Do Analysts Read the News?

Transcription

1 Do Analysts Read the News? Kaleab Mamo School of Accounting and Finance University of Waterloo September 2013 Abstract I examine whether and how corporate news releases influence equity analysts earnings forecasts. Combining large databases of U.S. corporate news releases and individual sell-side analysts earnings forecasts during 2000 to 2010, I investigate whether corporate news reports influence the direction and magnitude of equity analysts earnings forecast revision. I find that equity analysts forecast revision is significantly influenced by the tone of the news. Analysts revision directions and magnitude are both positively related to the tone of the news. Moreover, the relation between news and forecast revision is stronger for negative news and news that contains information regarding firm fundamentals. These results are obtained when earnings announcement periods are removed, highlighting the informational roles of corporate news in shaping analysts decisions. I am grateful to my supervisor Alan Huang for sharing the news database, and for his unfaltering support throughout this project. I thank Changling Chen, Ken Klassen, Kevin Markel, Pat O Brien, Hongping Tan, Christine Wiedman, fellow PhD students, and participants of the Brown Bag series for their helpful comments. 1

2 1 Introduction Financial analysts are primarily responsible for the analysis and interpretation of relevant information. 1 Prior research on sell-side equity analysts has examined different types of information events that shape analysts earnings forecast and stock recommendation decisions. This literature has mostly focused on whether earnings forecasts are influenced by specific types of information events such as earnings surprises (e.g., Abarbanell and Bernard, 1992), major stock price movements (e.g., Abarbanell, 1991; Conrad, Cornell, Landsman, and Rountree, 2006; Clement, Hales, and Xue, 2011), firm disclosure (e.g., Jennings, 1987; Rogers and Grant, 1997; Bowen, Davis, and Matsumoto, 2002), and other analysts forecasts (e.g., Welch, 2000; Clement et al., 2011). In this study, I use a comprehensive source of all corporate news releases to measure the general supply of firm-specific information and thus the relevant information, regardless of the information type or origin, to financial analysts. I then investigate whether and how analysts react to the arrivals of information. In addition to providing information regarding a wide variety of information events, corporate news releases may provide incremental qualitative information through the language of the report (e.g., Tetlock, Saar-Tsechansky, and Macskassy, 2008). The language (tone) of the news report may provide incremental information, for instance, because it reflects the analysis and interpretation of the financial press. This provides an opportunity to investigate whether equity analysts are influenced by other information intermediaries, which is the financial press in this case. Prior work on the capital market role of the press focuses on equity investors reaction to corporate news reports (e.g., Pritamani and Singal, 2001; Tetlock, 2007, 2010; Bushee, Core, Guay, and Hamm, 2010; Tetlock, 2011; Engelberg and Parsons, 2011; Huang, Tan, and Wermers, 2013). There is, however, no research examining how corporate news reports affect equity analysts. 2 This study aims to extend this literature by examining the effect of corporate news reports on earnings forecasts. Since analysts are arguably one of the most sophisticated capital market participants, this study can be considered as a strong test of the role of the financial press. The main hypothesis of this study is that public news stories, though their tone, influences 1 Presumably, the value-added activity of the analyst is, not surprisingly, analysis. (Bradshaw, 2011, page 5) 2 While there are studies (e.g., Frankel and Li, 2004) that examines the capital market impact of analysts and media separately, to the best of my knowledge, there is no research on how these two intermediaries affect each other. 2

3 analysts earnings expectations and result in forecasts revisions. To test this hypothesis, I use a quantitative measure of news tone, which is the proportion of negative (net of positive) words, and document its influence on forecast revision. 3,4 To this end, I combine a large database of U.S. corporate news releases (firm-specific public news stories) obtained from Factiva with individual analysts earnings forecasts from I/B/E/S. I match every forecast revision, issued within one year prior to the actual announcement date, with corporate news released between dates of the forecast revision and the previous forecast (i.e., during the revision period). I, then, hypothesize that the direction and magnitude of forecast revision are associated with the tone of the news stories released during the revision period. More specifically, equity analysts revise earnings forecasts upwards (downwards) if the tone of the news during the revision period is positive (negative), and the magnitude of such upward (downward) revision is higher if the tone is more positive (negative). If analysts reaction to public news is driven by the incremental information conveyed by the news, the association between news tone and forecast revisions should be stronger for more informative news. This argument is in line with Clement et al. s (2011) finding that analysts react more to stock prices and other analysts forecasts when these events are more likely to be informative. As a result, I hypothesize that the tone of public news influences forecast revisions more strongly when the news is more likely to be informative, i.e. when the news contains information that is directly relevant to earnings. Following Tetlock et al. (2008) and Huang et al. (2013) I use the frequency of the word root earn as a proxy for news informativeness. Lastly, consistent with the literature that finds an asymmetric reaction to positive and negative news (e.g., Chan, 2003; Tetlock et al., 2008; Veronesi, 1999; Brown and Ball, 1967), I hypothesize that negative news has a stronger impact on forecast revisions. The results generally support the hypotheses. First, I find that both the direction and magnitude of revision are positively associated with the tone of corporate news. The magnitude of forecast revision, defined as the change in EPS forecast scaled by stock price at the beginning of the fiscal 3 Prior research (e.g., Tetlock, 2007; Tetlock et al., 2008; Tetlock, 2010, 2011; Huang et al., 2013) use news tone to quantify the effect of news of news on investors. While most of these studies rely on the proportion of negative words, I follow Huang et al. (2013) who use the net proportion of negative words since this definition is more likely to reflect the actual tone of the news. 4 It is common in the analyst literature (e.g., Baginski and Hassell, 1990; Denis, Denis, and Sarin, 1994; Ely and Mande, 1996; Williams, 1996; Jennings, 1987; Conrad et al., 2006; Clement et al., 2011) to use forecast revision to document the influence of an information event on earnings forecasts. The logic is that if the arrival of a new information event changes analysts belief regarding their earnings forecasts, it will result in revision of the forecast. Thus, the information event should be associated with the forecast revision. 3

