Star sell-side analysts listed by Institutional Investor, Wall Street Journal and StarMine. Whose recommendations are most profitable?

Transcription

1 Star sell-side analysts listed by Institutional Investor, Wall Street Journal and StarMe. Whose recommendations are most profitable? Yury O. Kucheev 1,2, Kathryn M. Kamski 3,4, Felipe Ruiz 1, Tomas Sorensson 2,4 Abstract We document that the highest monthly abnormal return, 1.58 percent, of holdg a buy and a sell portfolio, is obtaed by followg the recommendations of the group of star sell-side analysts rated by StarMe s Top Earngs Estimators durg the period In this study we compare the profitability of vestment recommendations of analysts who were listed the four star rankgs by Institutional Investor magaze, StarMe s Top Earngs Estimators and Top Stock Pickers, and Wall Street Journal. The results dicate that the choice of analysts rankg is economically important when makg vestment decisions. Keywords: Star Analysts, Analyst recommendations, StarMe, Institutional Investor, Wall Street Journal JEL Classification Numbers: G1, G2 This draft: Feb 23, 2015 Acknowledgement: This research has been conducted as a part of the EMJD Programme European Doctorate Industrial Management (EDIM) funded by the European Commission, Erasmus Mundus Action, which is gratefully acknowledged. We would like to thank Joaqu Ordieres from the Technical University of Madrid for his help data processg. All errors are that of the authors. 1 Department of Industrial Management, Busess Admistration and Statistics, School of Industrial Engeerg, Technical University of Madrid (UPM)/Universidad Politécnica de Madrid, c/ José Gutiérrez Abascal, 2, Madrid, Spa 2 Department of Industrial Economics and Management, School of Industrial Engeerg and Management, KTH- Royal Institute of Technology, SE Stockholm, Sweden 3 Campbell & Company, LP, 2850 Quarry Lake Drive, Baltimore, MD 21209, USA 4 Swedish House of Fance, Stockholm School of Economics, Stockholm, Sweden 1 Correspondg author s contacts: kucheev@kth.se

2 1 Introduction This study analyzes whether vestors can profit from the recommendations of security analysts. We further exame if the vestor s choice of ratg agency matters. Academic theory and banks do not come to the same conclusion about the value of the security analysts. The semistrong form of market efficiency states that vestors should not be able to make profits from tradg on publicly available formation, such as analysts recommendations. However, banks and other firms spend large amounts of money on research departments and analysts dog security analysis, presumably sce they and their clients believe that such security analysis can generate large abnormal returns. That the market value security analysis and analysts is also manifested by the establishment 1998 and the growth of StarMe, a competitor to the Wall Street Journal s and Institutional Investor s rankg of analysts. StarMe writes on their homepage: StarMe is the world's largest and most trusted source of objective equity research performance ratgs (StarMe, 2015). The above observations provide a strong motivation for our study and distguish our analysis from studies which focus on corporate actions. Studies on dividend policy, share repurchases and stock splits, or firm characteristics, such as recent firm performance, actions not directly tied to how people vest their funds. In our study we analyze the economic value of security analysis an activity - that is carried out by thousands of professionals the fance dustry with the goal of improvg their clients return performance. The possibility that there could be profitable vestment strategies based on published recommendations from security analysts is supported by the research of Stickel (1995) and Womack (1996) who show that favorable (unfavorable) changes dividual analyst s recommendations are accompanied by positive (negative) returns at the time of their announcements. Womack documents a post-recommendation stock price drift for upgrades to last up to one month and for downgrades up to six months. Our paper s perspective, however, is different from that of Stickel and Womack. Sce they are terested measurg the average price reaction to changes dividual analyst s recommendations, they take an analyst and event-time perspective. Their perspective can only provide evidence as to whether, with no transaction costs, profitable vestment strategies could potentially be set up by the use of those recommendations. In comparison we take the calendartime perspective of vestors. Our research is related to Barber, Lehavy, McNichols, and Trueman (2001) sce we want to measure the abnormal returns to a number of vestment strategies, with a special focus on the difference between rankgs of security analysts, done by Institutional Investor, Wall Street Journal and StarMe, and the profitability of their 2

3 recommendations. By this approach we are able to establish if vestors can earn positive abnormal returns on vestigated strategies and if there is a difference profitability by usg different rankgs, comparg star analysts recommendations with non-star analysts recommendations and also comparg by differentiatg by which rankg they are ranked by. We use data from the Thompson Fancials Institutional Brokers Estimate System (I/B/E/S) Detail Recommendations File coverg the time period of We manually collected lists of star analysts from the Institutional Investor magaze (October 2003 October 2013), Wall Street Journal (May 2003 April 2013), and StarMe (October 2003 August 2013). The lists of Stars are matched with I/B/E/S by analysts names and broker affiliations. Our fal database contas 281,178 recommendations for 4,643 companies listed on NYSE, AMEX and NASDAQ markets which were announced from January 2002 to December The hand-collected database enables us to do research not done before on comparg the profitability of StarMe s rankgs of analysts with the rankgs by Institutional Investor, and Wall Street Journal. Usg this database, we track the vestment value of portfolios formed by the recommendations of the star analysts, non-star analysts and between stars from different rankgs with a calendar-time approach. In le with Barber et al (2001) and Fang and Yasuda (2013) we sort analysts accordg to their star/non-star status and form calendar-time buy and sell portfolios for each group. We separate our sample under two time frames: Year Before and Year After, which correspond to an evaluation year and to the one year period after particular star rankg is announced. We only use firms that are covered by a star analyst the year after becomg a star or durg the Year Before (evaluation year) and fd all other analysts coverg the same firms (group of Non-Stars). Thus, we only use firms that are covered by a star analyst our portfolios. The portfolio composition is formed accordg to the recommendations issued by a particular group of analysts. Every time an analyst reports to start coverg a firm, changg his or her ratg of a firm, the firm is cluded or excluded from the portfolio. In order to compare different rankgs, we use well-established buy-and-hold portfolio simulation strategy with a holdg period of 30 calendar days. We calculate a time series of daily returns and estimate standard risk-adjusted alphas for each portfolio. Any return that vestors might have earned from prior knowledge of the recommendations or from tradg the recommended stocks durg the day of recommendation is not cluded the return calculations. For our sample period, we fd that the recommendations of star analysts, generated higher excess returns, 1.36 percent, than recommendations by non-stars, 1.01 percent, although differences alpha differentials are not statistical significant. 3

