Market Reactions to Analysts Initiations of Coverage, Post Reg FD Delbert Goff Terrill Keasler Appalachian State University

Transcription

1 Market Reactions to Analysts Initiations of Coverage, Post Reg FD Delbert Goff Terrill Keasler Appalachian State University Abstract We examine the market reaction to equity analysts initiations of coverage during the period after the implementation of Regulation FD. By taking a large sample of analysts initiations over a period of several years and controlling for concurrent news and events, we overcome potential weaknesses of some previous studies. Consistent with previous research, we find statistically significant positive abnormal returns associated with positive initial recommendations and statistically significant negative abnormal returns associated with negative recommendations. For neutral initial recommendations we find a statistically significant negative market reaction a result that conflicts with some previous studies. We conclude that analysts initiations of coverage in the post-reg FD era are viewed by the market as valuable and informative. 1

2 Market Reactions to Analysts Initiations of Coverage, Post Reg FD 1. Introduction The value of equity analysts forecasts and recommendations has been the subject of extensive academic research. While most of the prior work has focused on the accuracy of analysts earnings forecasts and on changes in analysts recommendations, there has been less focus on the impact of analysts initiations of coverage for a stock. 1 Previous studies have found a statistically significant market reaction to analysts initiations of coverage; however, there are conflicting results among previous studies. These conflicts are most evident in the direction and magnitude of the reported market reactions to initiations that are neutral or negative. Analysts initiations provide a unique opportunity to examine the perceived value of analyst coverage for a company. When an analyst changes an existing recommendation for a stock, the direction and magnitude of the market reaction is dependent on the new recommendation and on the level of the prior recommendation. For an initiation of coverage, the analyst does not have an existing history of recommendations for the stock; therefore, the rating may be perceived differently than would a ratings change. It is also possible that initiations bring more new information to the market than do recommendation changes. Peterson (1987) provides evidence supporting this supposition in his analysis of initial reviews by the Value Line Investment Survey, and Irvine (2003) finds that the market responds more positively to analysts initiations than to recommendation changes. Some prior research has had one or more potential problems such as examining a very short time period, not controlling for concurrent events and information releases, or testing during a period when non-public transfer of information to analysts was common. This study overcomes those potential problems and provides additional evidence on the market reaction to analysts initiations. To our knowledge this is the only study to provide an in-depth analysis of the market reaction to analysts initiations of coverage occurring after the implementation of Regulation Fair Disclosure ( Reg FD ). Reg FD became effective on October 23, 2000 and, according to the SEC, companies were routinely disclosing nonpublic information to analysts prior to the passage of Reg FD. 2 If it was common practice for companies to release new information to analysts, then a statistically significant market reaction to the release of analysts opinions would be expected, because the market would infer that the opinions contained information that was not already publicly available. By examining analysts initiations during the post-reg FD information environment, we can better determine if analysts opinions truly convey information or value to the market in and of themselves rather than potentially being viewed as a conduit for 1 The terms initiations of coverage, analysts initiations, and initiations are used interchangeably to represent the first time that an analyst firm initiates coverage on a company. That is, the analyst firm does not have pre-existing coverage on a given company, although other analyst firms may already have coverage on the company. 2 For additional information on this topic see Selective Disclosure and Insider Trading, Release Nos , , IC-24599; File No. S Available at 2

3 nonpublic information (as was possible prior to the implementation of Reg FD). Reg FD does not prohibit managers of public companies from having private meetings with analysts; however, it does prohibit the disclosure of material non-public information. The skill of the analyst may involve the use of information that is not material, and they may use it to build a mosaic of material information. This point was explained in a speech by SEC Director of Enforcement, Richard H. Walker. Mr. Walker stated: Based on everything you've heard today, you may be asking yourself what information of value can issuers still share with analysts in one-on-one situations. In addition to information that already has been made public, the answer is what often is referred to as "mosaic information." By definition, mosaic information is not "material," but that does not mean it's not valuable to an analyst. An issuer may convey to an analyst information that might seem inconsequential to the typical investor, but which a skilled analyst, knowledgeable about the issuer and the industry, may use to form a mosaic that reveals a material conclusion. An issuer may reveal this type of data even if, when added to the analyst's own fund of knowledge, it contributes to the analyst's ultimate judgments about the issuer. 3 There is some evidence that Reg FD may have impacted both the public and private information available from companies. It is expected that analysts access to private information should have declined as a result of Reg FD, and Findlay and Mathew (2006) provide evidence indicating that access to private information did in fact decline. In addition, there is some evidence that Reg FD may have reduced the amount of information disseminated by companies. 4 To the extent that this is the case, Reg FD may have made the services of skilled analysts even more important because they may be better able to generate mosaic information than the average investor. Consistent with previous studies, we find positive initiations (i.e., recommendations of buy or strong buy ) are associated with a positive and statistically significant market reaction. For neutral initiations (e.g., recommendations of hold ) and negative initiations (e.g., recommendations of sell or underperform ) we find a negative and statistically significant market reaction. These findings are consistent with the findings of Barber, Lehavy, McNichols, and Trueman (2001) and contrary to the findings of Irvine (2003). The remainder of this paper is organized as follows. The literature review is provided in Section II, and the data and methodology are discussed in Section III. Section IV presents our results, and a conclusion is provided in Section V. 3 Speech by SEC Staff: Regulation FD-An Enforcement Perspective. By Richard H. Walker, Director of Enforcement, U.S. Securities & Exchange Commission, November 1, Available at 4 A summary of surveys indicating that Reg FD reduced the amount of information disseminated by companies is provided in Special Study: Regulation Fair Disclosure Revisited, by SEC Commissioner Laura S. Unger, December Available at 3