4 year, tends to be high (low) following revision periods dominated by positive (negative) news; moreover, the marginal effect of news tone on revision magnitude is approximately similar to that of cumulative abnormal return over the revision period. Upward (downward) revisions are more likely subsequent to revision periods dominated by positive (negative) news. Second, extending the regression model to include the interaction of news tone with the informativeness proxy ( earn frequency) confirms the hypothesis that analysts react more strongly to more informative news (news with fundamental content). This result also indicates that the influence of news on earnings forecasts is at least partly driven by the fundamental information available from corporate news reports. Third, dividing the sample into positive and negative news sub-samples and comparing estimated coefficients across these sub-samples shows that the association between news tone and forecast revision is stronger for negative fundamental news, but the association between general news tone and forecast revision (i.e., the main effect) is not statistically different for negative and positive news. This result differs from prior literature (Chan, 2003; Tetlock, 2007; Tetlock et al., 2008), which shows significant asymmetry between the effects negative and positive news. Finally, several additional tests show that the results hold generally regardless of cross-sectional variations in firm and analyst characteristics, and are robust to alternative measure of news tone. These additional tests also reveal potential factors that could determine the cross-sectional variation of the relationship between news tone and forecast revision. These factors include: analyst experience, analyst following, analyst forecast dispersion, trading volume (liquidity), and idiosyncratic volatility. These factors, with the exception of analyst experience, are usually considered as proxies of firm information environment. 5 Based on these observations, I conjecture that the association of corporate news and earnings forecasts is stronger for firms with poor information environment. This conjecture, however, is not thoroughly tested in the current study. This study, therefore, contributes to two branches of accounting and finance literature. First, it contributes to the equity analyst literature by providing evidence on how analysts react to the general supply of public information, regardless of the type and original source of the information. It, thus, generalizes the results in prior research, which has focused on specific types of information events such as earnings announcements, management disclosures, dividend changes, and major 5 Kim and Verrecchia (1994), Leuz and Verrecchia (2000), Landsman and Maydew (2002), Francis, Schipper, and Vincent (2005b), Francis, LaFond, Olsson, and Schipper (2005a), and Houston, Lev, and Tucker (2010) are examples of prior studies that treat one or more of these variables as proxies for firm information environment. 4

5 stock price changes. Second, it adds to the recent literature investigating the role of the financial press in capital markets. By documenting a significant impact of news tone on analysts earnings forecasts it provides strong evidence that corporate news releases contain incremental information and contribute to the firm information environment. Lastly, this study contributes to our understanding of the interaction between two different types of information intermediaries, particularly the influence of the business media on sell-side equity analysts. The remainder of this paper is organized as follows. Section 2 discusses prior research and develops the hypotheses. Section 3 describes the sample selection procedure and the research design. Section 4 presents the empirical results, and Section 4 concludes the paper. 2 Related Research and Hypotheses Development Equity analysts are important capital market information intermediaries. Their primary role is believed to be the analysis and interpretation of relevant information (e.g., Bradshaw, 2011), which leads to the question what information do analysts analyze?. 6 Since only the outputs of the analyst decision process are readily observable, prior research has focused on investigating what type information is incorporated in analysts outputs such as earnings forecasts, stock recommendations, and reports (e.g. Previts, Bricker, Robinson, and Young, 1994; Ely and Mande, 1996; Rogers and Grant, 1997; Bowen et al., 2002; Asquith et al., 2005; Conrad et al., 2006; Clement et al., 2011; Agarwal and Hess, 2012). Following this line of research, I examine whether and how public news stories influence equity analysts earnings forecasts. The main hypothesis is that public news stories provide incremental information to equity analysts, and therefore, influence their earnings forecasts. This research, in addition to furthering our understanding of the inputs into the analyst decision process, adds to evidence on the capital market role of the business media (e.g., Foster, 1979, 1987; Bushee et al., 2010; Tetlock, 2007; Tetlock et al., 2008; Tetlock, 2010, 2011). Prior research has documented analysts forecasts and recommendations consistently incorporate information contained in earnings surprises (e.g., Abarbanell and Bernard, 1992), stock price changes (e.g., Abarbanell, 1991; Conrad et al., 2006; Clement et al., 2011), dividend policy 6 While the current study and most prior research focus on the analysis or information interpretation role of analysts, analysts are also shown to have a significant information discovery role (e.g., Asquith, Mikhail, and Au, 2005; Chen, Cheng, and Lo, 2010). 5