4 Among the entire groups of stars, the best performance was observed for StarMe s Top Earngs Estimators with an excess return of 1.58 percent, while the worst performance was for the group of Institutional Investor stars with an excess return of 1.06 percent. However, on a detailed level the Institutional Investor s buy portfolio is the number two portfolio, but their sell portfolio is way below the other groups, which we terpret as that they focus more on buy recommendations. Comparg buy portfolios of the top-ranked analysts, we fd that number one-ranked analysts by Wall Street Journal and Institutional Investor had higher returns than the group of Non-Star analysts and that difference returns was statistically significant. The above results show that Star analysts who are ranked accordg to the method used by StarMe s Top Earngs Estimator which considers accuracy of earng forecasts as well as profitability of recommendations, issue more profitable recommendations than star analysts who are ranked based exclusively on the performance of recommendations (that are stars listed by the Wall Street Journal and StarMe s Top Stock Picker ). Our fdgs contribute to the existg literature on sell side star analysts sce we answer the question: Can any star rankg be an dicator of a future recommendations profitability or it is only a reflection of the past performance? In answerg this question it is of great importance to vestigate the lk between past and future performance terms of persistency issug accurate forecasts and profitable recommendations. Such persistency has been studied by Hall and Tacon (2010), Hess, Kreutzmann and Pucker (2011), Bilski, Lyssimachou and Walker (2013). However, to a lesser extent has been examed the relationship between analysts rankgs and the profitability of their recommendations (Emery and Li, 2009; Fang and Yasuda, 2013; Groysberg et al., 2008; Leone and Wu, 2007), which this paper focuses on. Fang and Yasud (2013), and Leone and Wu (2007), vestigated the effect of analysts reputation on profitability of recommendations and found that star analysts had better stock pickg ability than their non-star peers. In contrast, Emery and Li (2009), compared the profitability of the two prestigious star rankgs by Institutional Investor and the Wall Street Journal and concluded that they are popularity contests. Our contribution is that we compare four different star rankgs with the focus on the profitability of vestment recommendations usg a recent data set with unique (hand-collected) list of star analysts. Emery and Li (2009) used the formation ratio which is the t-statistic for the tercept of the regression estimation and that is not a direct performance measurement for profitability of recommendations as our study is focused on. In their paper, Fang and Yasuda (2013) discuss the difference returns for Institutional Investor stars comparison to all other 4

5 analysts (Non-Stars), when we are considerg only firms followed by star analysts our sample and we also primarily compare different rankgs among each other. In this study we contue to explore the relationship between reputation and profitability of recommendations by examg various star rankgs which use different evaluation approaches for selectg analysts. 1.1 Rankg evaluation approaches Star analysts are identified as trusted and rated fancial advisors to the dividual and stitutional vestors. An analyst is rated as a star accordg to his/her previous quality of reports, accuracy of forecasts and return that he/she might have generated for his/her clients (Loh and Mian, 2006). Ratgs of sell-side analysts could be maly divided to two groups accordg to the evaluation approach they use: (1) rankgs which are based exclusively on an vestment value of recommendations, such as Best on the Street issued by the Wall Street Journal (WSJ), and Top Stock Pickers issued by the Thomson Reuters StarMe (TSP); and (2) rankgs that use mixed evaluation methods, such as survey-based All-America Research Team issued by Institutional Investor (I/I) magaze, and Top Earngs Estimator issued by StarMe (TEE). In order to select the members the All-America Research Team rankg, Institutional Investor (I/I) magaze sends a questionnaire to buy-side vestment managers, askg them to evaluate various attributes of sell-side analysts. This list of stars is published October and is usually supplemented by 12 attributes that vestors view as the most important to possess. Attributes such as dustry knowledge and tegrity are listed among the most important, while stock selection and earngs estimates are among the least-ranked attributes. Thus the I/I rankg is not primarily focused on the stock pickg ability but rather covers a wide range of attributes which are perceived to directly or directly relate to the ability of an analyst to make profitable recommendations. Previous research shows mixed results regardg profitability of recommendations issued by I/I stars. Measurg the vestment value of recommendations durg , Fang and Yasuda (2013), reported that I/I stars outperformed the group of non-stars, havg Carhart 4-factor monthly alphas of 1.25 percent for Buy portfolios and 0.83 percent monthly alphas for Sell portfolios of I/I stars, compared with 1.09 percent and 0.71 percent for Buy and Sell portfolios for non-stars, respectively. Usg historical data from 1993 to 2005, Emery and Li (2009) vestigated the I/I and WSJ ratgs. The authors identified the determants of star status, and compared both rankgs on the basis of EPS accuracy and the dustry-adjusted performance of vestment recommendations the year before, and one year after analysts become stars. Emery and Li (2009) found that after becomg stars, star analysts 5

