delayed reaction, high frequency of analysts forecasts, and the large magnitude of drift associated with them, it is important to understand why investors consistently underreact to the information in analysts earnings forecasts. While prior work so far has largely focused on detecting the existence of the post forecast revision drift, one unaddressed issue is identifying the economic fundamentals to which investors underreact. In other words, viewing earnings news as signals about fundamentals (e.g., industry and firm factors), it is not clear whether and why investors consistently underreact to certain types of signals about fundamentals. Examining this issue deepens our understanding of the nature of the drift and how investors process the information in earnings that reflects fundamentals. We test whether investors delayed reactions to analysts forecasts are mainly attributable to their underreaction to the news in analysts forecasts about industrywide earnings or firm-specific earnings (hereafter industry-wide earnings news and firm-specific earnings news ). 2 The breakdown of analysts earnings forecasts into industry-wide earnings news and firm-specific earnings news is motivated by two important reasons. First, prior work finds that analysts forecasts provide rich industry- 1 Prior research has consistently documented a large drift following analysts earnings forecasts (Givoly and Lakonishok [1980]; Stickel [1991]; Chan, Jegadeesh, and Lakonishok [1996]; Gleason and Lee [2003]; Zhang [2006]). Gleason and Lee [2003] suggest that the delayed market response to earnings news is one of the most perplexing anomalies to emerge from accounting-based capital market research over the past 20 years (p. 194). 2 Following the literature (Brown and Ball [1967]; Ayers and Freeman [1997]; Elgers, Porter, and Xu [2008]), we focus on industry-wide and firm-specific earnings and are less concerned about market-wide earnings. Industry-wide earnings represent the common component of the earnings of all firms in the same industry, while firm-specific earnings are the deviations of individual firms earnings from the industry average. Market-wide earnings, on the other hand, are less interesting in cross-sectional settings. In addition, marketwide forces ultimately manifest themselves at the industry and firm levels. We give specific mathematical definitions of industry-wide and firm-specific earnings in Section 3. 1

3 wide information that helps the stock prices to reflect industry commonality instead of firm idiosyncrasy (Piotroski and Roulstone [2004]; Chan and Hameed [2006]). Unlike insiders who have access to firm idiosyncratic information, analysts are outsiders and more likely to gain comparative advantages by focusing their efforts on obtaining and mapping industry- and sector-wide information into earnings (Piotroski and Roulstone [2004]). We therefore expect that the post forecast revision drift should be closely related to how investors process the information about industry-wide earnings news in analysts forecasts. Second, we expect significant differences in the efficiency of prices in impounding industry-wide earnings news and firm-specific earnings news. Prior studies (Ayers and Freeman [1997]; Elgers, Porter, and Xu [2008]) argue that because investors can obtain industry-wide information from various sources but have only limited access to firmspecific information, investors are more likely to underreact to firm-specific earnings news (information availability hypothesis). On the other hand, investors are also likely to underreact to industry-wide earnings news due to their tendency to be functionally fixated on the bottom line numbers (e.g., Bernard and Thomas [1990]; Hand [1990]; Sloan [1996]). More specifically, economic theory suggests that owing to competition industry-wide earnings are more persistent than firm-specific earnings. If investors do not understand the difference in the persistence of earnings components, prices may underreact to industry-wide earnings news (functional fixation hypothesis). We test alternative explanations for the drift with a large sample of earnings forecasts in analyst industry report. An analyst industry report is issued by an analyst who follows multiple firms in the same industry and produces forecasts for these firms simultaneously based on common industry fundamentals (See an example in Appendix). We utilize the forecasts in analyst industry report to minimize the measurement errors in deriving both industry-wide earnings and firm-specific earnings news (Bhojraj, Lee, and Oler [2003]; Elgers et al. [2008]). 3 Because these industry analysts endogenously 3 Elgers et al. [2008], for instance, do not find evidence supporting the information availability hypothesis proposed by Ayers and Freeman [1997] in the context of post earnings announcement drift and point out that measurement errors in industry-wide earnings constructed based on Standard Industry Classification (SIC) 2

4 choose to follow a set of firms subject to common industry shocks, aggregating earnings of these firms should yield a cleaner measure of industry-wide earnings than grouping earnings of all firms according to broad industry classifications. In addition, because these analysts self-select to use a top-down industry perspective to estimate firm performance, using measures of industry-wide earnings news based on their reports takes advantage of analysts efforts to map common industry-wide economic fundamentals into earnings estimates. We collect analyst industry reports from Investext for firms in the manufacturing sector for the period To ensure that these forecasts are disseminated widely to investors, we manually match these forecasts with the forecasts recorded by I/B/E/S through the 2006 Brokerage-Translation File provided by I/B/E/S. Our final sample covers 25,195 annual earnings forecast revisions for 4,098 firms in the five-year sample period. To ensure the generalizability of our results, we also test our predictions with a sample of 18,456 stand-alone forecasts in I/B/E/S that are not issued by industry analysts (i.e., who do not appear in analyst industry reports). Unlike the forecasts in industry reports that likely contain rich industry-wide information, these stand-alone forecasts are likely to contain rich firm-specific information. Obtaining similar results using both the industry report forecasts and stand-alone forecasts should enhance the confidence in the conclusion of our study. Following Gleason and Lee [2003], we define analyst forecast revision as the earnings forecast in analyst industry report minus the same analyst s prior earnings forecast, scaled by lag stock price. Industry-wide earnings news is measured as the average of forecast revisions for all firms in the same industry report. Firm-specific codes are likely to significantly reduce the power of their empirical tests. Bhojraj et al. [2003] show evidence that industry classification based on financial analysts (i.e., The Global Industry Classifications Standard (GICS)) is better in grouping more homogeneous firms for capital market research than SIC codes and North American Industry Classification System (NAICS) codes. Firms in analyst industry reports should be even more homogeneous than the GICS, because there are much fewer firms in each industry report. For instance, there are on average 13.2 firms in each industry report in our sample. In contrast, there are on average 26 firms in each six-digit GICS industry reported by Bhojraj et al. [2003]. We further discuss the superiority of using the forecasts in analyst industry report to derive industry-wide earnings and industry-wide earnings news in Section 3. 3

5 earnings news is defined as the difference between a forecast revision and the industrywide earnings news. Our results are striking. We find a large drift associated with industry-wide earnings news in analysts forecast revisions. On the other hand, we find no drift associated with firm-specific earnings news. These results indicate that, on average, the post forecast revision drift is driven by investors underreaction to industry-wide earnings news as opposed to their underreaction to firm-specific earnings news. We test the following predictions to ascertain whether the drift associated with the industry-wide earnings news is consistent with the functional fixation hypothesis. First, economic theory suggests that abnormal earnings above or below industry average are more transitory. This leads to our prediction that industry-wide earnings are more persistent than firm-specific earnings. Second, if investors are functionally fixated on earnings news and do not understand the differences in the persistence of earnings components, we predict that initial price reactions place similar weights on both industry-wide earnings news and firm-specific earnings news. Third, because the drift associated with industry-wide earnings news is rooted in investors failure to appreciate the greater persistence in industry-wide earnings, we expect this drift to be more pronounced when industry-wide earnings are more persistent. Consistent with economic theory, we find that industry-wide earnings are more persistent than firm-specific earnings. Investors, however, behave as if they do not understand the greater persistence in industry-wide earnings. We find that the initial price reactions to analysts forecast revisions place similar weights on both industrywide earnings news and firm-specific earnings news. During the post forecast revision periods, prices drift in the direction of industry-wide earnings news. Furthermore, the drift associated with industry-wide earnings news is more pronounced when industrywide earnings are more persistent. These results provide strong support for the functional fixation hypothesis. That is, the drift associated with industry-wide news is likely attributable to investors failure to understand the greater persistence in industrywide earnings. 4