6 changes (e.g., Denis et al., 1994; Ely and Mande, 1996), mandatory disclosures such as financial statements and annual reports (e.g., Rogers and Grant, 1997), voluntary/management disclosures (e.g., Jennings, 1987; Baginski and Hassell, 1990; Williams, 1996; Bowen et al., 2002), other analysts forecasts and recommendations (e.g., Welch, 2000; Bernhardt, Campello, and Kutsoati, 2006; Clement et al., 2011), and macroeconomic news (Agarwal and Hess, 2012). In this study, I extend this literature by investigating whether public news stories contain incremental information that can influence in equity analysts earnings forecasts. The public news stories considered in this study differ in two ways from the other types of information events typically examined by the analyst literature. First, public news can be considered as a better proxy for the general supply of information since public news stories report a wide variety of events and gather information from various sources including firm disclosures, stock markets, and security analysts (e.g., Tetlock et al., 2008). The press, as an information intermediary, may also engage in private information production through journalism (e.g., Bushee et al., 2010). The investigation of the relationship between corporate news and earnings forecasts, therefore, provides a more comprehensive look at analysts reaction to news regardless of its type or origin. Second, the language in the news stories may contain additional information, possibly, because it reflects opinion regarding the implication of the news (e.g., Tetlock et al., 2008). Regardless of whether analysts already have access to the information prior to its release by the press, their analyses may be influenced by the language of this information. 7 Public news may, thus, affect equity analysts by providing new information or alternative interpretations of existing information. The ability of the press to influence analysts forecasts and recommendations depends on whether it has an incremental contribution to firms information environment. A relatively recent literature investigates the role of the financial press in capital markets and presents evidence consistent with the general hypothesis that the financial press is an important capital market information intermediary and contribute to firm information environment. 8 Bushee et al. (2010) find information asymmetry is lower for firms with higher media coverage and they attribute these 7 Note that this study does not attempt to differentiate the information interpretation and discovery roles of the press. 8 The earliest attempts to investigate the capital market impact of the financial press include Foster (1979, 1987), which show significant market reaction to Abraham Briloff s articles published by Barron s. Foster s studies are, however, specific to articles written by a single author making it difficult to generalize the findings to the general press. 6

7 findings to the press influence on firms information environment through information discovery, packaging and dissemination of publicly available information. Pritamani and Singal (2001) and Chan (2003) find a significant conditional drift (long-run momentum) in monthly returns subsequent to extreme price movements accompanied by public news. Tetlock (2007), Tetlock (2010) and Tetlock (2011) show that public news reduces information asymmetry in stock markets, particularly for small and high growth firms as well as individual investors. 9 Engelberg and Parsons (2011) provides a strong evidence of a causal effect of media coverage on investors trading behavior through the examination of the impact of the same news on investors with different access to media. Some of the evidence in these studies (e.g., Tetlock, 2007) suggests that unsophisticated investors, small stocks, and/or growth stocks drive the relationship between news and stock prices. There are, however, studies documenting the implication of corporate news for firm fundamentals and sophisticated investors. Tetlock et al. (2008) shows that a high fraction of negative words in corporate news predicts low future earnings and that the predictive ability of the news is stronger when the story is related to fundamentals. A recent working paper (Huang et al., 2013) shows that corporate news releases significantly affect institutional investors trading activities. These findings indicate the effect of news is not entirely due to over-reaction by unsophisticated investors for small and high growth stocks. While these findings show that public news has significant influence on investors information set, we do not know if corporate news adds to analysts information, and thus influence their decisions such as earnings forecasts. On the one hand, the evidence discussed above indicates corporate news contains fundamental information that predicts future earnings, influences stock prices, and trading decisions by sophisticated institutional investors. Institutional investors are different from other (e.g., individual) investors in that they are believed to be more rational and better informed with access to better resources (e.g., they employ their own analysts known as buy-side analysts), and sell-side analyst reports. If corporate news contains fundamental information that results in significant reaction from sophisticated institutional investors, it is possible that the news contains information that is not incorporated in outstanding analyst forecasts. If so, analysts should incorporate this new information in to their subsequent earnings forecasts Tetlock (2011) also shows that investors remain to be influenced by stale news, such as reprints of the same story. 10 This argument is based on the assumption that market prices efficiently incorporate information in existing analyst forecasts; therefore, persistent market reactions and trading activities by institutional investors subsequent 7