6 forecast accuracy of earngs per share (EPS) is not different from non-star peers; recommendations of I/I stars are not statistically better than for non-stars, while recommendations of WSJ stars are significantly worst. They concluded that both rankgs are largely Popularity Contests and do not provide any significant vestment value. In contract, Leone and Wu (2007) vestigated the vestment value of I/I stars recommendations issued from 1991 to 2000 and found that star analysts persistently issue profitable recommendations and such outperformance was not due to luck but could be explaed by a superior ability to pick the right stocks. Sce 1993, the Wall Street Journal publishes a list of Best on the Street analysts (before 2000 this rankg was named as All-Star Analysts ). This rankg is based on the score which an analyst obtaed durg the previous year calculated as a sum of one-day returns of recommendations (if an vestor would vest one day before a recommendation is announced and realize the return by the end of the recommendation day). Such evaluation methodology is focused on a short-term price forecast and it favors analysts who issue a higher number of recommendations and do so on the days when the price changes the most. At the same time it penalizes analysts who issue their recommendations before or after such a day of sharp price changes. Additionally, order to benefit from such recommendations, vestors should be able to receive the recommendation one day before it is announced which could be the case for a limited number of vestors with privileged access to analysts recommendations. Also, WSJ s evaluation method is bld for avoidg analysts who announced their recommendations on the same day but after a significant price change had already happened (Yaros and Imielski, 2013). All those concerns might lead to a substantial random selection of analysts to a star rankg. Emery and Li (2009) found that after becomg stars, WSJ star analysts issue recommendations which underperformed the group of non-stars. The authors terpret this result by the effect of the regression to the mean sce the short-term recommendation performance cludes a substantial random component. Annual Thomson Reuters StarMe Stock Pickg Awards (TSP) and Earngs Estimate Awards (TEE) have been issued sce 1998, and are based on a two-step measurement of the previous one-year profitability. The Coverage-Relative Ratg is the first evaluation step for both rankgs. It is based on the excess returns of a long-only portfolio constructed accordg to all recommendations of each analyst. It measures how well an analyst distguishes among the stocks he/she covers. The second step for the Stock Pickg Awards, all recommendations for each analyst are evaluated usg long and short buy-and-hold portfolio method adjusted to the market capitalization-weighted portfolio for a given dustry. For the second step of the Top 6

7 Earngs Estimator awards, the accuracy and timg of earngs forecasts is evaluated. Even though StarMe s rankgs appeared much later, they play an essential role sell-side research by providg fluential and an important reference the dustry (Kim and Zapatero, 2011). Accordg to Beyer and Guttman (2011) and Ertimur et al. (2011), many Wall Street firms use StarMe rankgs when defg the payments to their analysts. Recent work by Kerl and Ohlert (2013) vestigates the accuracy of earngs per share forecasts and target prices of StarMe analysts comparison with their non-star peers one year after the analysts became stars. They found that analysts possess a persistent ability to issue accurate earngs forecasts, as after becomg stars they keep issug more accurate earngs forecasts than non-star analysts. Regardg the accuracy of target prices, the authors could not fd any difference between the two groups of analysts. The reason for such significant difference TP forecasts could be due to the research methodology: star analysts with Stock pickg awards and Earngs estimate awards were grouped together order to compare their accuracy with non-stars, without splittg a sample of Stars to good stock pickers and accurate forecasters. However, accordg to the StarMe methodology for Stock pickg awards, analysts are not evaluated on the basis of accuracy of EPS. Thus, it is possible that even the year before they receive an award, the mixed group of stars does not outperform non-star peers accuracy of their forecasts. Furthermore, Kerl and Ohlert (2013) solely focus on accuracy of EPS or TP and factors that fluence accuracy, and do not provide any comparison of the performance of recommendation issued by star analysts with non-stars. 2 Data and descriptive statistics We use four data sources. The Thompson Fancials Institutional Brokers Estimate System (I/B/E/S) Detail Recommendations File gives standardized stock recommendations for all various brokers scales: mappg all recommendations on a fal scale from 1 to 5, where 1 corresponds to a Strong Buy, 2 to Buy, 3 to Hold, 4 to Sell and 5 to a Strong Sell. The Center for Research Security Prices (CRSP) Daily Stock File provides daily holdg period stock returns which clude dividends, price and cash adjustments. The Fama-French Factors Monthly Frequency database provides monthly returns for the factors of value-weighted market dex, for the size, book-to-market and momentum. We manually collected lists of star analysts from the Institutional Investor magaze (October 2003 October 2013), Wall Street Journal (May 2003 April 2013), and StarMe (October 2003 August 2013). The lists of stars are matched with I/B/E/S by analysts names and broker affiliations, and double checked for any possible consistencies (typos names, if analyst changed a broker a given year etc.). Our 7

8 sample does not clude analysts from some brokerage houses, particularly Lehman Brothers and Merrill Lynch, sce their recommendations are no longer available at I/B/E/S. To be able to make a fair comparison of performance different groups, we limit our sample to firms which are followed by Star analysts durg a one-year period after a particular list of stars was published or durg the previous calendar year (that is evaluation year). As a result, our sample contas only firm-year observations for which there is at least one recommendation by Star analyst durg the specified time period. Our fal database contas 281,178 recommendations for 4,643 companies listed on NYSE, AMEX and NASDAQ markets which were announced from January 2002 to December The entire sample of analysts is divided to the followg groups: (1) Stars and Non-Stars; (2) Institutional Investor (I/I), Wall Street Journal (WSJ), StarMe Top Stock Pickers (TSP) and Top Earngs Estimator (TEE); (3) Analysts ranked as Number one (Top-Ranked): WSJ-1, I/I-1, TSP-1, TEE-1. When a particular analyst is rated as a Star two different dustries, this analyst is cluded only once to a particular group of Stars. We compare these groups under two time frames: 1) The Year Before is the calendar year before the rankg is announced. For example, for WSJ list of stars announced May 2003, the previous year is the calendar year from January 2002 until December 2002 which is actually an evaluation year for WSJ ratg. As a result, an entire sample period for Year Before spans from January 2002 until December ) The Year After is the one year period from the day when a particular rankg is announced. For example, for WSJ announced May 12, 2003, the Year After begs on that day and ends on May 12, An entire sample period for Year After spans from May 2003 until December Because the last month is December 2013, we do not cover the whole Year After for groups of stars selected 2013 due to data availability. Table 1 shows the number of firms the sample, which was between 1075 year 2013 and 1887 year 2007, and also the percentage of firms covered by each group. On average, each group of star analysts covers about 50 percent of firms the sample (WSJ covers 57 percent, I/I 51 percent, TEE 49 percent, TSP 49 percent). This difference suggests that these groups have different firm coverage (issued recommendations for different universe of firms). Insert Table 1 here 8