6 While we do not observe a drift associated with firm-specific earnings news on average, we find evidence that the information availability hypothesis helps explain cross-sectional variation in the drift associated with firm-specific earnings news. Specifically, because firm-specific information is substantially more available to insiders, prices generally convey more firm-specific information when insiders trade more intensively (Piotroski and Roulstone [2004]). We thus predict a smaller drift associated with firm-specific earnings news among firms with high level of insider trading. Consistent with this prediction, we find that the magnitude of the drift associated with firm-specific earnings news is negatively associated with insider trading intensity. To ensure the generalizability of our results, we repeat our tests in the sample of stand-alone forecasts that are not issued by industry analysts and should therefore contain rich firm-specific information. Similar to the results obtained with the forecasts in the industry reports, we find that initial price reactions to stand-alone forecasts place similar weights on both industry-wide earnings news and firm-specific earnings news. During the post forecast revision periods, prices drift in the direction of industry-wide earnings news but not in the direction of firm-specific earnings news. To further corroborate the results that the drift is mainly driven by delayed reactions to industrywide news, we provide evidence that, when industry-wide information is a more important determinant of stock prices, the post forecast revision drift associated with the industry wide earnings news is larger. We also differentiate our underreaction story from risk-based explanations by testing whether the drift concentrates around major releases of earnings information. Consistent with our underreaction story, we find a greater than expected portion of the abnormal stock returns occur around future earnings announcements, suggesting that the drift is due to a delayed market response to current information about future earnings and this misperception is corrected when further information about future earnings is released. Our study contributes to the literature on the post analyst forecast revision drift by showing the information source of investors delayed reaction (Givoly and Lakonishok 5

7 [1980]; Stickel [1991]; Chan, Jegadeesh, and Lakonishok [1996]; Gleason and Lee [2003]; Zhang [2006]). We find that the drift is largely driven by investors underreaction to industry-wide earnings news and that there is little drift associated with firm-specific earnings news on average. We also find that both investors functional fixation on earnings numbers and lower availability of firm-specific information contribute to the post forecast revision drift. Our study also adds to the literature that shows investors likely misprice earnings components (e.g., Sloan [1996]; Xie [2001]; Hanlon [2005]; Richardson, Sloan, Soliman, and Tuna [2005]). Prior research shows variations in the persistence of earnings components and suggests that investors behave as if they do not understand differential persistence of earnings components. While the focus of prior research is on the managerial reporting discretion as the primary reason for the differences in the persistence of earnings components, we provide evidence that investors do not seem to fully understand the differential earnings persistence attributable to economic forces. Finally, our study contributes to the literature on the value of analysts forecasts in conveying industry-level earnings information (Piotroski and Roulstone [2004]; Chan and Hameed [2006]). While prior research shows that analysts forecasts appear to increase the relative amount of industry information in stock prices, it is not clear whether prices impound industry-level news in an efficient manner. We show evidence of prices underreacting to industry-wide earnings news in analysts forecasts. The remainder of our paper is organized as follows: Section 2 reviews the related literature and develops our main hypotheses; Section 3 outlines the sample selection and provides sample descriptive statistics; Section 4 reports results of hypothesis testing; Section 5 discusses additional empirical analyses; Section 6 concludes. 2. Related Literature and Hypotheses 2.1 RELATED PRIOR WORK Evidence that analysts forecasts are informative to investors dates back to the 1970s. (Griffin [1976]; Givoly and Lakonishok [1979]; [1980]; Elton, Gruber, and Gultekin 6

8 [1981]; and Imhoff and Lobo [1984]). Subsequent research shows that the short-window price reaction around forecast revisions is incomplete. Givoly and Lakonishok [1980] first report a post forecast revision drift. Stickel [1991] demonstrates that firms whose consensus forecast has been recently revised upward tend to earn higher abnormal returns over the next three to 12 months than firms whose consensus forecast has been recently revised downward. Chan et al. [1996] confirm Stickel s [1991] finding and show that it may be part of a general class of momentum strategies, in which the market reaction to recently released information is incomplete. Unlike these prior studies that mostly focus on changes in the consensus analyst forecast, Gleason and Lee [2003] examine price reaction to individual analysts forecast revisions and document crosssectional differences in the post forecast revision drift. They find that analyst visibility and analyst coverage mitigates the drift. In a similar vein, Zhang [2006] finds that drift is greater when information about the firm is scarcer. Prior research, however, has not viewed earnings news as a signal of firm fundamentals and examine whether news about firm-specific or industry-wide fundamentals causes investors underreaction. As a separate literature, Piotroski and Roulstone [2004] investigate the extent to which the trading activities of financial analysts, institutional investors, and insiders influence the relative amount of firm-specific, industry-level, and market level information impounded into stock prices, as measured by stock return synchronicity. They argue that financial analysts are outsiders who generally have less access to firmlevel, idiosyncratic information than insiders. As such, analysts are likely to focus their efforts on obtaining and mapping industry- and market-level information into prices. Viewing analysts and insiders as competitors of value-relevant information is consistent with the finding that number of analysts following is inversely related to the portion of the firm that is held by insiders (Moyer, Chatfield, and Sisneros [1989]). Prior evidence on financial analysts performance also suggests that analysts industry-level preferences exist. For example, Clement [1999] and Jacob, Lys, and Neale [1999] show that analyst accuracy improves with industry specialization, while Gilson, Healy, Noe 7

9 and Palepu [2001] show that the composition of analyst coverage changes after spin-offs and equity carve-outs. 4 Piotroski and Roulstone [2004] find that stock return synchronicity is positively associated with analyst forecasting activities, consistent with analysts increasing the amount of industry-level information in prices. In contrast, stock return synchronicity is inversely related to insider trades, consistent with these transactions conveying firmspecific information. Their results suggest that financial analysts forecasts contain significant industry-wide information because their relative information advantage is to aggregate and disseminate industry-wide information. Their results are further confirmed by Chan and Hameed [2006] with a sample of non-u.s. firms. It is unclear, however, from prior studies whether prices efficiently reflect industry-wide news in analysts forecasts. 2.2 HYPOTHESES Given that prior studies have documented the drift, our objective is to develop hypotheses on the existence of drifts associated with industry-wide earnings news and firm-specific earnings news and contrast them. We focus on two principal hypotheses related to the two types of earnings news: the information availability hypothesis and the functional fixation hypothesis. While the former views price discovery as a root cause of the drift, the latter is built on investors biases in processing earnings information Information Availability Hypothesis The information availability hypothesis is based on the argument that information about an industry is widely available to investors but firm-specific information is more difficult to obtain (Ayers and Freeman [1997]). Industry information, for example, can be extracted from various announcements of all firms in the same industry. In addition, the business press, industry associations, and government agencies provide data useful in 4 We further confirm that industry-wide earnings news is a significant component of total earnings news in analysts forecast revisions by regressing earnings news on indicator variables that represent individual industries. We find that industry indicator variables collectively explain 54% of the variation in total earnings news. 5 Ayers, Li, and Yeung [2010], for example, demonstrate with trading data that both price discovery and investors biased information processing play a role in explaining the post earnings announcement drift. 8