4 2. Literature Review Considerable attention has been given to the market reaction and information content of analysts recommendations, and these studies generally find that the market reacts to analysts recommendations. 5 The majority of previous research has focused on recommendation changes, and less attention has been given to analysts initiations of coverage. Peterson (1987) provides an early study that examines the market s reaction to the initial reviews of securities included in the Value Line Investment Survey. Peterson hypothesizes that initial reviews are most likely to contain new information and are therefore most likely to have the largest impact on stock prices. Peterson s empirical results support this idea. Branson, Guffey, and Pagach (1998) find a statistically significant market reaction to initial buy recommendations by analysts. They also found that the abnormal returns for firms that had coverage by only a few analysts were greater than the abnormal returns for firms that had no preexisting analyst coverage and firms that had heavy analyst coverage. Irvine (2003) finds a statistically significant positive abnormal return for initial buy and strong buy recommendations; however, the magnitude of the abnormal returns is smaller than documented by Branson et al. Irvine also finds that initial recommendations of hold (sell) were associated with positive (negative) abnormal returns that were not significantly different from zero. A concern regarding the Branson et al. study and the Irvine study is that they only include data for a very short time period. Branson, et al. only include data from 1992, and Irvine uses data from two quarters during Looking at a short time period may lead to conclusions that may not apply to other time periods. Another concern regarding Branson, et al. is the small sample size of only 277 observations. An interesting aspect of the Irvine (2003) study is that some of the results conflict with the findings of Barber, Lehavy, McNichols, and Trueman (2001). Using a sample of 80,216 initial analysts recommendations issued during the time period, Barber et al. find a statistically significant negative abnormal return associated with hold, sell, and strong sell recommendations. Barber et al. do not control for concurrent events that may have a confounding effect on abnormal returns, and they do not conduct an in-depth analysis of the initiations, focusing most of their analysis on changes in analysts recommendations. The studies mentioned above were conducted prior to the implementation of Reg FD. In the period after the initiation of Reg FD, there is a question as to why there is a market reaction to analysts initiations if analysts do not have access to material information that is not already publicly available. Two possible explanations for a market reaction to initiations are the liquidity hypothesis and the information hypothesis. McNichols and O Brien (1997) develop a model suggesting that the information in initial analysts reports is better than information in continuation reports. Their model indicates that the decision to initiate coverage is in part a function of information held by the analysts, and their empirical analysis supports this model. McNichols and O Brien find that initial analysts forecasts are more accurate than forecasts for which analysts are continuing coverage. In the Reg-FD era, analysts should not have access to 5 For example see Davies and Canes (1978); Bjerring, Lakonishok, and Vermaelen (1983); Beneish (1991); Barber and Loeffler (1993); Palmon, Sun, and Tang (1994); Sok, Lin, and Slovin (1997); Barber, Lehavy, McNichols, and Trueman (2001); Gintschel and Markov (2004); Mikhail, Walther and Willis (2004); and Ferreira and Smith (2006). 4