9 Table 2 displays the total number of analysts the sample on the election-year basis. On average, around 13 percent of analysts are listed as stars every year. Also, there are 13 percent of analysts among non-stars who have been elected as stars some other year but not this year (either before or after that year). Insert Table 2 here As can be seen Table 3, the group of Star analysts issues 28 percent of all recommendations our sample. Both WSJ and I/I Stars issue more recommendations than TEE and TSP. Insert Table 3 here The average overlap the sample years among rankg lists is presented Table 4, which shows the number of analysts listed by different rankgs, the number of the same analysts each pair of rankgs, and Portion of the same analysts for each rankg list. Panel A represents this data for entire groups of stars, while Panel B reports results for the number one-ranked stars. Thus, this table shows the percentage of analysts who appear another rankg. For example, Institutional Investor has on average 10 percent of analysts who were at the same years listed as Top Stock Pickers by StarMe, while on average there are 190 unique names on the Institutional Investor list. It can be seen that Top Stock Pickers and Wall Street Journal have the highest similarity of published lists, while Institutional Investor and Wall Street Journal have the lowest similarity. This terdependence is explaed part by the similarity evaluation methods. It shows how different the lists of star analysts are, as well as the difference the returns of their recommendations. Insert Table 4 here 3 Results: risk-adjusted portfolio returns. 3.1 Methods To measure the profitability of recommendations, we apply the well-established calendar-time portfolio simulation method the year subsequent to the year when rankgs had been assigned (denoted as Year After), and for the year durg which analysts were evaluated (denoted as Year Before) (Barber et al., 2006; Fang and Yasuda, 2013). We use a simulation of buy-and-hold calendar-time Buy and Sell portfolios for each sub-group of analysts. For each new Strong Buy or Buy recommendation, $1 is vested at the end of the recommendation announcement day (or at the close of the next tradg day if recommendation is issued on a non-tradg day) to the Buy portfolio. The stock is held the portfolio for the followg 30 calendar days if there are no recommendation revisions or recommendation changes from the same analyst. If 9

10 durg the followg 30 days the analyst changes recommendation level from Strong Buy or Buy to Hold or Sell or Strong Sell, then the stock is withdrawn from Buy portfolio and cluded to the Sell portfolio by the end of the same tradg day when a new recommendation is issued (or on the next tradg day if recommendation is issued on a non-tradg day). If there is a recommendation revision but the new recommendation is on the same level (that is Buy or Strong Buy) then the stock after this revision is kept the same portfolio for additional 30 calendar days or until the next recommendation change. The same procedures are applied for a Sell portfolio which cludes Hold, Sell, and Strong Sell recommendations. As a result of this strategy, the calendar day t gross return on portfolio p cludes from n=1 to N pt recommendations and could be defed as: N pt X n, t1ri n, t n1 R, (1) pt N pt n1 X n, t1 where X n,t-1 is the cumulative total gross return of stock i n from the next tradg day after recommendation was added to the portfolio to day t-1 which is the previous tradg day before t, that is: X (2) n, t1 Ri, recdat 1Ri, recdat 2 *...* R n n n n, recdatn t1 Monthly portfolio returns are obtaed from a geometric compoundg of the daily returns. Thus, a raw monthly return of a portfolio p is: n r R t 1, (3) t1 where n τ is the number of tradg days the month τ. Monthly excess returns for each group s Buy and Sell portfolios are estimated as an tercept (alpha) which is calculated accordg to the four-factor model proposed by Carhart, (1997): r rf ( rm rf ) s SMB h HML m UMD, (4) where rm τ is a monthly market return, rf τ is the risk-free rate of return, SMB τ is a size factor, that is the difference between value-weighted portfolio returns of small and large stocks, HML τ is a book-to-market factor, that is the difference between the value-weighted portfolio 10

11 returns of high book-to-market and low book-to-market stocks, UMD τ is a momentum factor, that is the difference the returns of stocks with a positive return momentum and those with a negative return momentum over months τ-12 and τ-2. The Alpha differentials (differences alphas) are statistically tested usg seemgly unrelated estimation accompanied with a test for statistical difference tercepts from various regressions (suest and test procedures STATA). 3.2 Results Table 5 represents monthly excess returns (alphas) for Stars and Non-Stars durg the year after rankgs have been published (Year After) and durg the evaluation year (Year Before). The first two rows the table show returns of Buy and Sell portfolios while the last (Buy-Sell) represents the total return of all recommendations for a particular group, which is the difference of the Buy mus Sell portfolio returns. As we see Table 5, Buy mus Sell portfolio of Stars with monthly alphas of percent outperformed Non-Stars with monthly alphas of percent leadg to a difference of 35 basis pots abnormal return for a Buy-Sell portfolio the year after rankgs where published. As could be expected, the evaluation year Stars had higher recommendation returns with percent than Non-Stars with percent. This shows that the group of stars persistently issued more profitable recommendations than their Non-Star peers. Insert Table 5 here. Table 6 shows excess returns of recommendations issued by entire groups of stars: I/I, TSP, TEE and WSJ. Comparg total monthly excess returns of Buy mus Sell portfolios (Buy-Sell) for the year after election, each entire group of Stars outperformed Non-Stars (see Table 5), with the highest return for TEE, percent, and the lowest return for the group of I/I stars, percent. In the year after election, each Buy portfolio of the sub-groups of Stars had higher returns than Non-Stars, while Sell portfolios of I/I and TSP performed worse than the Sell portfolio of Non-Stars. At the same time, the total return of the Buy-Sell portfolio for each entire group of Stars outperformed the Buy-Sell portfolio of Non-Stars. By comparg returns of each entire group of Stars (WSJ, I/I, TEE and TSP) with the returns of all Star analysts (first column Table 5), it can be seen that only Buy portfolios of I/I and TEE outperformed the average return of group of Stars, and only Sell portfolios of WSJ and TEE outperformed Sell portfolio of Stars. These results suggest that TEE outperformed not only Non-Star peers but also that they have higher returns than the total group of Stars. In the evaluation year, each sub-group of star analysts outperformed Non-Stars with the highest return of percent for 11

