درباره نقش اطلاعات در تجدید نظرهای سهام توصیه شده

We examine the information transmission role of stock recommendation revisions by sell-side security analysts. Revisions are associated with economically insignificant mean price reactions and often piggyback on recent news, events, long-term momentum, and short-run contrarian return predictors, typically downgrading after bad news and upgrading after good news. However, the revisions are usually information-free for investors. The findings go against the long-standing view that recommendations are an important means by which analysts assimilate information into stock prices. They disagree with the view of policymakers that analysts’ stock picks materially impact stock prices.

In an informationally perfect market, stock prices reflect all available information, indicating to investors the expected return on their investments. In reality, information is rarely perfect, and this allows economic agents to improve information efficiency by profiting from costly information discovery and incorporating their information into security prices (Grossman, 1976 and Grossman, 1995; Grossman and Stiglitz, 1980).1 In this study, we focus on the information role of security analysts through their recommendations to buy, hold, and sell stocks (sometimes called stock picking). In the information role, if analysts believe markets are reasonably efficient then they revise their recommendations based on new price-to-value comparisons built from private information and the belief in their superior ability to create information by processing public information. A widely accepted view is that analysts are information agents and they convey negative information through revision downgrades (e.g., revising a buy to a hold) and positive information through revision upgrades (e.g., revising a hold to a buy). Studies show that stock prices fall over 4% at downgrades and rise over 3% at upgrades. In the information role view, such returns are evidence of better stock picking that rewards analysts’ reputations and yields career opportunities. Although the analyst information role is widely accepted, there are reasons to call into question the interpretation of revisions in the information role and the revisions evidence. Because revisions provide the same information to all watchful investors at the same time, they are unlikely to be of much value to any single investor. This suggests that revisions could be an inefficient method for brokerages to profit from their information discovery effort, given their other means for profiting on their valuable information. Second, the evidence suggests that the value of the information transmitted through revisions is inordinately large. Simple calculations show the yearly value of revision stock returns, per brokerage, tops $100 billion, which exceeds brokerage equity value. Third, over a third of the revisions are contrary to the measured returns (see Conrad et al., 2006). Fourth, extant research questions the importance of the information role by advancing other purposes for revisions, which we collectively call the marketing role. These include influencing brokerage–client relations through exchange of revisions for management information or other quid pro quos (Schipper, 1991; Francis and Philbrick, 1993); building reputation for stock picking in analyst rankings crafted by Institutional Investor (I/I) and the Wall Street Journal (WSJ) ( Stickel, 1992); promoting brokerage investment banking business ( Lin and McNichols, 1998; Michaely and Womack, 1999; Ljungqvist et al., 2006; Kolasinski and Kothari, 2008); and boosting brokerage trading revenue ( Jackson, 2005; Irvine et al., 2007). Both the information role and the marketing role are consistent with the fact that brokerages annually spend large sums for analysts’ research. This paper reports new evidence of stock-returns behavior around revision announcements that overturns prior revisions evidence. Prior findings often use daily or overnight stock returns to measure the value of analyst information transmission. We show that almost 80% of the revisions are in response to corporate events, which frequently release firm-specific information about earnings and investments a few hours before revisions are announced. Thus, the daily and overnight return measures for analyst information have a basic identification problem as they contain reaction to the events, making them prone to erroneous inferences. To avoid this identification problem, we measure revision returns using a narrow return intervals around daytime revision announcements, similar to the approach of Graham et al. (2006) for identifying daytime dividend announcement returns from other event returns. We find the mean 40 minutes revision announcement returns are economically unimportant (−0.03% for downgrades and 0.03% for upgrades). These results are robust to wider windows of one hour and two hours and they agree with growing evidence showing that stock prices react in minutes to new information (Dann et al., 1977; Barclay and Litzenberger, 1988; Kim et al., 1997; Busse and Green, 2002; Chordia et al., 2008). Moreover, we report revision pre-returns, from the day before and until the revision announcement, are economically large and agree with the revisions, on average (−3.7% before downgrades and 1.1% before upgrades). Our analysis suggests the pre-returns are triggered by followed firm events. Revision post-returns trend in the direction of the revisions on average, falling 65 basis points (bps) after downgrades and rising 47 bps after upgrades. However, up to two-thirds of the post-return can be explained by the pre-return, pre-events, and known predictors of long-term momentum and short-run return reversals. The 19–20 bps residual post-return is economically small and below round-trip transaction costs. Although the residual could reflect analyst information, it is also true that it could reflect additional predictors of future returns and changes in expected returns that are allied with the corporate news and events. A central question raised by our findings is as follows: Why is there so much news just before analysts announce their revisions? Perhaps analysts’ leak their information just before revising. Because the news is mostly about corporate operations, management is a likely source for analysts’ information. However, because of Reg FD, after October 2000 management is barred from selectively disclosing privileged information to analysts. Thus, if analysts systematically leak their information, then more revisions should follow corporate events before Reg FD than after. Yet, we find the opposite. Moreover, many revisions are contrary to the news. Events and news give analysts rich opportunities to apply their superior skills to process news into new information. However, although such processing can occur, we find revisions are typically information-free. A third possibility is that analysts, in pursuit of their careers, strategically piggyback revisions on events, returns, and future return predictors, to better align their revisions with recent and future returns. This can improve analyst stock picking reputation and spur trading, boosting brokerage revenues and analyst income, and reducing the chance of job loss. The accepted WSJ analyst rankings, for example, which rely partly on picked stock returns measured from before to after the revision day, raise rank for larger announcement returns and better future return prediction.2 Our findings are consistent with this piggybacking explanation. Thus, piggybacking on the news can explain the appearance of large price reactions before revision announcements. We report new revisions evidence showing analyst revisions are typically information-free, contrary to the information role. Analysts appear instead to piggyback revisions on news of pre-events, pre-returns, and future return predictors, whose impact the market has already assimilated within a matter of minutes. Thus, the stock market is informationally more efficient than previously believed. The finding may be important in related literatures in which revision returns are a measure of analyst information, or serve as prima facie evidence of widespread informational inefficiency. It raises concerns with the usefulness of stock-return-driven practitioner measures of analyst stock picking performance. Moreover, Securities and Exchange Commission (SEC) reforms, such as Reg FD, rely in part on the presumption that analyst opinions in the form of recommendations materially impact stock prices. We note, however, that our findings do not rule out the notion that analysts’ are information agents in other ways, as presaged by Grossman, 1976 and Grossman, 1995 and Grossman and Stiglitz (1980). The findings broaden understanding of analyst behavior and could attach greater importance to the marketing role than is now recognized. The rest of the paper proceeds as follows. Section 2 reviews the associated literature. Section 3 describes the sample and shows stock returns around revisions and their informativeness. Section 4 considers pre-returns and pre-events. Section 5 investigates possible sources of error in the announcement returns. Section 6 examines the post-returns. Section 7 concludes the paper.