5 material information that is not public; therefore, to the extent that analysts initiations contain new information, it should be a function of the skill of the analyst in taking information that is not considered material and creating an information mosaic that is material. This mosaic could be considered private, proprietary information created by the analyst. The liquidity hypothesis indicates that an increase in analyst coverage results in an increase in informed traders, thereby leading to lower bid-ask spreads and a decrease in the required rate of return on the stock. If there is a decrease in the required rate of return as a result of an analyst s initiation, there should be a positive stock price reaction to the coverage initiation. Brennan and Subrahmanyam (1995) and Roulstone (2003) examine the relationship between analysts coverage and stock liquidity characteristics and conclude that increased analyst coverage has a positive impact on liquidity. Irvine (2003) examines analysts initiations and provides empirical evidence to support the liquidity hypothesis. First, he finds that all initiations in his sample, except sell recommendations, have positive average abnormal returns. For the sell recommendations, the average abnormal returns are negative but not significantly different from zero. Irvine also finds a stronger market response for initiations than for coverage continuations, and he concludes that this finding is consistent with the liquidity hypothesis. For positive initiations ( buy or strong buy ), the liquidity and information hypotheses appear to be complementary in their predicted market response with both predicting a positive response. For neutral initial recommendations ( hold ) the liquidity hypothesis predicts that there should be positive market response due to the increase in liquidity. The information hypothesis is less clear in this area. McNichols and O Brien (1997) make a case for analysts being optimistic when initiating coverage, and analysts self-select and initiate coverage on firms for which they can offer favorable reports. So, where does that leave neutral initial recommendations? If the market believes that analysts are generally optimistic and if the best recommendation that an analyst can give is neutral, it would appear that the market may react negatively because a neutral recommendation would be considered as bad news by the market (i.e., no news is bad news ). This study extends the previous work on analysts initiations by examining the market reaction to initiations during the post Reg-FD period. We use a larger sample of analysts initiations than was used by Branson, et al. (1998) or Irvine (2003), and we control for concurrent events that may have a confounding effect on abnormal returns. In addition, we conduct a year-by-year analysis of abnormal returns to determine if the results vary over time. This study contributes to the existing literature by providing additional information in an area where conflicts exist among the findings of previous studies. 3A. Data 3. Data and Methodology To identify companies for inclusion in the study, we began by randomly selecting 90 stocks from the S&P 400 index, 70 stocks from the S&P 500 index, and 165 stocks from S&P 600 index. Companies were selected from each of these indices to ensure that a variety of small-cap, midcap, and large-cap stocks were included in the sample. More small-cap (S&P 600) and mid-cap (S&P 400) firms are included in the sample than large-cap (S&P 500) stocks in order to ensure 5

6 that the sample size of initiations from each index would be approximately equal. 6 For the 325 firms included in the sample, we obtained information on analysts initiations for the years from Yahoo! Finance. 7 In order for a company to be included in the sample there must be at least one analyst s initiation in the period that is not associated with an initial public offering and that is not concurrent with a recommendation release by another analyst (e.g., upgrade, downgrade, or coverage initiation). Initiations associated with IPOs are eliminated due to the findings of Bradley, Jordan, and Ritter (2003) that 96% of initiation recommendations at the end of the IPO quiet period are positive, 4% are neutral, and none are negative. Initiations concurrent with another analyst s recommendation release are eliminated in order to isolate the impact of initiations. 8 The broad variety of terms used by analysts in their recommendations presents a challenge when working with analyst recommendations. We use two methods of classifying the recommendations. The first method classifies the recommendations as being positive, negative, or neutral, and Table 1 provides a list of the various recommendations included in each classification. Using this three-way classification helps avoid the problem of trying to distinguish between the equivalent of a strong buy recommendation and buy recommendation and reduces the probability of having recommendations misclassified. A disadvantage of this three-way classification is that it is not consistent with previous research. To further analyze the positive recommendations and to be more consistent with other research, we also divide the positive recommendations into two groups that we label strong buy and buy. In order to determine (and control for) the effect of concurrent news or events on the market reactions for analysts initiations, we searched the EDGAR Online database for 8-K filings that occurred during a period beginning two days before through one day after analysts initiations of coverage. For observations that do not have a concurrent 8-K filling, we also searched LexisNexis for other concurrent news items related to the company such as company press releases and external news. Daily stock returns are obtained from the Center for Research in Security Prices (CRSP), and market capitalizations are from COMPUSTAT Research Insight. After completing the data screens listed above, there are 2,402 analysts initiations for the 325 firms in the sample. Eliminating observations for which CRSP or COMPUSTAT data are not available yields a final sample of 2,335 observations. A summary of the data is given in Table 2. Panels A and B indicate there is good dispersion across years. For the positive (neutral) initiations, there are 326 (236) observations with concurrent news or SEC filings and 987 (659) observations without concurrent news or SEC filings. Consistent with the findings in previous studies, Panel C 6 Since small-cap and mid-cap firms have a smaller analyst following and thereby fewer analysts initiations than do larger stocks, it is necessary to have more small-cap and mid-cap firms in the sample than large-cap firms. 7 Use of Yahoo! Finance as a source of analyst recommendations data is consistent with Ferreira and Smith (2006). Yahoo! Finance obtains the data for its Upgrades & Downgrades History from Briefing.com. 8 Bradley, Jordan, and Ritter (2003) present a confirmation hypothesis that market reaction will be stronger when multiple analysts initiate coverage simultaneously. Their empirical analysis supports this hypothesis. 6

