Zahn Bozanic The Ohio State University Fisher College of Business

advertisement
Analyst Spin
Zahn Bozanic
The Ohio State University
Fisher College of Business
bozanic.1@fisher.osu.edu
Jing Chen
State University of New York at Buffalo
School of Management
jchen229@buffalo.edu
Xuan Huang
California State University at Long Beach
College of Business Administration
Xuan.Huang@csulb.edu
Michael J. Jung*
New York University
Stern School of Business
mjung@stern.nyu.edu
May 2016
* Corresponding author, NYU Stern School of Business, 44 West 4th Street, Suite 10-82, New York, NY
10012-1126, Tel. 212-998-0193. We thank Orie Barron, Jessica Halenda, Paul Fischer, Oded Rozenbaum,
Maya Thevenot, Brett Trueman, Yong Yu, and workshop participants at Boston University and New
York University for helpful comments.
Analyst Spin
Abstract
We examine a phenomenon in which sell-side equity analysts interpret earnings news in a
direction that appears to contradict conventional wisdom. We conjecture that an analyst “spins” a
firm’s earnings news to persuade investors in favor of the analyst’s view about the firm and its
stock, and that this activity has implications for the analyst’s career. Using a large sample of
analyst reports, we find that 11% of reports convey spin, where analysts place a positive spin on
earnings shortfalls more often than a negative spin on earnings beats by a ratio of roughly three
to one. We provide evidence that analyst spin relates to incentives for career advancement, as
analysts who spin tend to have greater credibility but work for less prestigious brokerage houses.
Indeed, analysts who spin are more likely to move up to a top-tier brokerage house in the
following year and less likely to exit the profession. Corroborating these career advancement
results, analysts who spin appear to be correct in both short- and long-term stock predictions.
Our evidence suggests that analyst spin reflects credible interpretations of firm news.
1. Introduction
We examine a phenomenon in which sell-side equity analysts interpret earnings news in a
direction that appears to contradict conventional wisdom. These instances, which we term
“analyst spin”, are important to institutional investors because they represent possible
opportunities to earn larger payoffs from an analyst’s stock call that at first may appear
unconventional but later turns out to be correct.1 According to the Merriam-Webster dictionary,
spin is “the activity of trying to control the way something (such as an important event) is
described to the public in order to influence what people think about it.” While the idea of spin
typically has a negative connotation in political media, we do not suggest that analyst spin is
negative in any way. Rather, we conjecture that under certain conditions, an analyst “spins” a
firm’s earnings news to persuade investors in favor of the analyst’s view about the firm and its
stock, and that this activity has implications for the analyst’s career.
To provide some context for analyst spin, consider the following analyst report from
RBC Capital Markets published on April 28, 2005 for semiconductor firm LSI Logic
Corporation. The firm reported 2005 first quarter revenue of $450 million and earnings-per-share
(EPS) of $0.06, beating the RBC analyst’s estimates of $407.5 million and $0.02 per share,
respectively. The analyst then increased his forecasts for revenue and EPS for the next seven
quarters and raised his target price from $5 to $6. Despite the firm exceeding the analyst’s own
estimates for the reported quarter and the analyst revising estimates upward for the next two
years, the analyst expressed his opinion with a pessimistic (or negative) report title: “Engenio
Weak; Competition at IBM Remains a Concern.” 2 This example illustrates one of many
instances in which conventional wisdom suggests that the analyst should be positive about the
Of course, it is equally important for institutional investors to try to avoid larger losses if an analyst’s
unconventional stock call turns out to be incorrect.
2
Engenio was a division of LSI Logic Corporation that was acquired in 2000 and sold in 2011.
1
1
firm and its stock based on the reported EPS and upward revisions in expectations of the firm’s
future performance. However, the inconsistency between the firm’s apparent earnings news and
the analyst’s opinion of the news leaves existing and potential investors in a peculiar position
with respect to how they should act on the stock.