10 predicting industry performance well before such information is available for individual firms. Therefore, it is possible that investors trade on industry information without hesitation; while firm information may require intensive, time-consuming analysis before investors are willing to act upon. Consistent with this hypothesis, Ayers and Freeman [1997] find that stock prices anticipate industry-wide earnings earlier than firm-specific earnings and that the post earnings announcement drift is due primarily to firm-specific earnings. While Elgers et al. [2008] revisit these issues empirically and find little evidence consistent with Ayers and Freeman s [1997] predictions, they attribute the lack of empirical support to the measurement errors in industry-wide and firmspecific earnings and do not refute the intuition behind the information availability hypothesis. In our context, we are able to utilize the information in analyst industry reports to reduce measurement errors in industry-wide earnings, and empirically test the following prediction based on the information availability hypothesis: 6 H1: During the post forecast revision period, abnormal stock returns are more positively associated with the firm-specific earnings news than with the industry-wide earnings news Functional Fixation Hypothesis Alternatively, we expect that investors are likely to underreact to industry-wide earnings news because i) industry-wide earnings have greater earnings persistence and ii) investors may fail to distinguish between earnings components and do not understand the difference in the persistence of them. The functional fixation hypothesis suggests that investors are fixated on the total earnings number and thus place similar weights on signals with different implications for future performance. For example, when investors observe an earnings surprise of $1, they apply a valuation multiple of 15 and prices move up by $15 accordingly. However, if $0.8 is more persistent than the other $0.2, different valuation multiples should have been assigned and prices would have been different. The initial mispricing leads to a drift associated with the component with 6 All of our hypotheses are stated in alternative form. 9

11 greater earnings persistence (e.g., Sloan [1996]). 7 To test the functional fixation hypothesis, we make the following specific predictions regarding earnings persistence of industry-wide earnings and investors reaction to industry-wide earnings during both announcement and post-announcement periods. We first predict that the association between industry-wide component in current earnings and future earnings (i.e., persistence of industry-wide earnings) is greater than the association between firm-specific component in current earnings and future earnings (i.e., persistence of firm-specific earnings). Economic theory has long suggested that abnormal profits above or below the industry norm are likely to whittle away by competition (Mueller [1977]; [1986]; [1990]; Waring [1996]). Industry fundamentals that determine firm performance (e.g., consumer taste, production technology, and regulatory environment) are relatively long-lasting attributes. On the other hand, firms competitive positions within an industry are much more dynamic. While more successful firms tend to lose their competitive edges over time due to new entrants and learning of other firms, unsuccessful firms tend to improve their performance by taking corrective actions and imitating industry leaders. To the extent that accounting earnings are noisy measures of economic profits, industry-wide earnings should be more persistent than firm-specific earnings. 8 We therefore test the following hypothesis: H2a: The persistence of industry-wide earnings is greater than the persistence of firm-specific earnings. While industry-wide earnings are more persistent than firm-specific earnings, prices may behave as if investors do not fully understand the greater persistence of industrywide earnings compared to that of firm-specific earnings. Prior research suggests that average investors are functionally fixated on reported earnings (Bernard and Thomas [1990]; Hand [1990]; Sloan [1996]). If investors do not understand the greater 7 More general forms of functional fixation hypothesis date back to much earlier studies on accounting choices (e.g., Ball [1972]; Watts and Zimmerman [1986]). 8 While we focus on earnings persistence in this hypothesis as our purpose is to test difference in drifts, our theory predicts that firm-specific earnings are transitory and likely be mean-reverting and industry-wide earnings are not. Consistent with this prediction, we find a significantly negative first-order autoregressive coefficient for changes in firm-specific earnings (-0.26, t =-4.12) in a pooled regression. On the other hand, the first-order autoregressive coefficient for changes in industry-wide earnings is small and statistically insignificant (-0.06, t =-1.15). 10

12 persistence of industry-wide earnings, they will underreact to news about industry-wide earnings by placing similar weights on signals of both industry-wide and firm-specific earnings. We therefore test the prediction that prices behave as if investors fail to understand the greater persistence of industry-wide earnings: H2b: Initial price reactions to analyst forecast revisions place similar weights on the industry-wide earnings news and the firm-specific earnings news in analysts forecasts. If investors initially underreact to both industry-wide earnings news and firmspecific earnings news, prices should self-correct during the post forecast revision period and move toward the correct levels as information about industry fundamentals and firm competitive position within the industry gradually arrives. Because investors initially place similar weights on both components but the persistence of industry-wide earnings news is greater than that of firm-specific earnings news, we expect that the magnitude of self-correction associated with industry-wide earnings news should be greater than that associated with firm-specific earnings news. We therefore predict that the drift associated with industry-wide earnings news is greater than the drift associated with firm-specific earnings news, and test the following hypothesis: H2c: During the post forecast revision period, the association between abnormal stock returns and the industry-wide earnings news is more positive than the association between abnormal stock returns and the firm-specific earnings news Cross-Sectional Variations in Drifts Our last set of hypotheses concern the cross-sectional variation in the drift associated with firm-specific earnings news and in the drift associated with industrywide earnings news. We expect that both the information availability hypothesis and the functional fixation hypothesis explain the cross-sectional variation in respective drift. Specifically, the information availability hypothesis predicts that when firm-specific information is more available to investors, its valuation implications are more transparent, and therefore investors are likely to trade on it in a timely fashion (Ayers and Freeman [2007]). Because more thorough initial trading is likely to result in a lowered drift (Ayers, Li, and Yeung [2010]), the drift associated with firm-specific 11

13 earnings news should be less pronounced when firm-specific information in general is more available to investors. Prior studies suggest that firm-specific information is more likely to be available to investors when insiders trade intensively. Piotroski and Roulstone [2004] show that insider trading intensity increases the relative availability of firm-specific information in stock prices, consistent with insiders in general having the unique firm-specific information advantage compared to other market participants (Seyhun [1992]; [1998]; Ke, Huddart, and Petroni [2003]; Piotroski and Roulstone [2005]; Jagolinzer, Matsunaga, and Yeung [2007]; Jagolinzer [2009]). We therefore make the following two hypotheses: H3a: The initial price reaction to firm-specific earnings news is more pronounced when insider trading is more intensive. H3b: The post forecast revision drift associated with the firm-specific earnings news is less pronounced when insider trading is more intensive. We also make a prediction that links the cross-sectional variation in the drift associated with industry-wide earnings news to the persistence of industry-wide earnings. Because the functional fixation hypothesis maintains that investors do not fully understand the greater persistence in industry-wide earnings and thus underreact to news about industry-wide earnings, we expect a greater drift associated with industry-wide earnings news when industry-wide earnings are more persistent and test the following hypothesis: H4: The post forecast revision drift associated with the industry-wide earnings news is more pronounced when industry-wide earnings are more persistent. 3. Sample Selection and Research Method 3.1 SAMPLE OF FORECATS IN ANALYST INDUSTRY REPORT Our sample starts with individual analyst industry reports in Investext. Analysts self-select to follow certain industries due to their own background and expertise. In the first part of their industry reports, analysts typically discuss the expected overall performance of an industry according to underlying industry fundamentals such as 12