7 indicates there are relatively few negative initiations in our sample. The small number of negative initiations limits the analysis that can be conducted on this type of initiation. Table 1 Classification of Analyst Recommendations The various recommendation schemes used by analyst firms are classified as being positive, neutral, or negative. The following table gives the various terms used for the firms in the sample. Recommendations Classified as Positive Recommendations Classified as Neutral Recommendations Classified as Negative Above Average Accumulate Mkt Underperform Add Average Reduce Aggressive Buy Equal-weight Sector Underperform Attractive Hold Sell Buy In-Line Underperform Buy Aggressive Maintain Position Underweight Long Term Accumulate LT Attractive Mkt Outperform Net Positive NT Accum/LT Accum NT Accum/LT Buy NT Buy/LT Buy NT Buy/LT Strong Buy NT Neutral/LT Accum NT Strong Buy/LT Strong Buy NT/LT Buy Outperform Outperform/Buy Overweight Positive Recommend List Sector Outperform ST Mkt Perform/LT Buy Strong Buy Market Perform Mkt Neutral Mkt Perform Neutral NT Market Perform NT/LT Mkt Performer Perform Sector Perform ST Mkt Perform/LT Mkt Perform The information in Table 2 indicates that during the time period, there may have been a shift in the type of initial recommendations made by analysts. In 2001, 71 percent of all initiations in the sample were positive. In 2005, only 45 percent of the initiations were positive. This change could be a reaction to public scrutiny regarding analysts activities during the dot.com boom of the 1990 s and also may indicate some influence from Reg FD; however, it is not possible to draw definite conclusions from this data. The information in Table 2 also indicates there has been a decline in the number of analysts initiations per year during the time period. This reduction may be due to consolidations within the brokerage industry and to cutbacks by Wall Street research firms. See Der Hovanesian and Borrus (2005) for additional information on the cutbacks by analyst firms. 7

8 Table 2 Summary of Data Distribution of analysts positive, neutral, and negative initiations of coverage divided by index and by year and segmented by those observations with versus without concurrent SEC filings or news. With indicates that an analyst initiation was made concurrent with an SEC filing or news event and W/O means that an analyst initiation was made without a concurrent SEC filing or news event. Panel A: Positive Initiations Segmented by SEC Filings/News S&P 400 S&P 500 S&P 600 All Indices With W/O Total With W/O Total With W/O Total With W/O Total Total Panel B: Neutral Initiations Segmented by SEC Filings/News S&P 400 S&P 500 S&P 600 All Indices With W/O Total With W/O Total With W/O Total With W/O Total Total Panel C: Negative Initiations Segmented by SEC Filings/News S&P 400 S&P 500 S&P 600 All Indices With W/O Total With W/O Total With W/O Total With W/O Total Total Panel D: All Observations Segmented by SEC Filings/News S&P 400 S&P 500 S&P 600 All Indices With W/O Total With W/O Total With W/O Total With W/O Total Total

9 3B. Methodology We use standard event study methodology with a two day announcement window. Using an estimation period of 255 days, ending 46 days before the event day, cumulative average abnormal returns () are calculated. To calculate the s, the parameters of the following market model equation are estimated using ordinary least squares regression: R j,t = α + β j R m,t + ε j,t (1) where R j,t is the rate of return on security j on day t, R m,t is the rate of return on the CRSP valueweighted index on day t, and α is the intercept term. β j is a term showing how returns of an individual stock are influenced by market movements, and ε j,t is an error term. After estimating the market model parameters, the measure of abnormal returns (AR) is calculated using equation (2), and the average abnormal returns (AAR) are calculated using equation (3): AR j,t = R j,t - (α j + β j R j,t ) (2) n 1 AARt = Σ AR j, t (3) n j=1 The measure of cumulative average abnormal returns for the event period day zero to day one is given by equation (4). 1 n 1 = Σ AR j, t (4) n 0,1 Σ t=0 j=1 The precision weighted cumulative abnormal return, as developed by Cowan (2000), is: n T 2 PW T,T = w 1 2 j AR j,t (5) j= 1 t= T where w j is a function of the standard deviations, s, of cross-sectional returns, and a measure of the portfolio weights implicit in the standardized cross-sectional test given by: j= 1 T 2 s t= T1 N T 2 t= Ta 1 2 1/ 2 Aj,t w j = (6) 2 1/ 2 ( s ) The nonparametric sign test statistic (generalized z-statistic) is described in Cowan (1992). Aj,t 9

