The last few decades have seen an extensive effort devoted to studying the herding behaviors in financial markets. Researchers have documented the economic effects of herding among peer analysts (e.
g. Trueman 1994; Welch 2000), institutional investors (e.g. Grinblatt et al.
1995; Wermers 1999), and managers (e.g. Zwiebel 1995). Individuals have a tendency to conform to their group for either strategic incentives (Welch 1992; Sias 2004) or behavioral reasons (Scharfstein and Stein 1990; Grinblatt et al.
1995). However, the behavior of outliers—agents who significantly deviate from the crowd—is less explored.Are outliers just noise that rational economic agents should ignore, or do they impose influence on people’s decisions and behaviors? This paper aims to understand the real effect of outliers. Since an outlier opinion comes from an individual instead of a cohort of agents and the extremism of the view can depend on the nature of the topic, it is empirically challenging to compare individuals’ tendency to exert an extreme behavior and to assess its real consequences directly.
We take on this challenge and examine these questions in the context of analysts’ extreme forecasts. Specifically, we study whether and how an analyst’s outlier forecast shapes the opinion of his peers and influences other market participants such as investors and managers.We start our analysis by identifying the most optimistic forecast for a firm’s annual earnings as the outlier forecast. The difference between the most optimistic forecast and the consensus from peer analysts measures the extremism of an outlier forecast.
We document the effects of outlier opinions on various types of market participants. First, peer analysts’ forecasts become more optimistic following the arrival of an outlier forecast. Psychology literature has long established that people’s assessments on value and probability are subject to the influence of a simple, sometimes even irrelevant, reference point (“anchor”), despite their effort and intension to avoid such an influence (e.g.
, Tversky and Kahneman 1974; Simmons, Robyn, and Nelson 2010).In the context of forecast dynamics, an outlier forecast appears to serve as an anchor that shapes peer analysts’ judgement. Second, investors are subject to the influence of outlier opinions: both abnormal returns and trading volume surrounding the issuance of an outlier forecast are significantly higher, compared to a non-outlier forecast. Lastly, there is evidence that managers respond to outlier forecasts: the extent of extremism of an outlier forecast is positively related to the degree of earnings management.
More importantly, the effect of extreme optimism on earnings management is prominent only when an outlier forecast subsequently generates more optimistic forecasts by peer analysts. Our results are robust to alternative definitions of outlier, inclusion of a range of year-, firm-, firm × year fixed effects, and taking into account potential impact from concurrent news events.In the second part of the paper, we investigate why certain analysts would voice extremely optimistic forecasts relative to their peers. We consider three plausible sources: private information, investment banking incentive, and career incentives.
An outlier forecast can engender stronger reactions from peer analysts, investors, and managers simply because it is more informative. We show, however, that outlier forecasts are less accurate than non-outlier forecasts. This indicates that analysts issuing outlier forecasts do not necessarily possess superior private information to justify their extreme positions. We then explore the effect of the Regulation Fair Disclosure (Reg FD), a regulatory event that reduces the availability of private information to analysts by prohibiting public companies from selectively disclosing information to certain market participants.
If superior private information enables analysts to convey extremely optimistic views, then the extent of the optimism in outlier forecasts should decline after the Reg FD. Instead, we find that the magnitude of extreme optimism increases after the Reg FD. Taken together, these results indicate that superior private information is unlikely to be the main cause for analysts to voice extreme forecasts.Do analysts broadcast extremely optimistic opinions in order to generate investment banking business? To evaluate this potential explanation, we exploit another regulatory event, the 2002 Global Settlement (GS), which prohibits analysts from being compensated for generating investment banking business for their brokerage firms.
If extreme optimism is motivated by investment banking incentives, the magnitude of extreme optimism should decline after the GS. Instead, we find no significant change in the level of extreme optimism post GS. This suggests that investment banking incentive is unlikely to be the dominating reason for issuing outlier forecasts.Lastly, we consider whether career incentives are behind an analyst’s decision to issue outlier forecast.
We find that an outlier forecast exhibits following traits: it is more likely issued by an analyst who has covered the firm for a longer period of time, who was less accurate in the past, and who is affiliated with a smaller broker. An outlier forecast also tends to become extreme when there is a higher level of ex ante uncertainty surrounding a firm’s information environment. Intuitively, this cultivates a stronger personal motive for risk-taking and attention-seeking, as the downside for such an analyst to voice an extreme view appears to be small (Clarke and Subramanian 2006).Overall, these tests suggest that private information and conflicts of interest are unlikely to be the dominant factors behind outlier forecasts.