14 demand and supply conditions, cost of production and pricing powers, and potential structural shifts in the industry. In the second part, analysts make earnings forecasts and buy-sell recommendations for selected individual firms within that particular industry. Appendix gives an example of an analyst industry report. We use the forecasts in analyst industry report to derive industry-wide earnings and industry-wide earnings news for two reasons. First, aggregating the earnings of the firms appearing in the same analyst industry report should considerably reduce the measurement errors in industry-wide earnings arising from industry misclassification. Aggregating firms earnings at industry levels as defined by, for example, Standard Industry Classification (SIC) codes may introduce significant measurement errors in capital market studies because of the judgments about how to define industry membership, the level of aggregation in classifying firms into industries, and the classification of firms engaged in multiple lines of business (i.e., Bhojraj et al. [2003]; Elgers et al. [2008]). In contrast, financial analysts endogenously choose to follow a set of firms and release forecasts of their earnings simultaneously in the same industry report. It is likely that these firms share such significant economic commonalities that it is more effective for analyst to forecast their earnings simultaneously. As such, aggregating the earnings of the firms appearing in the same analyst industry report considerably reduces the measurement errors in the empirical proxy for industry-wide earnings. Second, we expect to obtain high-quality industry-wide earnings news by aggregating the earnings forecasts in an analyst industry report. Because these analysts i) self-select to specialize and focus their expertise in knowing an industry and ii) choose to use a topdown industry perspective to forecast firm performance, measures of industry-wide earnings news based on their reports should have sufficient power to capture analysts efforts to map market-, sector- and industry-wide economic conditions into earnings estimates. We therefore expect that aggregating the earnings forecasts based on an 13

15 industry perspective should produce high-quality measures of industry-wide earnings news. 9 Because it is time-consuming to manually collect and process analyst industry reports from Investext, we focus on the manufacturing industries and collect 58,796 industry reports for the period For each analyst report, we collect the name of the brokerage house, the names of the analyst(s), the date of the report, and the firms appear in the report. To ensure that the forecasts in analyst industry reports are widely disseminated, we manually match industry reports with I/B/E/S through the 2006 version of the Brokerage-Translation File maintained by I/B/E/S, which records the names of analysts and their affiliated brokerage houses during Because we rely on the 2006 version of the Brokerage-Translation file, our sample period is the five-year window centered on Following prior studies (e.g., Stickel [1991]; Gleason and Lee [2003]), we focus on analysts forecasts of upcoming annual forecasts. 11 After merging with I/B/E/S, we have 31,559 annual earnings forecasts for the manufacturing sector. We require a prior annual forecast issued by the same analyst to compute forecast revisions (Gleason and Lee [2003]), which reduces our sample to 28,029 forecasts. After merging with markets data from Center of Research of Security Prices (CRSP), we have 26,509 forecasts. We further delete the outliers in the top and bottom one-percentile of the continuous variables in the regression analysis, and our final sample consists of 25,195 individuals representing 4,098 firm-year observations. The forecasts in our sample come from 1,347 distinct analysts and 128 distinct brokerage houses. Table 1 provides descriptive information on our sample firms. Panel A shows that our sample includes about 800 firms per year. The average number of revisions per firm ranges from 5.44 (2004) to 6.48 (2007). The median number of revisions per firm is about 3 or 4 forecasts, comparable to mean values. While our data come from a period 9 We note that random measurement errors in industry-wide earnings news bias against finding statistically significant drift associated with industry-wide earnings news (i.e., against our main conclusion). 10 I/B/E/S stopped providing Brokerage-Translation File for academic research after A second reason for focusing on annual forecasts is that all analyst industry reports provide annual earnings forecasts and only a subset of them provide quarterly earnings forecasts. 14

16 subsequent to the sample period of Gleason and Lee [2003], these numbers are much lower than theirs simply because we focus only on forecasts in analyst industry reports. Panel B of Table 1 reports the number of revisions in our sample, classified by firm size decile. We use beginning-of-year market capitalization for all New York Stock Exchange (NYSE) firms as the threshold values to calculate firm size deciles. As expected, the majority of the forecast revisions in our sample are for larger firms, which attract greater analyst coverage. However, firm-level distribution shows that the top six NYSE size deciles are evenly represented in our sample. It is important to note that the distribution across size deciles is similar to that documented by Gleason and Lee [2003], which indicates that firms in analyst industry reports are not tilted towards either large or small firms. Following Gleason and Lee [2003], in subsequent tests, we compute sizeadjusted abnormal returns for each firm by subtracting the mean buy-and-hold return of an equal-weighted portfolio of firms in the same NYSE size decile over our holding period. If a stock is de-listed during the return accumulation period, then we assume the proceeds are reinvested to earn the average return of the matching size decile portfolio. Panel C reports the distribution of the firms in our sample across two-digit SIC industry groups. 12 In comparison with the distribution of all firms in the intersection of Compustat, CRSP and I/B/E/S, we find that our sample consists of higher proportions of electronic equipment makers (two-digit SIC = 36) and chemicals (two-digit SIC = 28). We do not observe significant differences in other industry groups. 3.2 SAMPLE OF STAND-ALONE FORECASTS Earnings forecasts collected from analyst industry report are likely to contain rich industry-wide information. To ensure that our results can be generalized to the forecasts that contain rich firm-specific information, we also test our hypotheses with a sample of forecasts that do not appear in analyst industry reports (i.e, stand-alone forecasts). These stand-alone forecasts are issued by analysts who do not forecast earnings from industry perspective and thus are likely to contain rich firm-specific information. We 12 Analyst industry reports in Investext are organized by two-digit SIC codes. We therefore use SIC codes to tabulate the industry distribution of our sample. 15

17 examine whether the drift after stand-along forecast is attributable to industry-wide earnings news or firm-specific earnings news. To ensure the validity of using our measure of industry-wide earnings news as the proxy for industry-wide earnings news in stand-along forecasts, we require that these stand-alone forecasts be the I/B/E/S forecasts for firms in our sample and issued within five days after analyst industry report date (i.e., [+1, +5] window). Focusing on standalone forecasts issued immediately after industry forecasts is more appropriate because the industry environment is unlike to change significant in a short window. 13 We obtain 18,456 usable stand-alone forecast revision observations. 3.3 COMPUTATION OF KEY VARIABLES Our tests call for a breakdown of earnings into industry-wide component and firmspecific component. Following prior studies (e.g., Brown and Ball [1967]; Ayers and Freeman [1997]; Elgers et al., [2008]), industry-wide earnings represent the common component of the earnings of firms in the same industry, while firm-specific earnings are the deviations of individual firms earnings from the industry average. 14 Let Ei,j,t denote the earnings of firm i in industry j for year t, and assume that there are N firms in industry j, the industry-wide earnings of industry j for year t are defined as: IndE = 1 N j, t / N i = E 1 i, j, t, (1) and firm-specific earnings of firm i in industry j for year t are defined as FirmE i, j, t Ei, j, t IndE j, t =. (2) In all of our empirical analyses, we scale actual earnings per share by stock prices at the beginning of the fiscal year. 15 In a similar fashion, we define industry-wide and firm-specific earnings news using analysts forecasts in each analyst industry report. Specifically, let REVi,j,t represent an 13 Our results from the stand-alone forecasts are robust when we use a [+1, +10] or [+1, +20] window. 14 We define industry based on analyst industry reports. Specifically, firms in an industry are the union of firms that ever appear in the analyst industry reports for the same industry. For example, if firms A, B, and C appear in analyst 1 s report for industry j and firms B, C, and D appear in analyst 2 s report for industry j, industry j includes firms A, B, C, and D. 15 Our industry-wide earnings measures contain the impact of market-wide forces on each industry. Since the main variable of interest is abnormal stock returns that exclude market-wide earnings information, including market-wide earnings does not bias our results. 16