10 4. Results Abnormal returns and other statistics are presented in Table 3. Panel A shows the results for all observations divided according to the initial analyst s recommendation. As expected, a statistically significant positive mean is associated with positive initial recommendations. This finding is consistent with previous studies; however, the of 0.86% is lower than observed in previous studies. Statistically significant negative mean s are found for neutral and negative initial recommendations. This finding is consistent with the results documented by Barber, Lehavy, McNichols, and Trueman (2001) and conflicts with the results of Irvine (2003). This finding does not support the liquidity hypothesis. If the initiation of coverage by an analyst creates an increase in liquidity that is valued by the market, a neutral recommendation should be associated with a positive abnormal return. 9 It appears that the market interprets the neutral recommendation as bad news. If the market expects analysts to be optimistic, as indicated by McNichols and O Brien (1997), it is logical that neutral recommendations would in fact be viewed negatively. Panel B of Table 3 shows the abnormal returns for initiations with concurrent SEC filings or news, and Panel C shows the abnormal returns for initiations without concurrent SEC filings or news. The information in both Panels B and C is consistent with Panel A; however, the magnitude of the s varies. Panel D provides the results for a difference in means tests to determine if there is a statistically significant difference in the s reported in Panels B and C. The difference in means tests indicates that there is not a statistically significant difference in the market reaction for initiations with concurrent SEC filings or news versus those without a concurrent information event. 10 In order to provide a more direct comparison to previous studies, the positive initiations are divided into two groups by strength of recommendation. The s for this analysis are presented in Table 4. Both the strong buy initial recommendations and the buy initial recommendations are associated with statistically significant abnormal returns. These results are consistent with previous studies; however, the s are smaller than documented in previous studies. In Tables 5 and 6 the positive and neutral initiations are examined on a year-by-year basis. 11 For the positive initiations reported in Panel A of Table 5, the mean s for each year are positive and statistically different from zero. The same cannot be said for the neutral initiations. The results reported in Panel A of Table 6 indicate that the mean s for 2001 are not statistically different from zero. 9 The liquidity hypothesis cannot be absolutely rejected based on these results. In order to conclusively reject the liquidity hypothesis, variables related to liquidity such as bid/ask spreads should be considered; and the authors do not have access to this data. 10 The t statistic used for the final panel of Tables 3 through 7 is based on a pooled test that assumes that the two populations have equal variances. The Satterthwaite t statistic used does not assume that the populations have equal variances. 11 Negative initiations are excluded from this analysis due to the small number of observations. 10

11 Table 3 Market Reactions to Equity Analysts Initiations of Coverage Panel A shows s and other statistics for all initiations, Panel B for initiations that were concurrent with SEC filings or news events, and Panel C for initiations that were not concurrent with SEC filings or news events. Panel D provides statistics for a difference in means test to determine if there is a statistically significant difference in the s for initiations with vs. without concurrent SEC filings or news events. The symbols *, **, and *** denote statistical significance at the 5%, 1% and 0.1% levels, respectively. Rec. Type N Mean Precision Weighted Positive: Negative Z Generalized Sign Test Panel A: All Initiations Positive % 0.79% 795: *** 8.610*** Neutral % -0.44% 379: *** *** Negative % -1.43% 38: *** *** Panel B: Initiations with Concurrent SEC Filings/News Positive % 0.53% 190: *** 3.546*** Neutral % -0.65% 87: *** *** Negative % -1.32% 14: ** Panel C: Initiations without Concurrent SEC Filings/News Positive % 0.80% 604: *** 7.823*** Neutral % -0.41% 281: *** *** Negative % -1.49% 25: *** *** Panel D: Tests for Difference in Mean s (Panel B vs. Panel C) Difference in Mean s Positive -0.23% t = Satterthwaite t = Neutral -0.39% t = Satterthwaite t = Negative 0.46% t = 0.66 Satterthwaite t = 0.65 Panels B, C, and D of Tables 3 through 6 provide the results for tests to determine the impact of SEC filings and news events on returns. There is not a statistically significant difference in the s for observations with concurrent SEC filings or news events versus those without. Based on this finding, it does not appear necessary to screen for concurrent information events when working with analysts initiations. In order to examine whether company size impacts the market reaction to an analyst s initial recommendation, we divided the sample into size groups according to the attendant S&P Index in which the companies are included. The results for this analysis are presented in Table 7. For the positive initiations in Panel A, it is not surprising that firms in the S&P 600 have the highest mean s because smaller firms are typically covered by very few analysts and do not have 11

