The Shine of Star: The Effect of Star Analyst Title on Market Reaction to Financial Analysts’ Stock Recommendations Siyu Chen Runjing Lu Department of Economics Department of Economics National University of Singapore University of California, San Diego [email protected] [email protected] Abstract This paper studies how the award-winning titles of financial analysts affect the market reac- tion to their recommending stocks using a regression discontinuity (RD) design and a novel dataset. We find that, right after the award ceremony, investors react positively to the stocks previously recommended by winners, but negatively to the stocks recommended by finalists who fall short of being winners (failed finalists). We provide suggestive evidence that in- formed traders may know the list of finalists and buy their recommending stocks in the week before the ceremony, but sell the stocks by failed finalists after the ceremony. We further show that both the initial negative reaction to failed finalists and the positive reaction to winners completely reverse back to zero within six weeks after the ceremony. However, the market continues reacting more positively to subsequent stocks recommended by winners within one year after the ceremony, though the effect is much smaller. The short-term over- reaction to star title and the speculative tradings around the announcement indicate that the star analyst award is a new factor generating excessive volatility and inefficiency in the market. Keywords: financial analyst, title effect, institutional investors JEL classification: G14, G20, G24 Preprint submitted to Draft December 30, 2017 1. INTRODUCTION Sell-side financial analysts play a key role in collecting, interpreting, and disseminating company and market information to investors. Issuing “buy” and “sell” recommendations is an important part of an analyst’s job and one of the most visible ways for them to express their opinions on the stocks and markets they cover. In an information market such as the one of financial analysts, since the product is ex-ante hard to evaluate, investors may reply on outside certification, such as award-wining status of an analyst, to infer the quality of his or her recommendations. In line with this argument, a large body of literature in finance and accounting have documented that investors react abnormally more to stock recommendations by award-winning financial analysts (hereafter “star analysts”) than those by other analysts.1 However, precisely due to the correlation between analysts’ quality and the star analyst title, it is empirically challenging to cleanly identify the impact of star analyst award per se on the market. In other words, whether and how much do investors react differently to observably similar stocks recommended by similar analysts who only differ in star analyst title? Answers to this question have implication for the role of star analyst award in the market. If there is no significant difference in the market reaction between the recommen- dations by the two groups of analysts, we may interpret that investors react mostly to the underlying quality of stock recommendations rather than the star analyst title regardless of thequality. Then, thestaranalystawardfacilitatesthepricediscoveryprocessbyidentifying better financial analysts and attracting investors to these analysts’ higher quality recommen- dations which duly reflect market fundamentals. In contrast, if significant difference exists, investors are reacting to the star title regardless of the quality of analysts’ recommendations. Then, the star analyst award is a new factor generating short-term excessive volatility and 1For price reaction to stock recommendations see Stickel (1995); Leone and Wu (2007); Loh and Stulz (2011);Balakrishnan,Schrand,andVashishtha(2011). ForprofitabilityofstockrecommendationsseeEmery and Li (2009); Fang and Yasuda (2014); Kucheev, Ruiz, and Sorensson (2016). 2 inefficiency in the market, the role of which has not been documented and identified in pre- vious literature. Such disturbing effect will be even larger in emerging financial markets, like the one in China, where analysts’ recommendations are distorted by conflict of interests (O’brien, Mcnichols, and Lin 2005; Agrawal and Chen 2008), where the market is dominated by individual investors who are not sophisticated enough to distinguish informative recom- mendations versus biased analyst opinions (Jiang, Lu, and Zhu 2014), and where short-sales are restricted (Xiong 2013). The goal of this paper is to tackle the identification challenges and quantify the impact of star analyst award on the market. The test bed for our analysis is the “Star Financial Analyst” ranking by New Fortune (N/F) magazine. This ranking is the most recognizable competition among sell-side financial analysts in China, and is very similar to ”All American Financial Analyst” in the U.S. Each year, N/F magazine ranks participating analysts according to votes from institutional investors, privatelynotifiesfinalistsaroundoneweekbeforetheceremony, andrevealsranked winners and finalists in a widely publicized ceremony in the end of November. From now on, the Saturday the week before ceremony is referred to as “notification day”, the week between notification and ceremony is referred to as ”notification week”, the day of public ceremony is referred to as ”ceremony day”. The winners will experience increasing media coverage, special treatment from listed companies, and a huge pay rise of over half million. In contrast, analysts who are right below the announcement cutoff (hereafter “nobodies”), i.e., neither winners nor finalists, remain unknown to the public.2 With the proprietary data on unpublished ranking and analysts’ personal information from N/F magazine, we are able to exploit the quasi-random assignment of analysts’ winning statuses right at the cutoffs and estimate the first quasi-experimental estimate of the effect of star title on the market reaction to stock recommendations. UndertheRDdesign,wecomparethemarketreactiontostocksrecommendedbyanalysts 2Thepubliccanbackoutwhoenterthecompetitionbutdonotmakeittothefinaliststagebycomparing the list of participating analysts and the list of announced finalists. 