18 analyst forecast revision for firm i in industry j for year t, the industry-wide earnings news is defined as: N i = REV 1 i, j, t IndREV j, t = 1 / N. (3) Following Gleason and Lee s [2003] finding that price reactions are highly associated with the innovation in analysts own forecasts, we define an analyst forecast revision (REV) as the difference between the annual forecast in the industry report and the prior forecast issued by the same analyst for the same fiscal period, scaled by stock price two days before the revision date. Once we have a measure for industry-wide earnings news, we define firm-specific earnings news in analysts forecasts as the difference between the total earnings news and the industry-wide earnings news: FirmREV i, j, t REVi, j, t IndREV j, t =. (4) Firm-specific earnings news captures the deviation of earnings news of individual firms from industry average earnings news. Table 2 provides descriptive statistics of key variable of interest and control variables in our multivariate regression analyses (discussed in the next section). The distributions of IndREV and FirmREV indicate that the magnitudes of industry-wide and firm-specific earnings news are remarkably similar, suggesting that investors received similar level of industry-wide and firm-specific earnings shocks. Our data also indicate that industry-wide earnings (IndE) and industry earnings persistence (PersInd) are generally higher than firm-specific earnings (FirmE) and firm earnings persistence (PersFirm), consistent with industry rents being generally positive and persistent. Finally, mean and median values of stock return synchronicity (Synch) indicate that although stock prices of firms in our sample are more synchronized with market- and industry-wide price movements than the firms in Piotroski and Roulstone s [2004] sample are, market- and industry-news collectively only explain a little more than 10% of the variability of firm stock returns. 4. Empirical Results 17

19 4.1 UNIVARIATE ANALYSIS Panel A of Table 3 presents average abnormal (size-adjusted) returns during the [-1, +1] three-day window centered on analyst industry report date for the 25-portfolios sorted independently on both industry-wide earnings news (IndREV) and firm-specific earnings news (FirmREV). We observe that, within each quintile of firm-specific earnings news, abnormal returns increase almost monotonically with the magnitude of industry-wide earnings news. The spreads between the highest and the lowest IndREV quintiles range from 0.39% to 1.16% and are all statistically significant (t 2.71). We also observe that, within each quintile of industry-wide earnings news, abnormal returns generally increase with the magnitude of firm-specific earnings news. The spreads between the highest and the lowest FirmREV quintiles range from 0.31% to 0.69% and are all statistically significant (t 2.02). These results indicate that stock prices response to both industry-wide earnings news and firm-specific earnings news in the direction of analysts forecast revisions. Panel B presents the average abnormal (size-adjusted) returns for the same 25- portfolios during the [+2, +60] three-month period after analyst industry report date. We find that, within each quintile of firm-specific earnings news, abnormal returns generally increase in the direction of industry-wide earnings news. The spreads between the highest and the lowest IndREV quintiles range from 1.83% to 3.26% and are all statistically significant (t 2.59). Consistent with the functional fixation hypothesis, these results indicate a significant drift associated with industry-wide earnings news in analysts forecast revisions. On the other hand, we do not observe any pattern in abnormal returns within each quintile of industry-wide earnings news. The spreads between the highest and the lowest FirmREV quintiles are small and statistically insignificant ( t 1.49). These results indicate little post forecast revision drift associated with firm-specific earnings news. Our evidence in Table 3 thus suggests that while prices initially react to both industry-wide earnings news and firm-specific earnings news in analysts forecast revisions, they continue to drift only in the direction of industry-wide earnings news but 18

20 not in the direction of firm-specific earnings news. One possible explanation is that firmspecific information is available to investors through alternative sources such as informed trading of insiders (i.e., H3a and H3b). We further examine these issues in section PERSISTENCE OF INDUSTRY-WIDE AND FIRM-SPECIFIC EARNINGS The functional fixation hypothesis posits that industry-wide earnings (IndE) are more persistent than firm-specific earnings (FirmE) and that investors underreact to the news in analysts forecasts about industry-wide earnings. In this section, we test whether industry-wide earnings (IndE) are more persistent than firm-specific earnings (FirmE) by running the following ordinary least square (OLS) regression: Ei,j,t+1 = α0 + α1indei,j,t + α2firmei,j,t + ε1 i,j,t+1, (5) where Ei,j,t+1 (future earnings) is the actual earnings per share for year t+1 of firm i that appears in analyst industry report j scaled by stock price at the beginning of year t+1; IndEi,j,t (industry-wide earnings) is the average of actual earnings per share (scaled by beginning price) for year t of the firms in the analyst industry report j in which firm i appears. FirmEi,j,t (firm-specific earnings) is the difference between actual earnings per share of firm i for year t (scaled by beginning stock price) and IndEi,j,t (see more detailed definitions in section 3.2). Equation (5) has been employed by prior research (e.g., Sloan [1996]) to examine the differential persistence among different earnings components. In our context, we expect that the industry-wide earnings have greater implications for future earnings than firmspecific earnings do (i.e., α1 > α2). Because equation (5) requires one-year-ahead earnings, our sample period (i.e., t) for this test is We present the regression results in Table 4. Model (1) shows the base-line regression that predicts one-year-ahead earnings with current earnings. We find that the estimated coefficient for current earnings (E) is (t = 29.80). 16 Model (2) shows the results of estimating equation (5). We find that the estimated coefficient for industry-wide earnings (IndE) is (t = 20.86), while the estimated coefficient for 16 We assess statistical significance based on firm and forecast-date clustered standard errors. 19

21 firm-specific earnings (FirmE) is (t = 5.45). F-test indicates that these two coefficients differ significantly at two-tailed one-percent level (F = 4.59). Consistent with H2a, we find that the industry-wide component in accounting earnings is more persistent than the firm-specific component in accounting earnings. This result sets the stage for testing the underreaction to industry-wide earnings news as the major source of delayed price reactions to analysts forecast revisions. 4.3 PRICE REACTIONS TO INDUSTRY-WIDE AND FIRM-SPECIFIC NEWS To test H2b and H2c, we estimate the following two OLS regressions: CAR i,j,t = b0 + b1indrevi,j,t + b2firmrevi,j,t + ΣControls + ε2 i,j,t, (6a) and CARpost i,j,t = β0 + β1indrevi,j,t + β2firmrevi,j,t + ΣControls + ε3 i,j,t, (6b) where CAR and CARpost are cumulative abnormal stock returns during the forecast revision window and various post forecast revision windows. To maintain high comparability with prior research, we follow the model used by Gleason and Lee [2003] and include three control variables that empirically explain abnormal stock returns. Log(MV) is defined as natural log of the market capitalization at the beginning of the year. BM is the book to market ratio at the beginning of the year. Momentum is the market adjusted returns over the prior six months. We expect b1 > 0 and b2 > 0, because prices initially react to both industry-wide earnings news and firm-specific earnings news. In addition, following the prediction of H2b, we expect that investors fail to detect the greater persistence of industry-wide earnings and predict b1 = b2. During the post forecast revision periods, however, investors continue to trade in the direction of industry-wide earnings news (β1 > 0), and H2c predict that the drift associated with industry-wide earnings news is greater than the drift associated with firm-specific earnings news (β1 > β2). We present the regression results in Table 5. Model (1) shows the results of price reactions to analysts forecasts during the three-day (i.e., [-1, +1]) window centered on forecast revision dates. As expected, we find positive coefficients for IndREV (0.122, t = 5.11) and for FirmREV (0.092, t = 6.59). Although the former is slightly greater than the 20