12 as much publicly available information as do larger firms. Therefore, if analysts initiations of coverage convey information to the market, the information provided by increased analyst coverage should have a larger incremental impact for small firms than for larger firms. This finding is supported by the difference in means tests reported in Panel C, which indicates that the mean s for the smaller (S&P 600) firms are statistically greater than the mean CARRs for the larger (S&P 500) firms. The results reported in Panels A and C confirm that, for positive initiations, the mean s are impacted by firm size. 12 Table 4 Market Reactions to Equity Analysts Initiations of Coverage for Positive Recommendations Only Panel A shows s and other statistics for all positive initiations divided into two groups: (1) initial recommendations that are classified as strong buy (the analyst firm s most positive recommendation), and (2) initial recommendations that are classified as buy (positive recommendations that are below the highest recommendation). Panel B provides statistics for initiations that were concurrent with SEC filings or news events, and Panel C provides statistics for initiations that were not concurrent with SEC filings or news events. Panel D provides statistics for a difference in means test to determine if there is a statistically significant difference in the s for positive initiations with vs. without concurrent SEC filings or news events. The symbols *, **, and *** denote statistical significance at the 5%, 1% and 0.1% levels, respectively. Rec. Type N Mean Precision Weighted Positive: Negative Z Generalized Sign Test Panel A: All Positive Initiations Strong Buy % 1.03% 256: *** 6.295*** Buy % 0.73% 546: *** 6.621*** Panel B: Positive Initiations with Concurrent SEC Filings/News Strong Buy % 0.73% 63: * 3.306* Buy % 0.57% 129: ** 2.510* Panel C: Positive Initiations without Concurrent SEC Filings/News Strong Buy % 1.15% 195: *** 5.744*** Buy % 0.72% 418: *** 6.269*** Panel D: Tests for Difference in Mean s (Panel B vs. Panel C) Difference in Mean s Strong Buy -0.29% t = 0.65 Satterthwaite t = Buy -0.16% t = 0.08 Satterthwaite t = We also conducted the analysis using only firms without concurrent SEC filings or news. The results are essentially the same as reported here. 12

13 Table 5 Market Reactions to Equity Analysts Positive Initiations, With versus Without Concurrent SEC Filings/News, By Year Panel A shows a year by year analysis of s and other statistics for all positive initiations. Panel B (C) provides similar statistics for positive initiations that were (were not) concurrent with SEC filings or news events. Panel D provides statistics for a difference in means test to determine if there is a statistically significant difference in the s reported in Panels B and C on a year-by-year basis. The symbols *, **, and *** denote statistical significance at the 5%, 1% and 0.1% levels, respectively. Year N Mean Precision Weighted Positive: Negative Z Generalized Sign Test Panel A: All Positive Initiations % 0.70% 219: *** 3.171*** % 0.63% 177: ** 3.394*** % 0.78% 123: *** 3.934*** % 0.98% 154: *** 5.082*** % 0.84% 122: *** 4.150*** Panel B: Positive Initiations with Concurrent SEC Filings/News % 1.18% 61: ** 2.496** % 0.05% 31: % -0.60% 19: % 0.41% 41: ** % 0.81% 38: ** 2.020* Panel C: Positive Initiations without Concurrent SEC Filings/News % 0.49% 160: * 2.441** % 0.71% 146: *** 3.892*** % 0.86% 101: *** 3.643*** % 1.11% 111: *** 4.098*** % 0.83% 86: *** 4.002*** Panel D: Tests for Difference in Mean s (Panel B vs. Panel C) Difference in Mean s % t = 1.01 Satterthwaite t = % t = Satterthwaite t = % t = Satterthwaite t = % t = Satterthwaite t = % t = 0.33 Satterthwaite t =