3 above and below the announcement cutoffs. Since announced analysts can be either winners or finalists, there are two potential cutoffs — cutoff of winner and cutoff of finalist. Which cutoff matters more to the investors is an empirical question, and we will examine both in this paper. To lower the probability that analysts change stock recommending strategy after they know the award results, we focus on the latest stocks recommended by an analyst before notification day or before ceremony day, and examine the market reaction to the stocks after these information events.3 Our results show that star title does matter in the market, regardless of the quality of the analyst. Right after the ceremony, investors look back and react more to previously stocks recommended by winners. The 2-day CAR on the first trading day after the ceremony to the latest stocks recommended by analysts just above the cutoff of winner is 0.66% more than those just below, almost the same size as the baseline mean. One possible channel for this increase is “attention shock” channel proposed by Barber and Odean (2008). Retail investorswhofacetimeandattentionlimitinpickingwhichstockstopurchasearenet-buyers of attention-grabbing stocks. The high publicity of the star analyst award may induce retail investors to search and purchase winners’ latest recommending stocks. The higher market reaction concentrates in stocks recommended by first-time winners and by female winners. In addition, we find that the average 2-day CAR during the notification week for stocks previously recommended by analysts above the cutoff of finalist is 0.5% more than those by analysts right below. However, the difference in 2-day CAR on the Monday after the ceremony flips to negative 0.6%, when the low-ranking finalists are revealed as losing the title ofwinner(“failedfinalists”). Apotentialchannelforthispatternofmarketreactiontostocks by failed finalist is “coordinating device” channel, empirically supported by Balakrishnan, Schrand, and Vashishtha (2011). In the notification week, brokerage houses and financial 3PaperslikeEmeryandLi(2009),FangandYasuda(2009),andFangandYasuda(2014)provideevidence that analysts change strategies of recommending stocks after they are awarded. Since different cutoffs are revealed through different information events, we choose different stock sampling and outcome windows for each cutoff. We will further explain our choices later. 4 analysts may notify their connected traders of their finalist status. The informed traders can then buy in stocks by finalists ahead of time and sell stocks recommended by failed analysts once the actual ranking is revealed. We provide suggestive evidence that the higher CAR in the notification week and the lower CAR after the ceremony of stocks recommended by finalists are mostly driven by stocks recommended by analysts in brokerages with higher mutual fund trading commission.4 The abnormal market reaction to winners and failed finalists of star analyst award completely reverses back to zero within six weeks from the ceremony. The abnormal reaction and the subsequent reversal confirms that the star analyst award indeed generates short-term excessive volatility and inefficiency in the market. We also show that investors continue to react more to star analysts’ stocks recommended within one year from the ceremony even after the initial over-reaction has already dissipated. Furthermore, the positive abnormal market reaction to winners’ recommended stocks in the longer term is not justified by winners’ higher ability, proxy by analyst’s annual forecast error. The paper directly adds to a large body of literature in finance and accounting on the relationship between the star title of analysts and the price reaction to their stock recom- mendations (Stickel 1995; Leone and Wu 2007; Loh and Stulz 2011; Balakrishnan, Schrand, and Vashishtha 2011), and the relationship between the star titles and the profitability of the recommendations (Emery and Li 2009; Fang and Yasuda 2014; Kucheev, Ruiz, and Sorensson 2016). Most papers consider winning award in a certain year as an annual treat- ment and compare various performance of star analysts with all other non-star analysts in the year after the award, controlling for as many observables as possible or using various fixed effects. Their underlying assumptions are that winning the award is exogenous and is not correlated with time-varying unobservables. In contrast, we bypass these assumptions 4 MutualfundsareprimaryinstitutionalinvestorsinChina(Firth,Lin,andZou2010). Tradingcommis- sions on average constitutes about half of a brokerage firm’s operating income. Mutual funds can thus exert substantial influence over brokerages and brokerages have incentive to curry favor their connected mutual funds (Gu, Li, and Yang 2012; Firth, Lin, Liu, and Xuan 2013). 