8 On the other hand, equity analysts are believed to be experts with better access to information and better understanding of the key industry and firm success and risk factors. Equity analysts have better access to information, for instance, through their relationship with management and other firm insiders. Bradshaw (2011) shows that industry knowledge and management access are among the top five success factors for equity analysts. More specifically, he shows that the importance of management access has increased from 1998 to This is particularly important since it shows management access remains a key factor in analysts decision process even after the passage of Regulation Fair Disclosure (Reg FD) in While there is ample evidence (e.g., Heflin, Subramanyam, and Zhang, 2003; Gintschel and Markov, 2004; Francis, Nanda, and Wang, 2006; Agrawal, Chadha, and Chen, 2006; Mohanram and Sunder, 2006) suggesting Reg-FD has succeeded in minimizing selective disclosure, private meetings between management and analysts are still common in practice (Koch, Lefanowicz, and Robinson, 2013). A recent stream of working papers (e.g., Green, Jame, Markov, and Subasi, 2012a,b; Soltes, 2012) suggest that private meetings between management and analysts remain a significant information source to analysts. Such access to private information together with better understanding of the firms business may provide analysts with a superior information set compared to the financial press. Therefore, analysts are less likely to be influenced by corporate news as the information and the press s analysis may not add to their superior private information set. The above discussion leads to the main hypothesis that corporate news reports influence analysts earnings forecasts. Following the literature (e.g., Ely and Mande, 1996; Rogers and Grant, 1997; Conrad et al., 2006; Clement et al., 2011) which uses earnings forecast revision to measure the impact of information events on analysts forecasts, I test this hypothesis using earnings forecast revision. As in the prior literature, the effect of news is operationalized by constructing a quantitative measure of news tone (e.g., Tetlock et al., 2008; Huang et al., 2013). While many prior studies including Tetlock et al. (2008), Tetlock (2010), and Tetlock (2011) use the proportion of negative words as a measure of news tone (more specifically, negative tone), I follow Huang et al. (2013) to use the net proportion of negative words (net of positive words) to measure the overall tone of the news. If analysts incorporate information from corporate news reports in their earnings forecasts, we would be able to observe a significant association between the tone of the news released during to corporate news releases imply new information not included in existing earnings forecasts. 8

9 the revision period and their forecast revisions. 11 More specifically, public news may affect analysts forecast revision in two ways. First, it may influence the direction of revision with upward (downward) revisions being more likely subsequent to positive (negative) news stories. Second, it may influence the (signed) magnitude of the revision. That is, the more positive (negative) the tone of the news, the higher (lower) the forecast revision. These two sub-hypotheses are stated formally as follows. 12 Hypothesis 1. H1A. More positive (negative) news tone is associated with high (low) magnitude of forecast revision. H1B. More positive (negative) news tone is associated with a higher likelihood of upward (downward) forecast revision. While any relevant news is expected to influence analysts forecast revisions, some news stories may provide a direct and more informative signal regarding future earnings. For instance, Tetlock et al. (2008) find that the ability of news to predict earnings is higher when the news focuses on fundamentals. If analysts exhibit a reasonable level expertise, they should be able to identify and put higher weight on more informative news. Consistently, Clement et al. (2011) show that analysts respond more to stock price changes and other analysts forecasts when these signal are likely to be informative. Therefore, I hypothesize that more informative public news have greater impact on analysts earnings forecasts. While there may be many possible ways to measure the informativeness of public news, I follow Tetlock et al. (2008) and Huang et al. (2013) and use the frequency of the word root earn in the news story as a measure of news informativeness. This measure is designed to capture whether the news contains direct signal regarding earnings as well as the relative strength of the signal. It is, however, a noisy measure of news informativeness. To the extent that the measurement error is not systematic, the noise will only bias the coefficients towards zero. This leads to the prediction that the association between corporate news and analysts 11 The revision period, described in detail in section 3.2, is the period prior to the earnings forecast revision and after the previous forecast for the same firm by the same analyst. 12 Positive (negative) news could result in more upward (downward) or less downward (upward) revision depending on the expectation before the arrival of the news. Both result in a similar prediction for the relationship between the news and forecast revision variables. 9