12 recommendations issued by analysts who were the next year selected by WSJ, and the lowest excess return of percent for recommendations by the next year s I/I Stars. This result is as expected considerg the evaluation methodology applied by WSJ which is focused on the recommendations, and I/I which ranks stock pickg ability among the least important attributes for Star selection. Buy and Sell portfolios of the sub-groups of stars the evaluation year perform accordgly, havg the highest returns for WSJ stars and the lowest for I/I stars. As can be seen, WSJ and TSP stars show a decrease the performance after election, which can be explaed as the regression to the mean. At the same time, vestment value of recommendations by TEE and I/I crease after the rankgs are published. This crease the profitability could be attributed to the fluence of their reputation on the stock prices (Fang and Yasuda, 2013). Insert Table 6 here. Table 7 shows monthly excess returns for the top-ranked analysts, where the highest total return (Buy-Sell) was observed for I/I-1 with percent and the lowest total return for TSP-1 with percent the year after election. Buy portfolios of top-ranked I/I-1, TEE-1 and WSJ-1 analysts are better than Buy portfolios of each entire group of stars and had higher return than Buy portfolio of all Stars (see Table 5), while top-ranked TSP-1 had statistically significant return for their Buy portfolio (excess return equal to zero). Sell portfolios of I/I-1, TEE-1 and TSP-1 outperformed Sell portfolios of the correspondg groups of Stars, with the highest return for Sell portfolio of TEE-1 with 0.71 percent and the lowest for WSJ-1 with 0.25 percent. Among Buy portfolios the year after election, the highest excess return had the Buy portfolio of top-ranked I/I-1 analysts with percent and the lowest TSP-1 with 0.16 percent. In the evaluation year, TSP-1 performed better than all other sub-groups of top-ranked analysts with percent, while TEE-1 had excess returns equal to zero. Considerg an assumption of the persistency of stock pickg ability, WSJ-1, I/I-1 and TEE-1 perform better the year after election, while TSP-1 shows significant drop vestment value of their recommendations from percent excess return to 0.81 percent. Insert Table 7 here. In Table 8 we report the alpha differentials for comparisons between groups of stars the Year After, Year Before as well as Year After with Year Before. In Panel A which has alpha differentials for all Buy portfolios, we see that the Year After only TEE, and number oneranked WSJ-1 and I/I-1 had statistically significant outperformance above the returns of the entire group of Non-Stars. At the same time, I/I-1 stars has statistically significant difference returns with the entire group of Stars and also with WSJ. As soon as TSP-1 had significant 12

13 alphas of their Buy portfolios, we observe that they had statistically lower returns than most of the other groups of Stars. In the Year Before, the right bottom quadrant, most of the alpha differentials where statistically significant. However, the groups of I/I and TEE as well their number one-ranked analysts had statistically significant differences returns with the group of Non-Stars. As we see, the Year Before the differences returns were statistically significant, while the Year After groups of stars perform on the same level from the statistical pot of view, which we terpret as the regression to the mean effect. When we analyze the differences returns from the Year Before to the Year After, the left bottom quadrant Panel A, we conclude that Non-Stars keep performg on the same level which is reflected as significant difference their returns while entire group of Stars had lower returns the Year After (difference is significant). The diagonal of this quadrant is representg the differences for each group with itself two different time periods. We see that groups of WSJ, WSJ-1, TSP and TSP-1 the Year After perform worse than they did the Year Before. This result proves the assumption that rankgs which are based exclusively on the performance of recommendations have low predictive power due to regression to the mean for the best performers from the previous time period. At the same time, groups of I/I, I/I-1 and TEE-1 the Year After had significantly better returns than the Year Before which could be attributed to the fluence of the reputation of those rankgs on the recommended stock prices. Panel B of the Table 8 shows alpha differentials and their statistical significance for all Sell portfolios of various groups of analysts. As we see, only I/I stars had statistically different returns the Year After because they had the lowest return for their Sell portfolio among all groups. In the Year Before the pattern of differences is similar to what we see for Buy portfolios: rankgs which are based on the performance of recommendations, WSJ and TSP, outperform group of Non-Stars while TEE and I/I had alphas statistically not different from that of Non- Stars. When we compare returns from the Year Before with Year After, we fd that for all Sell portfolios of various groups of stars the returns are significantly different from that of the Year After (diagonal of the left bottom quadrant Panel B). We terpret this result as the less priority beg given to sell recommendations by analysts, and a limited possibility of the market to react on such recommendations. Insert Table 8 here. Figure 1 shows a comparison of frequency of months when a particular sub-group appears to be the best group compared to other sub-groups with the same comparison pool. For example, usg raw monthly returns, Stars are compared with Non-Stars: number of months when Stars 13