14 Table 6 Market Reactions to Equity Analysts Neutral Initiations, With versus Without Concurrent SEC Filings/News, By Year Panel A shows a year by year analysis of s and other statistics for all neutral initiations. Panel B (C) provides similar statistics for neutral initiations that were (were not) concurrent with SEC filings or news events. Panel D provides statistics for a difference in means test to determine if there is a statistically significant difference in the s reported in Panels B and C on a year-by-year basis. The symbols *, **, and *** denote statistical significance at the 5%, 1% and 0.1% levels, respectively. Year N Mean Precision Weighted Positive: Negative Z Generalized Sign Test Panel A: All Neutral Initiations % -0.52% 64: % -0.52% 71: * * % -0.42% 89: * % -0.40% 73: ** % -0.40% 82: ** ** Panel B: Neutral Initiations with Concurrent SEC Filings/News % -1.16% 12: * % -0.71% 18: % -0.76% 17: * * % -0.58% 21: * % -0.45% 19: * * Panel C: Neutral Initiations without Concurrent SEC Filings/News % -0.44% 48: % -0.62% 53: * % -0.23% 71: % -0.44% 49: * * % -0.39% 60: * * Panel D: Tests for Difference in Mean s (Panel A vs. Panel B) Difference in Mean s % t = 0.16 Satterthwaite t = % t = Satterthwaite t = % t = 0.04 Satterthwaite t = % t = 0.21 Satterthwaite t = % t = Satterthwaite t =

15 The results for the neutral initiations reported in Panels B and D of Table 7 are more difficult to interpret. The mid-cap S&P 400 firms have the largest average market reaction. The average market reaction for these firms is statistically larger than the reaction for the S&P 600 firms. There is not a significant difference between the mean s for the S&P 500 and S&P 600 firms. Therefore, the market reaction for neutral initiations does not appear to be strongly related to firm size. Table 7 Market Reactions to Equity Analysts Positive and Neutral Initiations, by Firm Size Panel A (B) shows s and other statistics for all positive (neutral) initiations segmented by firm size. Firm size is determined by S&P index in which the firm is listed. Panels C and D provide difference in means test results for mean s reported in Panels A and B. The symbols *, **, and *** denote statistical significance at the 5%, 1% and 0.1% levels, respectively. Index N Mean Precision Weighted Positive: Negative Z Generalized Sign Test Panel A: Positive Initiations S&P % 0.56% 267: *** 4.559*** S&P % 0.43% 223: *** 2.469** S&P % 1.25% 291: *** 6.418*** Panel B: Neutral Initiations S&P % -0.90% 107: *** *** S&P % -0.19% 149: * S&P % -0.55% 117: ** * Panel C: Tests for Difference in Mean s for Positive Initiations S&P 400 vs. S&P 500 t = 1.51 Satterthwaite t = 1.51 S&P 400 vs. S&P 600 t = Satterthwaite t = S&P 500 vs. S&P 600 t = -3.01* Satterthwaite t = -3.09* Panel D: Tests for Difference in Mean s for Neutral Initiations S&P 400 vs. S&P 500 t = Satterthwaite t = S&P 400 vs. S&P 600 t = -2.19* Satterthwaite t = -2.17* S&P 500 vs. S&P 600 t = 0.80 Satterthwaite t = 0.79 To examine further the effect of firm size on the market reaction to analysts initiations, the following equation is estimated: j,t = α 0 + α 1 SIZE j,t + ε j,t (7) where SIZE j,t is the log of the market capitalization for firm j on day t. The results from the estimation of Equation 7 are presented in Table 8. For the positive initiations the intercept coefficient is positive and statistically significant, indicating that there is a positive market reaction to analysts initiations of coverage after controlling for firm size. Likewise the intercept coefficients for the neutral and negative initiations are negative and statistically significant, indicating a negative market reaction after controlling for firm size. This finding supports the 15