Our interest in examining analyst spin is motivated by: 1) many of our own observations
of analysts interpreting positive earnings surprises negatively (and vice versa) and 2) anecdotes
of institutional investors rewarding analysts who convey unconventional opinions. Each year,
institutional investors vote in Institutional Investor magazine’s rankings of the best sell-side
equity analysts and often comment that they voted for analysts who made unconventional stock
calls. 3 As a result, we believe that analyst spin is an interesting and important institutional
phenomenon that may be related to analyst career concerns and has not been previously
examined in the analyst literature.
In addition, we believe that this setting can be used to test theories about analyst
credibility, investors’ differential interpretation of firm news, and the empirical predictions of
these theoretical models for market reactions. Theoretical work by Fischer and Stocken (2010)
motivates our prediction that when the precision of firm news is low, resulting in a wider range
of possible interpretations about whether the news is good or bad, an analyst with high credibility
may spin the news in a particular direction to inform investors. If this is the case, and analysts
are recognized and rewarded for their unconventional advice, then we should observe better
career outcomes for such analysts. Alternatively, if analyst spin represents “cheap talk”
Many anecdotes each year from Institutional Investor’s annual ranking of the All-American Research Team
indicate that institutional investors value analysts who convey unconventional views. For example, in the 2009
rankings, the #1 Electric Utilities analyst was described by a portfolio manager as “frequently out in front of the
herd on calling a change in trends.” Similarly, in the 2008 rankings, the #3 Internet analyst was described as “not
afraid to go against the consensus.” In 2007, the #3 Pharmaceuticals/Major was described as “not afraid to stick her
neck out in the quest for upside.” In 2005, an investor stated about the #2 Telecom Equipment/Wireless analyst,
“His advice seemed crazy, but it worked well for us.”
3
2
(Crawford and Sobel 1982; Farrell and Rabin 1996), then analysts who spin should not
experience better career outcomes and may possibly experience negative career outcomes.
Regarding investors’ reactions to analyst spin, analytical work by Kim and Verrecchia (1991,
1994) and empirical evidence from Bamber, Barron, and Stober (1997) suggest that if analyst
spin leads to, or is a reflection of, greater disagreement among market participants in interpreting
public signals, then there will be greater price changes, trading volume, and bid-ask spreads on
the days in which analyst reports convey spin relative to days in which analyst reports do not
convey spin.
To empirically capture analyst spin, we use a large sample of analyst reports and focus on
the quantitative earnings news contained in each report and the qualitative tone of the report’s
title. For reports that review a firm’s recent earnings announcement (38% of our sample), we
code the earnings news to be positive (negative) if the reported quarter’s EPS is above (below)
the analyst’s estimate, and for all other reports, we code the earnings news to be positive
(negative) if the analyst revised upward (downward) the EPS forecast for the next quarter. Then
using a dictionary customized for analyst vernacular, we classify a report to contain positive spin
when the tone of the title is positive while the direction of the earnings news is negative, and vice
versa.
We focus on the report’s title rather the entire report for three reasons. First, while we do
not imply that the report’s body is uninformative, we argue that an analyst writes a title to be the
first and primary text to influence investors’ beliefs. Our argument is grounded in research in
psychology (Asch 1946; Murdock 1960), behavioral economics (Rabin and Schrag 1999), and
accounting (Bowen, Davis, and Matsumoto 2005; Files, Swanson, and Tse 2009; Huang,
Nekrasov, and Teoh 2013), which suggests that the title alone is sufficient to shape investors’
3
beliefs.4 For example, Paiva, Lima, and Paiva (2012) find that short titles presenting results or
conclusions predict higher citations in medical research journals. Second, the investor inattention
literature (Hirshleifer and Teoh 2003; Hirshleifer, Lim, and Teoh 2009) suggests that a timeconstrained investor who wants to simply assess an analyst’s overall opinion of a firm and its
stock is more likely to do so based on a one-sentence “takeaway” displayed at the top of the
report rather than on several hundred sentences in the body of a report. Third, retrieving and
coding a large sample of analyst report titles is much less costly than doing so for the full content
of each report, allowing us to increase our sample size substantially. For these reasons, we view
an analyst’s report title to be a sufficient summary statistic that captures the analyst’s overall
qualitative opinion about the firm and its stock.