14 outperformed Non-Stars is divided by the total number of months the sample period. These results are le with abnormal returns analyzed above. We observed that for 55 percent of months our sample period, the total group of Stars outperformed Non-Stars. In the pool with the entire groups of stars, I/I and TEE had the highest number of months when their Buy portfolio outperformed Buy portfolio of other star groups. While the Sell portfolio of WSJ had the highest frequency, the lowest was for the TSP Sell portfolio. For top-ranked analysts, Buy portfolios of I/I-1 and WSJ-1 analysts showed the highest frequency, while TSP and TEE the lowest. However, Sell portfolios of TEE-1, TSP-1 and WSJ-1 showed almost the same frequency of months leavg I/I-1 at 5 percent behd. Insert Figure 1 here. Results of comparg portfolio returns and analyzg the frequencies of months when particular group outperformed the others suggest that star analysts who were listed as Top Earng Estimators by StarMe (TEE) outperformed all the other groups of stars as well as their non-star peers. The Top Stock Pickers (TSP) appear to perform the worst, thus suggestg that this rankg has the lowest predictive power for a future profitability of recommendations. This might be a result of an effect of the regression to the mean, accordg to which the previous year best performers should show results closer to the average on subsequent year. This explanation should lead to the similar regression to the mean for WSJ analysts, however we observed significantly positive returns which outperformed non-stars the year after selection, even though there is a decle performance comparg with the evaluation year. In contrast, returns for I/I and TEE crease after the rankg is published, which could be attributed to the better evaluation process and the effect of reputation of these ratgs on prices of recommended stocks. 4 Conclusion Previous research by Mikhail et al., (2004) found that sell-side analysts are consistent issug profitable recommendations. Logical deduction from those fdgs would be that star rankgs focused on the profitability of recommendations should provide an vestor with the list of analysts who have better stock pickg ability which then will be reflected the profitability of the future recommendations. In our study we found that the entire group of Star analysts shows persistency the performance of their recommendations, however, the complex approach utilized by Top Earngs Estimator rankg leads to substantially higher predictive power than was observed for rankgs which are based exclusively on recommendations. We also found that Institutional Investor s stars outperform non-stars by only 2 basis pots, but their number oneranked stars had the highest return of the Buy as well as Buy-Sell portfolios which might be 14

15 explaed by the fluence that these top-ranked analysts have on the market, by affectg the stock prices by their recommendations. Specifically, our fdgs could be summarized as follows: - Abnormal returns of Star recommendations are higher than Non-Stars for both Buy and Sell portfolios after star rankgs are issued as well as durg the evaluation year. Additionally, Stars outperformed Non-Stars measured by the frequency of months when they had higher raw returns. - Each Buy portfolio of Star analysts (WSJ, I/I, TEE, and TSP) performed better than the Buy portfolio of Non-Stars. However, Sell portfolios of I/I and TSP had lower alphas. All four overall portfolios (Buy-Sell) of stars are better than the Buy-Sell portfolio for Non-Stars. - Buy portfolios of analysts who were ranked exclusively on the performance of their recommendations (WSJ and TSP) performed worse than TEE and I/I the subsequent year. - WSJ and TSP have the highest percentage of the same analysts appearg both rankgs at the same year (with the average of 30 percent of analysts), while WSJ and I/I have the lowest percentage of terdependence (9 percent on average). Similar results are observed for Number one-ranked analysts among the four vestigated groups. - The performance of WSJ and TSP analysts decles the year after election, while TEE and I/I show an crease vestment value of their recommendations. Thus, there is strong evidence that star rankgs that use a mixed evaluation approach are able to identify analysts who have a persistent stock pickg ability. Their recommendations outperform Non-Stars and the other groups of analysts who were ranked only accordg to the past profitability of recommendations. An important fdg is that the survey based rankg by the Institutional Investor magaze, of the number one ranked analysts, shows the highest vestment value of recommendations, even though their evaluation methodology is qualitative and they have low excess returns durg evaluation year comparison to other groups of star analysts. In summary, the choice of which analysts to work with is of a great importance for the longterm growth of the value of the vestor s portfolio. In our study we provided empirical evidence of which Star rankgs of sell-side analysts a potential vestor should have relied on. Additionally, our results show that the stock pickg ability is a reflection of the set of skills which could be captured by the mixed evaluation methods such as surveys or those considerg recommendations and earngs forecasts simultaneously. 15