10 forecast revisions is stronger when the news is more informative (i.e., news stories with relatively high frequency of the word root earn ). Chan (2003), Tetlock (2007), and Tetlock et al. (2008) provide evidence of asymmetric effects of positive and negative corporate news. Chan (2003) finds a strong downward stock price drift subsequent to bad news. Tetlock (2007) finds media pessimism predicts a price decline, whereas the media optimism has weak predictive ability. Tetlock et al. (2008) find negative words in corporate news reports predict firm earnings, but the predictive ability of positive word is much weaker. Asymmetric reaction to positive and negative news is also observed in other economic literature (e.g., Veronesi, 1999; Brown and Ball, 1967). Consistently, I predict that the impact of corporate news reports on analysts forecast revision is stronger when the news is negative. These two predictions are formally stated in the second hypothesis. Hypothesis 2. The association between news tone and forecast revisions, both in terms of direction and magnitude of revision, is stronger when: H2A. there is a higher frequency of earn in the news article. H2B. the news has a negative tone. 3 Sample Selection and Research Design 3.1 Sample Selection The sample is composed of data from four major datasets. Data for news tone, news informativeness, and other related public news variables are obtained from Factiva. 13 Individual analysts annual earnings forecasts and other related information are obtained from the I/B/E/S details file, and the data for other necessary information such as firm financials and stock market variables are obtained from the Compustat and CRSP databases. This news sample is constructed using corporate news stories for all U.S. firms from the Top Sources in the Factiva database between January 1, 2000 and December 31, It uses only 13 While the original source of the data is Factiva, I obtain the news dataset, including the measures of news tone and earn frequency from professor Alan Huang. The description of sampling procedure is, therefore, adapted from Huang et al. (2013), who provide more details on the construction of this database. 14 The Dow Jones News Archives include all news releases from Dow Jones newswire and The Wall Street Journal. The Top Sources of Factiva include: Dow Jones Newswire, Major News and Business Publications, Press Release Wires, Reuters Newswire and The Wall Street Journal. 10

11 the 2.2 million news stories that contain at least fifty words in total, and mention a company identity in the first twenty five words. To minimize false identification of news to a company, a news story is assigned to a particular company only if it mentions the company s identity three or more times. If a news story mentions two or more company identities, it is assigned to the firm that has the highest frequency of company-identity mentions in the article provided that the frequency of mentions of the second highest firm is less than 90% of that of the highest firm. Otherwise, the news is excluded from the sample. Moreover, observations that cannot be matched with a Compustat Gvkey are excluded from the sample. After these sampling screens, there are nearly 1.7 million news articles left. News released on a trading day before market closes at 4:00 PM is assigned to that trading day. If the news is released on a trading day after market closes at 4:00 PM or on a non-trading day (e.g., Weekends and Holidays), it is assigned to the next trading day. If there are two or more news releases on the same day for the same firm, they are combined into a single composite story. 15 This reduces the number of news observations to about 1.1 million for 15,540 firms. I apply additional screens and restrict the news sample to those firms that have annual analyst earnings forecasts in I/B/E/S. This reduces the sample to 495,718 news stories for 5180 distinct firms. I also remove news stories within the [-5,+5] days window around quarterly and annual earnings announcements resulting in 391,924 news stories for 5116 distinct firms since news in this window is likely to be dominated by earnings announcement news. The initial analyst sample contains over 2.89 million observations of annual earnings forecasts issued within two years prior to the actual announcement date for 8,206 unique firms for the fiscal years between 1999 and I exclude about 55 thousand observations for which: (i) I/B/E/S forecasts activation dates (I/B/E/S variable name, ACTDATS) precede the forecast announcement dates (I/B/E/S variable name, ANNDATS), (ii) forecast revisions dates (I/B/E/S variable name, REVDATS) precede either forecast announcement (ANNDATS) of activation dates (ACTDATS), (iii) forecast activation, announcement, or revisions are dated after the actual earnings announcement date % of the sample contains more than one news for a single firm in a single trading day. 16 These observations appear to be errors. The ANNDATS is the date the analyst announced (or, issued) the forecast, ACTDATS is the date I/B/E/S includes the forecast in its database, and REVDATS is the last date the analyst confirms (or, reiterates) the outstanding forecast as valid. Therefore, the natural order of these dates should be ANNDATS, ACTDATS, and REVDATS; moreover, they should all precede the actual earnings announcement date. 11