16 results reported previously, indicating that analysts initiations convey information to the market and the finding holds even after controlling for firm size. Table 8 Analysis of the Relationship Between Firm Size and s This table gives the estimates for the following ordinary least squares regression equation: j,t = α 0 + α 1 SIZE j,t + ε j,t. The dependent variable j,t is the two day (0, +1) cumulative average abnormal return for firm j on day t and SIZE j,t is the log of market value for firm j on day t. The regression equation is estimated using all initiations (including observations with and without concurrent SEC filings and news). The t-statistic is provided in parentheses below the estimated coefficient. The symbols *, **, and *** denote statistical significance at the 5%, 1% and 0.1% levels, respectively. Variable Positive Initiations Neutral Initiations Negative Initiations INTERCEPT (4.06)*** (-2.73)** (-2.29)* SIZE (-3.39)*** (2.32)* (1.81) Adjusted R F-Statistic 11.49*** 5.39* Conclusions There are conflicts among the results of previous studies examining the market reaction to analysts initiations of coverage. We contribute to this literature by examining the market reaction for a sample that is drawn from multiple years for firms of various size during the post- Reg FD era. Consistent with all previous studies on this topic, we find a positive and statistically significant market reaction to initiations of analyst coverage when the initial recommendation is positive. For initiations where the recommendations are neutral and negative, the abnormal returns are negative and statistically significant. This result is consistent with the findings of Barber, Lehavy, McNichols, and Trueman (2001) but conflicts with the findings of Irvine (2003). Furthermore, our results do not support the liquidity hypothesis, and it appears that the market interprets neutral initiations as being bad news. We also find that the market reaction is greatest for positive analysts initiations for small-cap firms; however, for neutral initiations the market reaction is greatest for mid-cap firms. We conclude from our findings that, on average, analysts initiations of coverage in the post-reg FD era are viewed by the market as valuable and informative. 16

17 References Barber, Brad, Reuven Lehavy, Maureen McNichols, and Brett Trueman, 2001, Can Investors Profit from the Prophets? Security Analyst Recommendations and Stock Returns, The Journal of Finance, 56 (2), p Barber, Brad, and Douglas Loeffler, 1993, The Dartboard Column: Second-Hand Information and Price Pressure, Journal of Financial and Quantitative Analysis, 28 (2), p Beneish, Messod, 1991, Stock Prices and the Dissemination of Analysts Recommendations, The Journal of Business, 64 (3), p Bjerring, James, Josef Lakonishok, and Theo Vermaelen, 1983, Stock Prices and Financial Analysts Recommendations, The Journal of Finance, 38 (1), p Bradley, Daniel J., Bradford D. Jordan, and Jay R. Ritter, 2003, The Quite Period Goes out with a Bang, The Journal of Finance, 58 (1), p Branson, Bruce, Daryl Guffey, and Donald Pagach, 1998, Information Conveyed in Announcements of Analyst Coverage, Contemporary Accounting Research, 15 (2), p Brennan, M., Subrahmanyan, A., 1995, Investment Analysis and Price Formation in Securities Markets, Journal of Financial Economics 38, p Cowan, Arnold, 1992, Nonparametric Event Study Tests, Review of Quantitative Finance and Accounting 2, p Cowan, Arnold, 2000, Eventus User s Guide, Cowan Research LC, page 107. Davies, Peter and Michael Canes, 1978, Stock Prices and the Publication of Second-Hand Information, The Journal of Business, 51 (1), p Findlay, Scott, and Prem Mathew, 2006, An Examination of the Differential Impact of Regulation FD on Analysts Forecast Accuracy, The Financial Review 41, p Ferreira, Eurico J., and Stanley D. Smith, 2006, Effect of Regulation FD on Analysts Recommendations, Financial Analysts Journal, 62 (3), p Gintschel, Andreas, and Stanimir Markov, 2004, The Effectiveness of Regulation FD, Journal of Accounting and Economics 37, p Der Hovanesian, Mara, and Amy Borrus, 2005, Can the Street Make Research Pay? Business Week 3918, p

18 Irvine, Paul, 2003, The Incremental Impact of Analyst Initiation of Coverage, Journal of Corporate Finance, 9, p McNichols, Maureen, and Patricia O Brien, 1997, Self-Selection and Analyst Coverage, Journal of Accounting Research, 35, p Mikhail, Michael B., Beverly R. Walther, and Richard H. Willis, 2004, Do Security Analysts Exhibit Persistent Differences in Stock Picking Ability? Journal of Financial Economics, 74, p Peterson, David, 1987, Security Price Reactions to Initial Reviews of Common Stock by the Value Line Investment Survey, The Journal of Financial and Quantitative Analysis, 22 (4), p Roulstone, Darren, 2003, Analyst Following and Market Liquidity, Contemporary Accounting Research, 20 (3), p Securities and Exchange Commission, 2000, Selective Disclosure and Insider Trading, Release Nos , , IC-24599; File No. S Available at Securities and Exchange Commission, 2000, Speech by SEC Staff: Regulation FD-An Enforcement Perspective. Available at Securities and Exchange Commission, 2001, Special Study: Regulation Fair Disclosure Revisited, Available at Sok, Tae Kim, Ji-Chai Lin, and Myron Slovin, 1997, Market Structure, Informed Trading, and Analysts Recommendations, Journal of Financial and Quantitative Analysis, 32 (4), p