16 References Barber, B.M., Lehavy, R., McNichols, M., and Trueman, B. (2001). Can Investors Profit from the Prophets? Security Analyst Recommendations and Stock Returns. J. Fance 56, Barber, B.M., Lehavy, R., McNichols, M., and Trueman, B. (2006). Buys, holds, and sells: The distribution of vestment banks stock ratgs and the implications for the profitability of analysts recommendations. J. Account. Econ. 41, Barber, B.M., Lehavy, R., and Trueman, B. (2007). Comparg the stock recommendation performance of vestment banks and dependent research firms. J. Fanc. Econ. 85, Beyer, A., and Guttman, I. (2011). The Effect of Tradg Volume on Analysts Forecast Bias. Account. Rev. 86, Bilski, P., Lyssimachou, D., and Walker, M. (2013). Target Price Accuracy: International Evidence. Account. Rev. 88, Carhart, M.M. (1997). On Persistence Mutual Fund Performance. J. Fance 52, Emery, D.R., and Li, X. (2009). Are the Wall Street Analyst Rankgs Popularity Contests? J. Fanc. Quant. Anal. 44, 411. Ertimur, Y., Mayew, W.J., and Stubben, S.R. (2011). Analyst reputation and the issuance of disaggregated earngs forecasts to I/B/E/S. Rev. Account. Stud. 16, Fang, L.H., and Yasuda, A. (2005). Analyst reputation, conflict of terest, and forecast accuracy. Rod. WHITE Cent. Fanc. Res.-Work. Pap.- 7. Fang, L.H., and Yasuda, A. (2013). Are Stars Opions Worth More? The Relation Between Analyst Reputation and Recommendation Values. J. Fanc. Serv. Res. Groysberg, B., Lee, L.-E., and Nanda, A. (2008). Can They Take It With Them? The Portability of Star Knowledge Workers Performance. Manag. Sci. 54, Hall, J.L., and Tacon, P.B. (2010). Forecast accuracy and stock recommendations. J. Contemp. Account. Econ. 6, Hess, D., Kreutzmann, D., and Pucker, O. (2011). The Good, The Bad, and The Lucky: Projected Earngs Accuracy and Profitability of Stock Recommendations. In AFA 2012 Chicago Meetgs Paper,. Kerl, A., and Ohlert, M. (2013). Star-Analysts Forecast Accuracy and the Role of Corporate Governance. Work. Pap. Available SSRN Kim, M.S., and Zapatero, F. (2011). Competitive compensation and dispersion analysts recommendations. Available SSRN Leone, A., and Wu, J. (2007). What does it take to become a superstar? Evidence from stitutional vestor rankgs of fancial analysts. Evid. Institutional Invest. Rank. Fanc. Anal. May Simon Sch. Bus. Work. Pap. No FR Loh, R.K., and Mian, G.M. (2006). Do accurate earngs forecasts facilitate superior vestment recommendations? J. Fanc. Econ. 80, Mikhail, M.B., Walther, B.R., and Willis, R.H. (2004). Do security analysts exhibit persistent differences stock pickg ability? J. Fanc. Econ. 74, Starme (2015). About us. downloaded February 4 th, Stickel, S.E. (1995). Anatomy of the performance of buy and sell recommentastions.pdf. Fanc. Anal. J. 51, Womack, K.L. (1996). Do brokerage analysts recommendations have vestment value? J. Fance 51, Yaros, J.R., and Imielski, T. (2013). A Monte Carlo Measure to Improve Fairness Equity Analyst Evaluation. Work. Pap. 16

17 Tables and Figures Table 1. Number of firms the sample and percentage of firms covered by each group calculated on the election-year basis. Rankgs done by Wall Street Journal (WSJ), Institutional Investor (I/I) the Thomson Reuters StarMe Top Stock Pickers (TSP), and Top Earngs Estimator issued by StarMe (TEE). When it is dexed by -1, it means the number one-ranked analyst. Each group of star analysts covers about 50 percent of firms the sample. This shows that the coverage universe is different for various groups of stars. Number one-ranked I/I analysts cover a half of firms of the entire group of I/I stars (which cludes 3 other rankg positions). Election year Total Number of Firms Portion of Firms covered by Entire groups of Stars Portion of Firms covered by Number-one ranked Stars WSJ I/I TEE TSP WSJ-1 I/I-1 TEE-1 TSP % 65% 35% 38% 17% 28% 14% 14% % 62% 40% 44% 20% 30% 18% 18% % 62% 40% 41% 17% 27% 20% 18% % 58% 50% 51% 16% 24% 23% 25% % 54% 46% 45% 14% 24% 23% 21% % 57% 55% 55% 18% 29% 29% 29% % 46% 58% 56% 19% 27% 24% 27% % 31% 55% 55% 17% 31% 28% 26% % 31% 57% 54% 18% 31% 26% 25% % 48% 55% 50% 18% 23% 24% 22% % 51% 52% 52% 19% 28% 25% 22% Average % 51% 49% 49% 18% 27% 23% 23% Overall % 58% 73% 77% 46% 36% 48% 53% 17

18 Table 2. Number of analysts and percentage each group represented the sample on the election-year basis. Rankgs done by Wall Street Journal (WSJ), Institutional Investor (I/I) the Thomson Reuters StarMe Top Stock Pickers (TSP), and Top Earngs Estimator issued by StarMe (TEE). When it is dexed by -1, it means the number one ranked analyst. On average there were 13 percent of star analysts per year. Election year All Analysts Non Stars ever elected as stars Portion of Analysts Entire groups of Stars Portion of Analysts Number-one ranked Stars Stars WSJ I/I TEE TSP WSJ-1 I/I-1 TEE-1 TSP % 12% 4% 6% 2% 3% 1% 1% 1% 1% % 13% 5% 6% 3% 3% 1% 1% 1% 1% % 14% 5% 7% 3% 3% 1% 1% 1% 1% % 14% 5% 6% 4% 4% 1% 1% 1% 1% % 14% 5% 6% 4% 4% 1% 1% 1% 1% % 15% 5% 6% 4% 4% 1% 1% 1% 1% % 12% 4% 3% 4% 4% 1% 1% 1% 1% % 12% 5% 1% 4% 4% 1% 1% 2% 1% % 12% 5% 1% 4% 4% 1% 1% 1% 1% % 13% 3% 4% 4% 4% 1% 1% 1% 1% % 14% 3% 4% 5% 5% 1% 2% 2% 2% Average % 13% 4% 5% 4% 4% 1% 1% 1% 1% Overall % 19% 11% 5% 8% 9% 3% 1% 4% 4% 18