12 These screens result in a sample of over 2.36 million observations. For simplicity, I further restrict the sample to forecasts with one year horizon. 17 That is, only forecasts issued subsequent to the earnings announcement for the previous fiscal year and prior to the earnings announcement of the current fiscal year, are included in the sample. 18 This results in over 1.19 million earnings forecasts for 5,449 unique firms. I remove observations with fiscal year beginning price less than $1 since price is used as scaling variable, and earnings forecasts issued within the [-5, +5] days window around quarterly and annual earnings announcement dates since these forecasts are likely to be in response to earnings announcement news. These screens result in 532, 917 forecasts for 5202 unique firms. The analyst sample is matched with the news sample where each forecast revision is matched with news released between the previous forecast date and the revision date, which is referred to as the revision period. 19 If there are two or more news releases during the revision period, I take the average of the news variables as composite measures for the revision period. The combined sample, at this point, has 309,177 firm-year-analyst-forecast level observations for 4252 unique firms and 8524 unique analysts. I further require non-missing observations for scaling and control variables such as price at the beginning of the fiscal year (P rice), the net number of downward revisions by other analysts during the revision period (N etdownothers), the deviation of the analysts outstanding forecast from the consensus forecast during the revision period (Dist2Consensus), as well as forecast dispersions (Dispersion), cumulative abnormal returns (CAR), and idiosyncratic volatility measured both during the revision period and the previous fiscal year. These additional screens result in the final sample with 145,424 observations for 2978 unique firms and 6418 unique analysts. 3.2 Research Design I test the hypotheses by examining how analysts one-year-ahead earnings forecast revision is influenced by the tone of firm-specific public news released prior to the revision date and subsequent to 17 Corporate news may have an impact on earnings forecasts with two (or, more) years horizon; however, this study focuses on one year horizon for simplicity. 18 Forecasts for the one year horizon correspond to earnings forecasts for which the I/B/E/S Fiscal Year Indicator (FYI) equals See section 3.2 below for a detailed description of the research design. 12

13 the previous forecast date (the revision period). 20 A typical analyst issues multiple EPS forecasts for each firm and forecast end period, I consider only forecasts issued between the annual earnings announcement dates for the previous fiscal year (i.e., year t 1) and the current fiscal year (i.e., year t). The second forecast subsequent to the actual earnings announcement for the preceding fiscal year is, therefore, considered as the first forecast revision. Assuming a December 31 fiscal year end, the time line can be demonstrated as follows: 29-Jan Feb Mar May-2011 Revision Period 1 Revision Period 2 Earnings announcement for fiscal year 2010 First earnings forecast in 2011 for fiscal year 2011 Second earnings forecast in 2011 for fiscal year 2011 First earnings forecast revision Third earnings forecast in 2011 for fiscal year 2011 Second earnings forecast revision In this time line, the first revision date is March 15, At each forecast revision date (for each firm, fiscal year, and analyst), I measure the direction and magnitude of the revision. The magnitude of revision (RevM ag) is defined as the difference between the revised and the previous EPS forecasts scaled by stock price at the beginning of the fiscal year. That is, RevMag = RevisedESP P riorep S. P rice The direction of revision (RevDir), is captured by an indicator variable which takes the value one if the revision is downward, and zero otherwise. 1 if RevMag > 0 RevDir = 0 otherwise. The primary dependent variable is RevM ag, while RevDir is used to test the hypotheses regarding the direction of revision, for instance H1B, and as a robustness check for the other hypotheses. The two variables, RevM ag and RevDir, are designed to capture different dimensions of forecast revision. RevDir measures the direction of the revision and does not differentiate between small and 20 This research design is in line with Agarwal and Hess (2012), who investigate the influence of macroeconomic news on earnings forecasts. 13

14 large revisions of the same sign, a dimension captured by RevM ag. The moderate 53% correlation between the direction and magnitude of forecast revision supports this claim. The period between two consecutive forecast dates is referred to as the revision period. For example, in the time line above, the revision period associated with the first forecast revision on March 15, 2011 spans the days between, but not including, February 3, 2011 (the date of the first forecast) and March 15, 2011 (the date of the second forecast and the first revision). All firm-specific news stories released during the revision period are matched with the corresponding forecast revisions in order to examine the effect of news tone on the forecast revision. Thus, the independent and most of the control variables are measured during the revision period. News tone is measured using the average value of NegNet i, which is the net proportion of negative words (net of positive words), for all the news stories i released during the revision period. Huang et al. (2013) define NegNet i for each news event i as: NegNet i = No. of Negative W ord Occurrences No. of P ositive W ord Occurrences T otal Number of W ords in the News based on the word list of Loughran and McDonald (2011). The main independent variable in this study (NegNet) is, therefore, the equally weighted average of NegNet i for all the news stories i released during the revision period. I use the variable EarnF req to identify more informative news (or, news with fundamental content). Similar to NegNet, EarnF req is the equally weighted average of EarnF req i for all news stories i released during the revision period. Huang et al. (2013) define EarnF req i, for every news i, as the number of occurrences of the word root earn in news story i. While this research design allows examining the cumulative effect of corporate news released during the revision period on analysts earnings forecast revision, public news arrival is not the only important factor analysts observe during the revision period. For instance, there may be a significant information event (e.g., a major decline in stock price) during this revision period that leads to forecast revisions. Control variables measured during the revision period are, thus, included in the empirical analyses. It is also possible that the relationship between corporate news and analysts earnings forecasts are mediated by firm or analyst characteristics. To control for such cases, I include in the analyses control variables measured during the preceding fiscal year. 14