22 latter, the magnitudes are statistically indifferent (F = 1.30, p = 0.25). These results are consistent with H2a in that price reactions place similar weights on industry-wide earnings news and firm-specific earnings news in analysts forecast revisions. Results in Models (2) to (4) show evidence of delayed price reactions during various post forecast revision windows. We find positive coefficients for IndREV in regressions of cumulative abnormal returns during the 1 st to the 3 rd month window (i.e., [+2, +60]), the 4 th to the 6 th month window (i.e., [+61, +120]), and the 7 th to the 12 th month window (i.e., [+121, +240]). The estimated coefficients are large 0.305) ( in comparison with the estimated coefficient in Model (1) and highly significant (t 2.96). These results indicate strong evidence of post forecast revision drift associated with industry-wide earnings news. On the other hand, we find no evidence of drift associated with firm-specific earnings news. Specifically, estimated coefficients for FirmREV are small and statistically insignificant ( t 0.32). Consistent with H2c, we find that the estimated coefficients for IndREV are significantly greater than the estimated coefficients for FirmREV (F 7.51, p 0.01). 4.4 DRIFT AND INSIDER TRADING INTENSITY While we do not observe a significant drift associated with firm-specific earnings news on average, we expect that the information availability hypothesis should help explain the cross-sectional variation in this drift. According to this hypothesis, when firm-specific information is less available to investors, there should be a greater drift associated with firm-specific earnings news. Because insider trading increases the availability of firm-specific information in stock prices (Piotroski and Roulstone [2004]), we predict that, when firm-specific information in general is more available to investors, the initial price reactions are stronger to firm-specific earnings news (H3a) and drift associated with firm-specific earnings news is less pronounced (i.e., H3b). To test these predictions, we run the following multivariate regressions: CAR i,j,t = c0 + c1indrevi,j,t + c2firmrevi,j,t + c3indrevi,j,t Insider i,j,t + c4firmrevi,j,t Insider i,j,t + ΣControls + ε4 i,j,t, (7a) and 21

23 CARpost i,j,t = γ0 + γ1indrevi,j,t + γ2firmrevi,j,t + γ3indrevi,j,t Insider i,j,t + γ4indrevi,j,t Insider i,j,t + ΣControls + ε5 i,j,t, (7b) where Insideri,j,t is insider trading intensity of firm i in industry j for year t, defined as the absolute value of the difference between the sum of total shares purchased and total shares sold by insiders during the year, scaled by total trading volume during the year (Piotroski and Roulstone [2004]). For ease of interpreting the results, we convert Insider into decile ranks. If prices convey more firm-specific earnings information when insider trading intensity is high, we expect firm-specific earnings news is easier to interpret and trigger more thorough initial trading from investors (i.e., c4 > 0). In addition, we expect the more thorough initial reaction results in a lower drift associated with firm-specific earnings information (i.e., γ4 < 0). For a more balanced analysis, we also include the interaction of Insider and IndREV in both regressions. We present the regression results in Table 6. Model (1) shows the results of initial price reactions to analysts forecasts during the three-day (i.e., [-1, +1]) window centered on forecast revision dates. We find a significant positive coefficient for FirmREV Insider (0.011, t = 1.75), suggesting that investors react more strongly to firm-specific earnings news during the initial reaction window when insider trading is generally more intensive (H3a). On the other hand, we find negative coefficients for FirmREV Insider during the 1 st to the 3 rd month window (i.e., [+2, +60]) and the 4 th to the 6 th month window (i.e., [+61, +120]), indicating that the drift associated with firm-specific earnings is smaller when firm-specific information is more available to investors (H3b). It is also comforting to document the absence of similar patterns regarding the coefficients for IndREV Insider during both the initial three-day window and the post forecast revision windows. Overall, the results in Table 6 are consistent with the notion that the information availability hypothesis helps explain the cross-sectional variation in the drift associated with firm-specific earnings news. 4.5 DRIFT AND PERSISENCE OF INDUSTRY-WIDE EARNINGS 22

24 While our earlier results in Table 5 support H2b and H2c and indicate that the initial price reactions to industry-wide earnings news are insufficient and that there is a drift associated with it, we further link the delayed price reactions to investors underestimation of the persistence of industry-wide earnings (i.e., H4) and estimate the following two OLS regressions: CAR i,j,t = d0 + d1indrevi,j,t + d2firmrevi,j,t + d3indrevi,j,t PersIndE j,t + d4firmrevi,j,t PersFirmE j,t + ΣControls + ε6 i,j,t, (8a) and CARpost i,j,t = δ0 + δ1indrevi,j,t + δ2firmrevi,j,t + δ3indrevi,j,t PersIndE j,t + δ4firmrevi,j,t PersFirmE j,t + ΣControls + ε7 i,j,t, (8b) where PersIndEj,t is persistence of industry-wide earnings of industry j for year t, defined as the coefficient for current year (t) industry-wide earnings in a regression predicting one-year-ahead firm earnings (t+1), and PersFirmEj,t is persistence of firmspecific earnings of industry j for year t, defined as the coefficient for current year firmspecific earnings in a regression predicting one-year-ahead firm earnings. Persistence of industry-wide earnings and firm-specific earnings is obtained by estimating equation (5) for each industry-year. 17 For ease of interpreting the results, we convert PersIndE and PersFirmE into decile ranks. If investors understand the cross-sectional differences in the persistence of both industry-wide and firm-specific earnings, we expect d3 > 0 and d4 > 0. On the other hand, if investors underestimate the persistence of industry-wide earnings, we expect that the delayed reactions to industry-wide earnings news is more pronounced when industry earnings are more persistent (i.e., δ3 > 0). In contrast, if investors do not underestimate the persistence of firm-specific earnings, we expect little variation in the delayed reactions to firm-specific earnings news that is associated with the persistence of firmspecific earnings (i.e., δ4 = 0). 17 Because one-year-ahead earnings are required to estimate earnings persistence, we drop the observations in the last year of our sample. To run cross-sectional regression, we further require at least six observations in each industry for each year. 23

25 We present the regression results in Table 7. Model (1) shows the results of price reactions to analysts forecasts during the three-day (i.e., [-1, +1]) window centered on forecast revision dates. We find a significant coefficient for IndREV PersIndE (0.022, t = 2.04), suggesting that investors understand the cross-sectional differences in the persistence of industry-wide earnings. We also find a significant coefficient for FirmREV PersFirmE (0.016, t = 2.18), suggesting that investors also understand the crosssectional differences in the persistence of firm-specific earnings. For the post forecast revision periods, results in Models (2) to (4) show evidence that delayed price reactions to industry-wide earnings news are more pronounced when industry-wide earnings are more persistent. Specifically, we find positive coefficients for IndREV PersIndE during the 1 st to the 3 rd month window (i.e., [+2, +60]), the 4 th to the 6 th month window (i.e., [+61, +120]), and the 7 th to the 12 th month window (i.e., [+121, +240]). The estimated coefficients are all statistically significant (t 1.95). On the other hand, we find insignificant coefficients for FirmREV PersFirmE, suggesting that the persistence of firm-specific earnings does not affect the magnitude of post forecast period returns associated with firm-specific news. Overall, the results in Table 7 indicate that investors behave as if they only partially understand the persistence of industry-wide earnings and therefore underreact to industry-wide earnings news in analysts forecasts. This is consistent with H4 that the post forecast revision drift associated with the industry-wide earnings news is more pronounced when industry-wide earnings are more persistent. 4.6 STAND-ALONG FORECASTS ANALYSIS Our evidence so far indicates that prices do not fully react to industry-wide earnings news in the forecasts in analyst industry reports and that the drift after these forecasts is largely attributed to the underreaction to industry-wide earnings news. Because earnings forecasts in analyst industry report are likely to contain rich industry-wide information, it is not clear, however, whether our results would hold for forecasts that contain rich firm-specific information. We thus test whether prices underreact to the industry-wide earnings news in a sample of forecasts issued by other analysts who do 24