5 and employ an RD design with a focus on the market reaction immediately after the award announcement to stocks recommended before the announcement. This paper thus provides the first quasi-experimental estimates of how star title affects short-term market reaction to stock recommendations. This paper contributes to the literature on the factors causing instability and inefficiency in the financial market. Biases in information transmission between investors and financial intermediaries can lead to bubbles and crises in the market. Na¨ıve investors may not rec- ognize distortions in analysts’ recommendations and follow whichever analysts with better titles. Meanwhile, sophisticated investors who can recognize the distortions and discount the recommendations may take advantage of the na¨ıve ones through speculation (Malmendier and Shantikumar 2007), which contributes to the instability of the market. The star analyst award induces speculative tradings and is a new factor amplifying the disturbing effect of biased information transmission, and causing excessive volatility and inefficiency. Our paper is the first to identify this role of star analyst award in the market. This paper also speaks to a broader class of economics literature on the effect of title in various contexts. In the school accountability literature, information on school ranking may affect children’s school choices and their performance in school (Hastings and Wein- stein 2008; Andrabi et al. 2007; Camargo et al. 2017; Mizala and Urquiola 2013). In the restaurant setting, information on restaurant ranking (Luca 2016), dish ranking (Cai et al. 2009) and hygiene report cards (Jin and Leslie 2003) can change customers’ patronizing be- haviors. In the academic setting, scientists selected to be Howard Hughes Medical Institute Investigator see an increase in the citation of their previous papers (Asoulay et al. 2013). However, the stock market is a different context than the above to study the title effect. First, investment decisions in the financial market are high-stake compared to consumption decisions in most other contexts. Second, the financial market is built upon beliefs. Investors may follow the recommendations by star analysts, not because they think the recommenda- tion matches fundamentals, but because they believe that other investors will do the same 6 (higher-order beliefs), and they’d better follow first and gain from later price momentum. Such “coordinating device” channel rarely exists in other contexts studying the title effect.5 Taken together, this paper quantifies the impact of the star analyst award on the market. Given the short-term overreaction to star titles and the potential speculative behaviors of sophisticated traders during the award period, the star analyst award is a potential new factor generating excessive volatility and inefficiency in the market. 2. BACKGROUND AND DATA 2.1. ”Star Financial Analyst” Competition The “Star Financial Analyst” ranking competition by N/F magazine is the earliest and the most recognizable competition among sell-side financial analysts in China. Starting from 2003, N/F magazine has ranked and publicized top financial analysts in each industry each year. The time line of the annual competition is summarized in Figure 1. Each year around August, large institutional investors, such as mutual fund managers managing over a certain amount of fund, register to be voters for the competition. In September, individual analysts (or teams) apply to be candidates for certain industries for that year. In mid- to late- October, registered institutional investors rank top five financial analysts (or teams) from a list of candidates for each industry.6 For a given voter and N/F industry, the analyst that he or she ranks first in the industry gets five vote counts, second gets four vote counts, ..., and fifth gets one vote count. All the vote counts are then weighted by the amount of 5 Balakrishnan,Schrand,andVashishtha(2011)provideevidencethatanalystrecommendationsinfluence traders’ higher-order beliefs, and that stock downgrades by all-star analysts appear to be coordinating event leading investors’ trading strategies to become common knowledge and induce the dot-com crash in 1999/2000. 6Inpractice,therearetwotypesofindustries,onewithmorethan20candidates,andtheotherwithfewer than 20 (the number of candidates in these industries is usually lower than 15). For the larger industries, voters rank top five analysts, and N/F magazine announces top five winners and top seven finalists. For smallerindustries,votersranktopthreeanalysts,andN/Fannouncestopthreewinnersandtopfivefinalists. Wefocusonthelargerindustriesinthispaper, becauseinvestorspaymoreattentiontotheseindustriesand star titles. 7 fund the voter is managing.7 N/F magazine sums up all the weighted vote counts that an analyst (or team) gets in a certain industry, and rank the analysts (or teams) by their total vote counts from top to bottom in each industry. In about one week before the ceremony, namely the notification week, N/F magazine privately informs analysts who are finalists without revealing the their actual rankings.8 In late November or early December, always a Friday or a Saturday afternoon, N/F magazine announces the actual ranking and total vote counts for top seven finalists to the public and gives physical award to the top five winners in a widely publicized award ceremony. In contrast, analysts who are right below the announcement cutoff remain unknown to the general public.9 From now on, we refer to top seven analysts as finalists, top five analysts as winners, sixth to seventh analysts as failed finalists, and analysts below seventh as nobodies. 