Instead, in a highly competitive industry with only the very few top performers reaping huge rewards, analysts’ career incentives, such as attention-seeking or risk-taking, are more likely to be the source for them to voice extremely optimistic opinions.Our paper contributes to the emerging literature studying the impact of financial analysts on investors and firms. On the positive side, researchers have shown that analysts help reduce information asymmetry (Brennan and Subrahmanyam 1995; Kelly and Ljungqvist 2012) and act as external monitors (Yu 2008; Chen, Harford, and Lin 2015; Irani and Oesch 2013). On the negative side, there is evidence that analysts impose pressure on managers and induce managerial short-termism (Graham, Harvey, and Rajgopal 2005; He and Tian 2013).
Most of the studies use the total number of analysts to measure the scope of coverage. Instead of the aggregate effect of analysts, we focus on the heterogeneity and dynamics within the analyst group. By examining how individuals with extreme opinions affect their peers, investors, and managers, we provide first-hand evidence on the real outcomes of outlier opinions.Our findings are also related to the literature on analyst optimism.
This strand of literature generally measures analyst optimism as the difference between the average of the forecasts by analysts as a group and the realized firm earnings, thus focuses on cross-firm comparisons of optimism. In contrast, our setting explores within-firm dynamics between the most extreme view and these from peer analysts, thereby allowing us to abstract from all firm-level factors and explore the interactions and behaviors within the same analyst cohort. By documenting the effect of an outlier forecast on the subsequent forecast optimism of peer analysts, we identify a different source for the overall analyst optimism. In particular, forecast optimism of the entire cohort of analysts can be driven by an individual analyst’s personal motives, in addition to the well-documented investment banking incentive of brokerage houses (Lin and McNichols, 1998; Michaely and Womack, 1999).
The rest of the paper proceeds as follows. Section 2 describes the empirical framework, hypotheses, data sources and sample construction. Sections 3 and 4 present the empirical results. Section 5 provides discussions of additional tests of causality and robustness.
Section 6 concludes the paper. Variable definitions and constructions are in the Appendix.
Empirical Framework And Data
The Merriam-Webster Dictionary defines “outlier” as an observation that is “markedly different in value from the others of the sample”. An outlier opinion thus exhibits two essential traits: being markedly distinct from the others and coming from an individual instead of a cohort of agents.
Using earnings forecasts to evaluate the effect of outlier opinions offers several advantages. First, unlike corporate events such as security issuances and mergers and acquisitions, in which the timing of the event is endogenous and individuals who voice their opinions can be sporadic and random, corporate earnings disclosure is mandated on a regular basis, and analyst following tends to be stable. Second, there is a clean and systematical way to define and identifyoutlier opinions: individual forecasts that are most distinct from the group consensus. Lastly, the extent to which a forecast deviates from the group’s view is measurable and comparable across firms.
In an attempt to capture the two traits of an outlier opinion, we identify as an “outlier forecast” if an earnings forecast deviates the most from the consensus forecasts by fellow analysts. To better reflect the extremism nature of an outlier, we focus on the most optimistic forecast. Accordingly, we label an analyst issuing the outlier forecast as the “outlier analyst”, and as a “peer analyst” if an analyst covers the same firm in the same year as the outlier analyst does.Besides the dummy variable for outlier forecast, we construct “Extreme Optimism”, a firm-year level variable defined as the difference between the outlier forecast and the consensus peer forecast, which is the average of forecasts issued by other analysts covering the same firm in the same year.
To ensure the comparability across firms with distinct characteristics such as size and growth opportunities, we then scale this difference by the stock price of the previous calendar year. By construction, “Extreme Optimism” quantifies the extent to which an outlier’s opinion differs from that of the group. In this respect, this variable stands in sharp contrast to the traditional measure for analyst optimism, which is typically calculated as the difference between the average of forecasts of all analysts and the realized firm earnings (e.g.
, Bradshaw et al. 2001; Drake and Myers 2011).Both the dummy for outlier forecast and “Extreme Optimism” focus on the optimistic side of analyst opinions. This is because given the incentive structure, an analyst is reluctant to voice extremely pessimistic forecasts, as he can always opt for the alternative—terminating his coverage on firms that he views negatively (McNichols and O’Brien 1997; Hayes 1998).
Put differently, the observed pessimistic forecast less likely reflects the actual degree of extreme pessimism, thus introduces greater biases than the observed optimistic ones.By selecting the forecast that deviates the most from the consensus forecasts, we ensure one outlier forecast per firm per year. This setting offers an opportunity to study within-firm dynamics between the outlier analyst and peer analysts, thereby allowing us to control for all cross-firm-level factors and focus on the direct interactions and behaviors within the same analyst group. We do, however, acknowledge that there is no general consensus on how an outlier should be defined, and in particular, how extreme an opinion should be in order to be classified as an outlier.
As such, an outlier forecast may not occur for each firm each year. In the robustness section, we explore several alternative ways to identify outlier forecasts.