26 not simultaneously issue earnings forecasts of other firms in the same industry (i.e., stand-alone forecasts). Panel A of Table 8 presents average abnormal (size-adjusted) returns for the 25- portfolios sorted independently on industry-wide earnings news (IndREV) and firmspecific earnings news (FirmREV) during the [-1, +1] three-day window centered on forecast revision date. Except for the first FirmREV quintile, average abnormal returns increase almost monotonically with the magnitude of industry-wide earnings news within each FirmREV quintile. The spreads between the highest and the lowest IndREV quintiles range from 0.22% to 2.07%. We also observe that, within each quintile of industry-wide earnings news, abnormal returns generally increase with the magnitude of firm-specific earnings news. The spreads between the highest and the lowest FirmREV quintiles range from 0.70% to 1.85% and are all statistically significant (t 3.85). These results indicate that prices initially react to both industry-wide and firmspecific earnings news in the stand-alone forecasts. Panel B presents average abnormal (size-adjusted) returns for the same 25-portfolios during the [+2, +60] three-month period after analyst forecast revision date. We find that, within each FirmREV quintile, abnormal returns generally increase in the direction of industry-wide earnings news. The spreads between the highest and the lowest IndREV quintiles range from 1.28% to 3.19% and are all statistically significant (t 1.89). These results indicate that initial price reactions to the industry-wide news in analysts forecast revisions are insufficient. On the other hand, we do not observe increases in abnormal returns in the direction of firm-specific earnings news within each IndREV quintile. For IndREV quintiles 1, 2, and 4, the spreads are significant negative, indicating price reversals associated with firm-specific earnings news in these standalone forecasts. Overall, results in Table 8 suggest that initial price reactions to industry-wide earnings news in stand-alone forecasts are insufficient and that during the post forecast revision period, prices continue to drift in the direction of industry-wide earnings news but not in the direction of firm-specific earnings news. 25

27 We also run multivariate OLS regressions to test for the underreaction to industrywide earnings news in the sample of stand-along forecasts. Specifically, we estimate equations (6a) and (6b) and report the results in Table 9. Model (1) shows the results of price reactions to analysts forecasts during the three-day (i.e., [-1, +1]) window centered on forecast revision dates. As expected, we find positive coefficients for IndREV (0.207, t = 3.17) and for FirmREV (0.257, t = 6.59). The difference in their magnitudes is statistically insignificant (F = 0.56, p = 0.46). These results are consistent with H2b in that price reactions place similar weight on industry-wide earnings news and firmspecific earnings news in analysts forecast revisions. Results in Models (2) to (4) show evidence of delayed price reactions during various post forecast revision windows. We find positive coefficients for IndREV in regressions of cumulative abnormal returns during the 1 st to the 3 rd month window (i.e., [+2, +60]), the 4 th to the 6 th month window (i.e., [+61, +120]), and the 7 th to the 12 th month window (i.e., [+121, +240]). The estimated coefficients are large 0.323) ( in comparison with the estimated coefficient in Model (1) and statistically significant (t 2.34). These results indicate strong evidence of post forecast revision drift associated with industry-wide earnings news in stand-alone forecasts. On the other hand, we find no evidence of drift associated with firm-specific earnings news. The estimated coefficients for FirmREV are small and statistically insignificant ( t 0.99). Consistent with H2c, we find that the estimated coefficients for IndREV are significantly greater than the estimated coefficients for FirmREV (F 5.07, p 0.02). Consistent with H4, non-tabulated results further indicate that the post forecast revision drift associated with the industry-wide earnings news is more pronounced when industry-wide earnings are more persistent. 5. Supplemental Analyses 5.1 RETURNS AROUND SUBSEQUENT EARNINGS ANNOUNCEMENTS To discriminate between the market-inefficiency and risk-based explanations for the post forecast revision drift, we examine abnormal returns in three-day window centered on subsequent earnings announcements. If the post forecast revision drift is due to a 26

28 delayed market response to current information about future earnings and this misperception is corrected when further information about future earnings is released, then subsequent abnormal returns should cluster around the future announcements of earnings news such as earnings announcements. Conversely, if the price drift is due to omitted risk variables, we should not observe higher abnormal returns concentrated around the release of subsequent earnings news. Table 10 reports the results of these tests with our main sample. Specifically, we calculate the mean size-adjusted returns of a hedge portfolio over three-day windows centered on the next one- to four-quarterly earnings announcements. Hedge portfolios are formed by long the firms in the top decile industry-wide earnings news (IndREV) and short the firms in the bottom decile industry-wide earnings news. Our results show that the average hedge return is positive around all four subsequent earnings announcements, ranging from 0.236% to 0.755%. Assuming expected returns do not vary daily, we expect 0.063% to 0.328% around these earnings announcements. The difference between observed abnormal returns and expected returns are statistically significant (t 2.37) as well as economically significant (i.e., at least 2.3 times the expected returns). The higher proportion of abnormal returns observed around earnings announcements suggests that at least a portion of the delayed price response is due to misperception about future earnings, which is corrected around future earnings release dates. 5.2 DRIFT AND THE IMPORTANCE OF INDUSTRY-WIDE NEWS One of our main results is that the post forecast revision drift is primarily driven by delayed reaction to industry-wide news. One implication of this result is that the post forecast revision drift should be stronger when industry-wide news is a more important determinant of firm value. Following Piotroski and Roulstone [2004], we use return synchronicity as a measure of the importance of market- and industry-wide information for stock prices. 18 Higher return synchronicity implies that industry-wide news is a more 18 More specifically, return synchronicity (Synch) is estimated as log(r 2 /(1-R 2 )), where R 2 is the r-square from the regression RETi,t = a + b1marketi,t+b2marketi,t-1 + b3indreti,t + b4indreti,t-1 + εi,t. In this regression, RET is firm s weekly return of week t, MARKET is the value weighted market return of week t, and INDRET is the value weighted industry return of week t. We require at least 40 weekly observations in each firm-year regression and our sample in this test is reduced to 20,

29 important determinant of stock prices. We thus test the prediction that the drift associated with the industry-wide earnings news is more pronounced when return synchronicity is higher. To test this prediction empirically, we run regression models similar to equations (6a) and (6b) and add two interaction terms, IndREV Synch and FirmREV Synch, in the model. We expect a positive coefficient for IndREV Synch during the post revision periods. Table 11 present the multivariate regression results with our main sample. Model (1) shows the results of price reactions to forecast revisions during the three-day (i.e., [-1, +1]) window centered on forecast revision dates. We find that return synchronicity does not affect the strength of short-term market reactions. On the other hand, we find significantly positive coefficients for IndREV Synch during the [+2, +60] and the [+61, +120] return windows (t 2.35). These results are consistent with the prediction that drift associated with the industry-wide earnings news is more pronounced when return synchronicity is higher. In contrast, we find insignificant coefficients for FirmREV Synch during the post forecast revision periods. Overall, the results in Table 11 corroborate our main results that post forecast revision drift is highly associated with industry-wide earnings news. 6. Conclusions Prior research has consistently documented a large drift following analysts earnings forecast revisions, indicating that investors are slow in processing earnings news in analysts forecast revisions. We examine whether investors delayed reactions to analysts forecast revisions are mainly attributable to their underreaction to the news about industry-wide earnings or firm-specific earnings. To minimize the measurement errors in the proxies of industry-wide earnings and industry-wide earnings news, we test our predictions with a sample of earnings forecasts in analyst industry report, in which an analyst produces forecasts for multiple firms in the same industry simultaneously. We find that the post forecast revision drift is largely attributable to the drift associated with industry-wide earnings news. In addition, we find no significant drift 28