15 All control variables are discussed in detail in the empirical analyses part, and the definition and measurement period of all variables are presented in Appendix A. 3.3 Summary Statistics Table (2) presents the summary statistics for the data. In general, the sample is consistent with prior research. The mean (median) number of firms an analyst follows is approximately 12 (14). The mean (median) revision duration is 62 (56) days, and a vast majority (over 75%) of the analysts revise their forecasts within three months. 21 Roughly 60% of the forecast revision are downward and the mean (median) revision magnitude is -0.19% (-0.07%) expressed as percentage of the price at the beginning of the fiscal year. Thus, the average analyst follows 12 firms, revises down his/her earnings forecast every two months by about -0.19%. This is consistent with prior literature such as Richardson, Teoh, and Wysocki (2004), who document that analysts tend to walk down to beatable forecasts over the year. The average firm, for a given fiscal year, has total assets of $30B, average daily trading volume of 1.8M shares, over 23 analysts following it with an average price scaled forecast dispersion of 2%, and roughly 19 news stories. During an average revision period for a given analyst, a typical firm has roughly 8 analysts actively issuing EPS forecasts and at least two news stories. These descriptive statistics closely resemble the statistics from Agarwal and Hess (2012). Table (3) presents the Pearson correlation coefficients for the main regression variables. 22 This table provides the first evidence supporting my hypotheses. Both the direction and magnitude of the revision have significant correlations with N egn et with the expected signs. More strict tests of these hypotheses are presented in the next section. 4 Empirical Results 4.1 The Regression Model The hypotheses are generally tested by regressing a measure of forecast revision on news measures and control variables. The regression models take the form described in equation (1). 21 Note that the forecast revisions around earnings announcement dates are not included in this sample. 22 The correlation matrix for all the variables is available upon request. 15

16 RevV ar = α + β 1 NegNet + Φ CONT ROLS + ɛ (1) where RevV ar is either the price scaled magnitude of revision RevMag or the direction of revision RevDir, N egn et is the average net negative tone of news stories released during the corresponding revision period, and CON T ROLS is a vector of control variables, which include measure other analysts activities and stock markets measures during the revision period and the preceding fiscal year. As briefly mentioned in the research design section, public news releases are not likely to be the only information sources the analyst observes during the revision period. For instance, other information events such as quarterly earnings announcements may occur during the revision period. To minimize the confounding effect of quarterly earnings announcements, I exclude forecast revisions issued within 5 days of the announcement dates. Other information events documented by prior research to have a significant influence on analysts forecast revisions include dividend changes (Denis et al., 1994; Ely and Mande, 1996), management disclosures (Jennings, 1987; Baginski and Hassell, 1990; Williams, 1996), and major stock price changes (Conrad et al., 2006). While these information releases are public and likely to be incorporated in the news measures, it is not the aim of this study to show the effects of such information events. I, therefore, include cumulative abnormal stock returns and idiosyncratic volatilities, both measured during the revision period, as proxies for such information events. I use stock return variables as controls for other information events following (Conrad et al., 2006) who treat price changes as proxies for major public information events. 23,24 Prior research has also identified other factors that influence analysts forecast revisions. For instance, Welch (2000) shows that analysts tend to herd suggesting that an analysts forecast revision may be influenced by other analysts forecasts and/or the consensus forecast during the revision period. I control for the revision activity by other analysts using the variable RevM ag(others) 23 In reasonably rational markets, stock price changes occur due to of investors reaction to unexpected news (e.g., earnings surprises, dividend policy changes, or M&A announcements); therefore, these control variables are expected to be good proxies for major information arrival. It is, however, unlikely that they will completely remove the effects of these information events. 24 In untabulated results, the regression models were estimated including the number of days with negative return as additional control variable resulting in similar conclusion. This variable is not included in the main analyses since it is highly correlated with the CAR and idiosyncratic volatility measures. 16

17 and N etrevdown(others), which are respectively the average revision magnitude and the difference between the numbers of downward and upward revisions by other analysts during the revision period. Following Clement and Tse (2005), I also include Dist2Consensus, which is the distance between the analysts initial forecast and the consensus forecast during the revision period, as an additional control for herding. Equity analysts are important components of the firm information environment, and their decisions shape and are shaped by the information environment (Beyer, Cohen, Lys, and Walther, 2010). Moreover, the availability and quality of earnings signals is presumably a function of firm information environment. The firm s information environment is, thus, likely to play a role in whether and how corporate news stories influence analysts forecast revision decisions. For instance, if management voluntary disclosure provides sufficiently informative signal to produce EPS forecasts with acceptable accuracy level, analysts may not rely on other information sources such as pubic news stories. Therefore, I include proxies of firm information environment as control variables. The proxies, which include analyst following, analysts forecast dispersion, idiosyncratic volatility, and daily trading volume are selected based on prior research. Kim and Verrecchia (1994), Leuz and Verrecchia (2000), Landsman and Maydew (2002), Francis et al. (2005b), Francis et al. (2005a), and Houston et al. (2010) are among the prior work that treat one or more of these variables as proxies for firm information environment. I use the total number of distinct analysts who issued forecast during the revision period (F ollowanalysts(revp er)), the dispersion of these forecasts (Dispersion(RevP er)), the average daily trading volume (AvgV olume(revp er)), and idiosyncratic stock volatility (Idio.V olatility(revp er)) as proxies for firm information environment during the revision period. To control for proxies the long term component of firm information environment, I use analyst following (F ollowanalysts(lagged)), analyst forecast dispersions (Dispersion(lagged), average daily trading volume (AvgV olume(lagged)), idiosyncratic stock volatility (Idio.V olatility(lagged)), and size, all measured during the previous fiscal year. The ability of an analyst to efficiently gather and process information may also influence whether and how the analyst responds to public news releases. Such efficiency may depend on experience (considered as a proxy for expertise) and time constraint. I use the general and firm-specific analyst experience (measured in years) to control for experience, and the number of firms the analysts follows to control for time constraint. 17