2.2. Data Sources The first data comes from N/F magazine. N/F magazine provides us with proprietary data for top 15 financial analysts (or teams) in each N/F industry from 2005 to 2014. The data contains information on the year of the competition, the industry and brokerage firm the analyst (or team) works for, the vote count and ranking the analyst (or team) gets, as well as demographic information, such as the analyst’ name, gender, highest academic degree, and work history up till the point when he or she last submitted application to N/F. The second dataset we use is the China Stock Market Accounting Research (CSMAR) database, whichiswidelyusedinresearchontheChinesefinancialmarket. CSMARcontains almostallpubliclyavailableChineseanalystreports.10 Bothinstitutionalinvestorsandretail 7The voting scheme only changes slightly from year to year, and we use year fixed effects to control for these changes. 8Thegeneralretailinvestorsdonothaveaccesstothisinformation. Wesearchonmajorfinancialwebsites in China, and do not find public articles on the finalist name list for ”Best financial analyst” award before the day of ceremony during 2005-2014. The N/F magazine does not keep track of the exact day of notice, but they say the notification is usually within one week before the award ceremony. And the magazine specifically asks the notified brokerages and analysts not to give out the information of finalists. 9In theory, investors can back out analysts who are N/F candidates but are not finalists by comping the list of candidates and finalists. 1098% of the analysts in the final sample have recorded recommendations in CSMAR during the same 8 investors are able to gain access to the recommendation reports studied here. We obtain all stock recommendations on A share companies listed on the Shanghai Stock Exchange and the Shenzhen Stock Exchange from 2005 to 2014. Each stock recommendation consists of a unique report ID, names of the analysts, the brokerage firm the analysts work for, the date of issuance, stock ID, rating of the stock, and the expiration date of the rating. The rating is standardized to a five-point scale: strong sell=1, sell=2, neutral=3, buy=4, and strong buy=5. We also extract daily stock price, daily and monthly stock returns with reinvestment of cash dividends, quarterly market values and annual book values of listed companies from CSMAR. The third dataset is the work history of N/F candidates that we manually compiled from various online sources. We extract data from resumes of licensed analysts posted on the Security Association of China (SAC) (http://www.sac.net.cn/xxgs/cyryxxgs), home pages of brokerage firms, personal web pages of the analysts, and online resume sites for financial industry, such as Golden Compass (http://stock.sohu.com/s2011/jlp) and Ifeng Finance (http://star.finance.ifeng.com). These data are publicly available, and subject to verification from SAC, brokerage firms, analysts and the public. This information supplements the work history provided by N/F magazine. The forth dataset is Choice Financial Terminal (Choice). Choice compiles information from mandatory filings of brokerage firms and mutual fund companies in China as mandated by the China Securities Regulatory commissions (CSRC). We obtain the semi-annual/annual amount and composition of revenue for each brokerage house, the total amount of stock trading commissions payments to each brokerage, the distribution of the commissions among brokerages, and the stock holdings of each mutual fund. year of competition. And the probability of analysts having recommendations recorded in CSMAR changes smoothly across the cutoff of winner and finalist. 9 2.3. Sample Construction First, we split the N/F ranking data from team level to analyst level and CSMAR stock recommendationdatafromreportleveltoanalystlevel.11 WethenassigneachN/Fcandidate a unique ID using information on demographic and work history from N/F magazine and from online sources. We cannot distinguish two analysts who share the same name and who work in the same brokerage firm at the same time, so we assume them to be the same person following Cohen et al. (2010). With this unique analyst id, we can merge analysts’ N/F vote counts and rankings to their stock recommendations in CSMAR.12 Over 98% of N/F candidates in our sample are matched with their outcomes in the CMSAR. Next, we follow Loh and Stulz (2011) and Daniel et al. (1997) to calculate two-day cumulative abnormal return (CAR) for the recommended stocks. To avoid confounding factors brought by stock suspensions, we treat the value of CAR as missing from one day before to one day after the suspension period. We take the opposite sign of a stock’s CAR on a certain day if the rating in the recommendation is less than or equal to neutral.13 A positive CAR for a stock on a certain day implies that the performance of that stock cannot be replicated by a portfolio of stocks with the same market value, book-to-market ratio, and momentum on that day. In other words, the market responds abnormally more to this stock. We have three regression samples in this paper. The first one is constructed to study the immediate effect of star titles on market reaction at the cutoff of winner. We examine the two-day CAR on the first trading day (Monday) after the ceremony of the latest stocks recommended by the analysts within 1-30 days before the ceremony. Our second sample is used to identify the immediate market effect at the cutoff of finalist. Since finalists are privately informed of their status around one week before the ceremony, analysts above and below the cutoff of finalist may change recommendation behaviors discontinuously given the 11When splitting the N/F ranking data, we assign the same rank to all analysts in one N/F team, under the assumption that investors perceive analysts on the same N/F team as having equal ability. 12IfananalystparticipatesinmorethanoneN/Findustrycompetitioninacertainyear, weonlykeephis orhermainindustryinwhichtheyissuethemoststockrecommendationsduringtheyearofthecompetition. 13More detailed construction of CAR is in the appendix. 10