19 Table 3. Number of recommendations and percentage of Star recommendations done on the election-year basis by each rankg. Rankgs done by Wall Street Journal (WSJ), Institutional Investor (I/I) the Thomson Reuters StarMe Top Stock Pickers (TSP), and Top Earngs Estimator issued by StarMe (TEE). When it is dexed by -1, it means the number one ranked analyst. Election year Entire sample All Stars Entire groups of Stars Number-one ranked Stars WSJ I/I TEE TSP WSJ-1 I/I-1 TEE-1 TSP % 11% 15% 5% 6% 2% 3% 2% 2% % 11% 14% 5% 6% 2% 3% 2% 2% % 10% 13% 6% 6% 2% 3% 2% 2% % 9% 13% 9% 8% 2% 3% 3% 4% % 12% 12% 7% 10% 5% 3% 2% 3% % 12% 11% 9% 9% 5% 3% 4% 4% % 12% 7% 8% 11% 4% 3% 2% 3% % 9% 3% 8% 8% 2% 3% 3% 3% % 9% 3% 8% 7% 2% 3% 3% 3% % 6% 9% 7% 6% 2% 2% 2% 2% % 9% 8% 9% 8% 3% 3% 3% 3% Average % 10% 10% 7% 8% 3% 3% 3% 3% Overall % 14% 12% 10% 11% 4% 3% 4% 4% 19

20 Frequency Frequency Frequency Figure 1. Frequency of months when particular group of analysts outperformed the other groups. Rankgs done by Wall Street Journal (WSJ), Institutional Investor (I/I) the Thomson Reuters StarMe Top Stock Pickers (TSP), and Top Earngs Estimator issued by StarMe (TEE). When it is dexed by -1, it means the number one ranked analyst. 60% 50% Stars and NonStars Buy Sell 54% 46% 45% 55% 40% 30% 20% 10% 0% NonStars Group of Analysts Stars 35% 30% 25% 20% 15% Entire groups of Stars Buy Sell 31% 27% 28% 27% 26% 23% 23% 15% 10% 5% 0% I/I TEE TSP WSJ Group of Analysts 35% 30% 25% 20% Top-Ranked Star anlysts Buy Sell 32% 30% 27% 26% 27% 21% 19% 19% 15% 10% 5% 0% I/I-1 TEE-1 TSP-1 WSJ-1 Group of Analysts 20

21 Table 4. Average percentage of terdependence among rankgs, average number of analysts listed particular group and a portion of the same analysts for each rankg list. Panel A represents the data for entire groups of stars, while Panel B reports results for number one-ranked stars. The last le shows the average value for each value. Comparison is made on the election-year basis. Rankgs done by Wall Street Journal (WSJ), Institutional Investor (I/I) the Thomson Reuters StarMe Top Stock Pickers (TSP), and Top Earngs Estimator issued by StarMe (TEE). When it is dexed by -1, it means the number one-ranked analyst. The highest correlation among WSJ and TSP, Lowest for WSJ and I/I. Panel A. Entire groups of Stars Year Number of Star Analysts Number of the Same Analysts Portion of the Same Analysts I/I WSJ TSP TEE I/I & WSJ I/I & TSP I/I & TEE WSJ & TSP WSJ & TEE TSP & TEE WSJ I/I TSP I/I TEE I/I I/I WSJ TSP WSJ TEE WSJ I/I TSP WSJ TSP TEE TSP % 11% 15% 19% 26% 12% 25% 45% 13% 37% 23% 14% % 8% 12% 12% 28% 10% 18% 47% 11% 26% 17% 10% % 8% 12% 12% 28% 9% 16% 40% 14% 23% 12% 14% % 7% 16% 11% 32% 10% 11% 41% 23% 24% 12% 22% % 9% 11% 7% 28% 12% 15% 37% 21% 16% 15% 19% % 10% 11% 8% 26% 10% 15% 32% 22% 16% 11% 21% % 17% 13% 10% 23% 14% 13% 27% 21% 9% 16% 19% % 16% 11% 3% 25% 13% 6% 33% 18% 4% 16% 17% % 6% 9% 2% 29% 11% 2% 38% 19% 3% 14% 19% % 7% 10% 7% 25% 7% 8% 20% 22% 10% 5% 20% % 12% 11% 6% 27% 9% 11% 20% 18% 9% 6% 17% Avg % 10% 12% 9% 27% 11% 13% 35% 18% 16% 13% 18% I/I TEE WSJ TEE TSP TEE

22 Panel B. Number one-ranked Analysts Year Number of Star Analysts Number of the Same Analysts Portion of the Same Analysts I/I-1 WSJ-1 TSP-1 TEE-1 I/I-1 & WSJ-1 I/I-1 & TSP-1 I/I-1 & TEE-1 WSJ-1 & TSP-1 WSJ-1 & TEE-1 TSP-1 & TEE-1 WSJ-1 I/I-1 TSP-1 I/I-1 TEE-1 I/I-1 I/I-1 WSJ-1 TSP-1 WSJ-1 TEE-1 WSJ-1 I/I-1 TSP-1 WSJ-1 TSP % 2% 9% 7% 17% 0% 2% 17% 5% 13% 0% 5% % 2% 7% 2% 14% 2% 2% 14% 2% 9% 2% 2% % 5% 5% 0% 18% 5% 7% 16% 9% 6% 4% 8% % 3% 5% 2% 22% 5% 3% 15% 17% 5% 4% 18% % 4% 2% 0% 21% 0% 4% 14% 9% 2% 0% 9% % 5% 5% 0% 15% 5% 5% 10% 10% 5% 4% 11% % 5% 5% 0% 16% 5% 4% 11% 7% 4% 4% 7% % 5% 4% 2% 14% 2% 5% 11% 7% 3% 2% 7% % 2% 4% 2% 9% 5% 2% 7% 5% 4% 4% 5% % 3% 5% 0% 13% 0% 2% 9% 13% 4% 0% 13% % 0% 2% 0% 7% 0% 0% 5% 5% 2% 0% 5% Avg % 3% 5% 2% 15% 3% 3% 12% 8% 5% 2% 8% TEE-1 TSP-1 I/I-1 TEE-1 WSJ-1 TEE-1 TSP-1 TEE-1 22