18 The nature of the research design necessitates controlling for the following additional variables. Analysts do not revise their forecasts on regular basis, and, news does not arrive on a regular basis. Therefore, the revision period associated with each revision varies in length, and the number of news stories during the revision period varies from period to period. To make sure that the relationship between news and forecast revision is not an artifact of the difference in revision period or the number of news stories during the revision period, I control for the length of the revision period RevDur. 25 The variable RevDur is similar to the variable W alk used in Agarwal and Hess (2012) as a control for the walk down effect in analysts earnings forecasts reported in Richardson et al. (2004) Do Equity Analysts Incorporate Public News into their Earnings Forecast Revisions? To test the first part of hypothesis 1 (H1A), I estimate model (1) for the dependent variable RevM ag. The estimation results using the full sample are presented in Table (4). Columns (1) shows the results for regression of RevMag on NegNet controlling for revision period activity by other analysts following the firm, and stock market variables. Consistent with hypothesis 1, the significant negative coefficient on NegNet suggests that analysts react to the tone of public news releases during the revision period. More specifically, consistent with H1A, the magnitude of forecast revision tends to be lower (higher) following a revision period dominated by negative (positive) news. Column (2) repeats the regression with industry and year fixed effects, and provides similar results. Columns (3) - (7) expand the set of control variables to include firm and analyst characteristics measured during the previous fiscal year. The relationship between news tone and magnitude of revision remain unaffected. The coefficients on the control variables are also consistent with expectations based on prior research. Consistent with Agarwal and Hess (2012), the negative coefficient on RevDur in table captures the walk down effect documented in Richardson et al. (2004). However, the coefficient on RevDur is significant only in column (1) becoming its statistically insignificant when industry and 25 The number of news stories released during the revision period is not included as a control variable in the tabulated results due its high correlation with RevDur. The results remain similar when it is used instead of RevDur 26 An alternative, but highly correlated control variable for the Walk down effect is length of the time interval between the forecast date and the actual announcement date. While this variable is not included in the main regression analyses presented in this paper, the results remain qualitatively similar when it is added to the regression model. 18

19 year fixed effects are included in the regression. The mean revision magnitude by other analysts during the revision period RevM ag(others) loads consistently with positive coefficient suggesting forecasting activity by other analysts is one of the important factors in analysts forecast revision decision. This is consistent with prior work that documents herding in analysts earnings forecasts (e.g., Welch, 2000). This evidence is strengthened by the results for the difference between the consensus forecast during the revision period and the analyst s initial forecast (Dist2Consensus). The coefficient on Dist2Consensus is also consistently positive suggesting that analysts tend to converge to the consensus. Consistent with the hypothesis that stock price movements contain information relevant to analysts EPS forecasts (e.g., Conrad et al., 2006), the revision period cumulative abnormal return (CAR(RevP er)) significantly loads with positive coefficient. That is, the magnitude of forecast revision tends to be high (low) subsequent to periods of significantly high (low) CAR. The remaining coefficients indicate that forecast revision by a typical analyst is related to the firm specific analyst experience (F irmexper), analyst following (AnalystF ollow(lagged)), the number of other analysts actively issuing forecasts during the revision period (AnalystF ollow(revp er)), the dispersion of these forecasts (Dispersion(RevP er)), as well as idiosyncratic volatility (Idio.V olatility(revp er)) and trading volume (Avg.V olume(revp er)) during the revision period. The magnitude of forecast revision decreases with F irmexper, AnalystF ollow(revp er), Dispersion(RevP er), and Idio.V olatility(revp er) while increasing with Avg.V olume(revp er) and AnalystF ollow(lagged). These results suggest that forecast revision is sensitive to firm information environment as measured by analyst following, forecast dispersion, stock return volatility and trading volume. To sum up, the results in table (4) show analysts forecast revision are significantly influenced by the tone of public news releases in a manner consistent with hypothesis 1. Moreover, since I control for a host of measures designed to capture information in other analysts and stock market activity during the revision period, this evidence suggests public news stories contain incremental information. 19