30 associated with firm-specific earnings news. Our evidence also indicates that both investors functional fixation on earnings numbers and lower availability of firm-specific information contribute to the post forecast revision drift. Specifically, while we find that industry-wide earnings are more persistent than firm-specific earnings, investors behave as if they do not fully understand the greater persistence in industry-wide earnings and therefore underreact to industry-wide earnings news, consistent with the functional fixation hypothesis. While we do not observe a drift associated with firm-specific earnings news on average, we find evidence that the magnitude of the drift associated with firm-specific earnings news is negatively associated with the availability of firmspecific information, consistent with the information availability hypothesis. To ensure the generalizability of our results, we replicate these tests in a sample of stand-alone forecasts that are not issued by industry-analysts. Our study contributes to the literature on the post analyst forecast revision drift by showing the information source of investors delayed reaction. Our results indicate that the drift is mainly driven by investors underreaction to industry-wide earnings news. Our study also adds to the literature about investors mispricing of earnings components. Prior research shows variations in the persistence of earnings components and suggests that investors behave as if they do not understand these differences. While the focus of prior study is on the managerial reporting discretion causing the differences in the persistence of earnings components, we show evidence that investors do not seem to fully understand the differential earnings persistence attributable to economic forces. Finally, our study contributes to the literature on the value of analysts forecasts in conveying industry-level earnings information. While prior research shows that analysts forecast revisions appear to increase stock return synchronicity, it is not clear whether prices impound industry-level news in an efficient manner. We show evidence of prices underreacting to industry-wide earnings news in analysts forecasts. 29

31 REFERENCES Ayers, B; and R. Freeman. Market Assessment of Industry and Firm Earnings Information. Journal of Accounting and Economics 24 (1997): Ayers, B; O. Li; and P.E. Yeung. Investor Trading and the Post Earnings Announcement Drift. The Accounting Review (2010): forthcoming. Ball, R.J. Changes in Accounting Techniques and Stock Prices. Journal of Accounting Research 10 (1972): Brown, P.; and R. Ball. Some Preliminary Findings on The Association Between The Earnings of A Firm, Its Industry, And The Economy. Journal of Accounting Research 5 (1967): Bernard, V.; and J. Thomas. Evidence That Stock Prices Do Not Fully Reflect The Implications of Current Earnings for Future Earnings. Journal of Accounting and Economics 13 (1990): Bhojraj, S.; C. Lee; and D. Oler. What s My Line? A Comparison of Industry Classification Schemes For Capital Market Research. Journal of Accounting Research 41 (2003): Chan, K.; and A. Hameed. Stock Price Synchronicity and Analyst Coverage in Emerging Markets. Journal of Financial Economics 80 (2006): Chan, L. K.; N. Jegadeesh; and J. Lakonishok. Momentum Strategies. Journal of Finance 51 (1996): Moyer, R., R. Chatfield and P. Sisneros, Security Analyst Monitoring Activity: Agency Costs and Information Demands. Journal of Financial and Quantitative Analysis (1989): Clement, M. Analyst forecast accuracy: Do Ability, Resources, And Portfolio Complexity Matter? Journal of Accounting and Economics 27 (1999): Elgers T.; S. Porter; and L. Xu, The Timing of Industry and Firm Earnings Information In Security Prices: A Re-Evaluation. Journal of Accounting and Economics 45 (2008): Elton, E. J.; M. J. Gruber; and M. Gultekin. Expectations and Share Prices. Management Science, 27 (1981): Gilson, S.; P. Healy; C. Noe; and K. Palepu. Analyst Specialization and Cnglomerate Stock Breakups. Journal of Accounting and Research 39 (2001): Givoly, D.; and J.Lakonishok. The Information Content of Financial Analysts Forecasts of Earnings: Some Evidence on Semi-strong Efficiency. Journal of Accounting and Economics 1 (1979): Givoly, D., and J. Lakonishok. Financial Analysts Forecast of Earnings: The Value to Investors. Journal of Banking and Finance, September (1980), Gleason, C. A.; and M.C. Lee. Analyst Forecast Revisions and Market Price Formation The Accounting Review 78 (2003): Griffin, P.. Competitive Information in The Stock Market: An Empirical Study Of Earnings, Dividends And Analysts Forecasts." Journal of Finance 31 (1976): Hand, J. R. M. A Test of the Extended Functional Fixation Hypothesis. The Accounting Review 65 (1990):

32 Hanlon, M. The Persistence and Pricing of Earnings, Accruals, and Cash Flows When Firms Have Large Book-Tax Differences. The Accounting Review 80 (2005): Imhoff, E.; and G. Lobo. Information Content Of Analysts Composite Forecast Revisions. Journal of Accounting Research 22 (1984): Jacob, J.; T. Lys; and K. Neale. Expertise in Forecasting Performance of Security Analysts. Journal of Accounting and Economics 28 (1999): Jagolinzer, A. SEC Rule 10b5-1 and Insiders Strategic Trade. Management Science 55 (2009): Jagolinzer, A.; S. Matsunaga; and P.E. Yeung. An Analysis of Insiders Use of Prepaid Variable Forward Contracts. Journal of Accounting Research 45 (2007): Ke, B., S. Huddart; and K. Petroni. What Insiders Know About Future Earnings And How They Use It: Evidence From Insider Trades. Journal of Accounting and Economics 35 (2003): Mueller, D.C. The Persistence of Profits Above the Norm. Economica 44 (1977): Mueller, D.C. Profits in the Long Run. Cambridge University Press. (1986). Mueller, D.C. The Persistence of Profits in the U.S. Dynamics of Company Profits: An International Comparison. Cambridge University Press (1990): Piotroski, J. and D. Roulstone, The Influence of Analysts, Institutional Investors, And Insiders On The Incorporation Of Market, Industry And Firm-Specific Information Into Stock Prices. The Accounting Review 79 (2004): Piotroski J.; and D. Roulstone. Do Insider Trades Reflect Both Conttraraian Belie and Susperior Knowledge about Future Cash Flow Realizations? Journal of Accounting and Economics 39 (2005): Seyhun, H. N.. "Why Does Aggregate Insider Trading Predict Future Stock Returns?" Quarterly Journal of Economics 107 (1992): Seyhun, H. N.. Investment Intelligence from Insider Trading, MIT Press, Cambridge, MA, (1998). Richardson, S. A.; R. G. Sloan; M. T. Soliman; and I. Tuna. Accrual Reliability, Earnings Persistence and Stock Prices. Journal of Accounting and Economics 39 (2005): Sloan, R. G. Do Stock Prices Fully Reflect Information in Accruals And Cash Flows About Future Earnings? The Accounting Review 71 (1996): Stickel, S. E. Common Stock Returns Surrounding Earnings Forecast Revisions: More Puzzling Evidence. The Accounting Review, 66 (1991): Waring, G.F. Industry Differences in the Persistence of Firm-Specific Returns. American Economic Review 86 (1996): Watts, R.L. and J.L. Zimmerman. Positive Accounting Theory. Englewood Cliffs, NJ: Prentice Hall (1986). Xie, H. The Mispricing of Abnormal Accruals. The Accounting Review 76 (2001): Zhang, X.F. Information uncertainty and stock returns. Journal of Finance 61(2006):

33 Appendix: An Example of Industry Analyst Report 32