Our findings show the revision of analyst recommendations to buy and sell stocks is typically information-free. They often quickly follow large stock prices reactions to corporate events and news, downgrades after negative news and upgrades after positive news, on average. Their announcement returns are small and their short-term post-returns drift mildly with the revisions. We examine if the pre-returns could reflect information implied by the revisions, perhaps because of measurement problems or systematic anticipation. But the evidence contradicts these explanations. In an obvious instance, over a third of the revisions are contrary to their large pre-returns. We examine the modest short-term returns, which might be registering analyst information with some delay. However, those post-returns tend to reflect common predictors of future returns, such as price–momentum indicators, PEAD predictors, and return reversal predictors, leaving less than 20 basis points of post-return unexplained, which is below round-trip transaction costs. While the post-return residuals could reflect investor reaction to analyst information, they could also reflect changes in expected future returns indicated by news of corporate earnings and investment opportunities and other long-term drift factors. The results show that analyst revisions piggyback on recent news and events, trending with short- and long-term momentum factors and contrarian with return reversal predictors, and their revisions are information-free. Revisions could be motivated by strategic career concerns, as better ranking of stock picking performance seems to both credit analyst with return performance and boost their pay and employment opportunities. Our findings differ markedly from prior studies which report significant and large return reactions to revisions. We suggest the difference arises because analysts revise recommendations quickly after corporate news. This piggybacking blurs daily and overnight return measures for analyst information with reaction to the news, and prior studies have not effectively separated the revision-specific returns from that news. Using intraday data to measure the stock returns around recommendation revisions allows separating the revision period return from the pre-revision period return, as well as from post-announcement momentum and return reversals. We note other concerns with the information view. The notion that analysts simultaneously transmit valuable information to several savvy investors with new recommendations appears to be an inefficient way for brokerages to profit from expensive information discovery. Brokerages have a variety of ways to profit from their valuable investment information. We note peculiarities with past measures of revision information value. For example, when annually aggregated per broker, the implied value appears to be unusually large, exceeding brokerage market value of equity. Moreover, past measures appear to be wrong for the large fraction of contrarian revisions. The new findings expand knowledge of analyst recommendation behavior and suggest the marketing role for revisions could be more important in future research. Among the implications that can be drawn from our results, we note first that they indicate the securities markets are informationally more efficient than previously believed. The findings are also of policy interest, as policymakers have relied on the view that analyst stock picking materially impacts stock prices. Another implication is that a need exists for a more comprehensive understanding of analysts’ economic role and behavior, both theoretically and empirically. The findings call for more evidence about the information role, the marketing role and the analyst revision decision process. More empirical evidence is needed concerning measures of analyst information based on long-term returns following revisions. Our findings suggest an identification problem could be present in those measures, due to confounding events and news.