Descriptive evidence reveals that analysts spin earnings news quite often. Using a sample
of 396,188 analyst reports written by 2,787 analysts from 202 brokerage houses covering 2,005
firms between 2005 and 2010, we find that 43,829 of the reports contain spin. That is,
approximately 11% of analyst reports have a title whose tone is inconsistent with the direction of
the earnings news. Analysts place a positive spin on earnings shortfalls more often than a
negative spin on earnings beats by a ratio of roughly three to one. Analysts from the top-10
brokerages (ranked by number of reports in our sample) tend to spin proportionately less often
than analysts from smaller brokerages, and the industries in which analysts tend to spin more
frequently are technology, financial, and energy.
Beyond documenting the prevalence of analyst spin, we attempt to gain insights into how
analyst spin relates to constructs commonly used in prior studies on analysts: stock ratings,
4
The importance of titles is widely recognized in the news media, social arena, and the popular literature.
Newspaper editors and reporters strive for punchy headlines. Publishers and movie producers pressure authors and
directors to change titles to increase commercial success. Consider the outcome if George Orwell had not changed
The Last Man in Europe to 1984, or if Kathryn Bigelow had gone with God and Country instead of Zero Dark
Thirty.
4
forecast errors, and forecast dispersion. Analysts who spin negative earnings news in a positive
light tend to have a positive rating (Buy or Strong Buy) on a firm’s stock, and analysts who spin
positive earnings news in a negative light tend to have a negative rating (Hold, Sell or Strong
Sell) on a firm’s stock. This correlation suggests that analysts who spin do so to “defend” their
particular view of a firm and their rating on the stock. In terms of forecast error, for analyst
reports published immediately after a firm’s earnings announcement, the average absolute
forecast error (reported EPS less the analyst’s estimate) is greater for the analysts who spin
compared to analysts who do not spin. For all analyst reports, the average dispersion in analyst
forecasts (standard deviation of all analysts’ estimates) is smaller when analysts who spin are
excluded from the analysis. These findings indicate that analysts who spin earnings news tend to
have forecasts that are farther from the eventual reported EPS and consensus EPS forecast. This
evidence also suggests one explanation for analyst spin: standing out from the crowd. That is, an
analyst spins earnings news (or any firm news) to persuade investors in favor of the analyst’s
contrarian view (rating) of the firm and its stock.
We next turn to an examination of the determinants of analyst spin. Drawing on the
analytical work in Fischer and Stocken (2010), we hypothesize that analyst credibility is
positively associated with spin. Using analyst experience, breadth and depth of industry
knowledge, and coverage effort as proxies for analyst credibility, we find evidence consistent
with our hypothesis. We also find that analysts who work for less prestigious brokerages and
who cover smaller and lower performing firms tend to spin earnings news, which suggests that
experienced and knowledgeable analysts who work for less prestigious brokerages may be
attempting to gain recognition among institutional investors by arguing contrarian views about
smaller firms.
5
We further examine this explanation by testing our second prediction that analyst spin is
related to career outcomes. Recall that we identify analyst spin when an analyst espouses a view
that runs counter to apparent earnings news, which may ultimately hurt or help an analyst’s
career. On the one hand, if analyst spin misleads investors about a firm’s prospects and future
value, then we would expect career outcomes to be worse for analysts who spin. On the other
hand, if analyst spin helps investors when there is greater uncertainty about the firm’s future,
then we would expect career outcomes to be better for analysts who spin. We examine all
analysts who changed employers during our sample period and find evidence consistent with the
latter scenario, i.e., analysts who spin are more likely to move up to a top-tier brokerage in the
following year and also less likely to exit the profession. We corroborate these results by
showing that analysts who spin tend to be correct in their short- and long-term stock calls. In
contrast, we do not find that analysts who spin are more likely to decline in their careers.
In our final set of analyses, we test our third prediction that analyst spin is reflective of
greater investor disagreement about firm news and is associated with a greater market reaction.
Specifically, we predict and find higher return volatility, trading volume, and bid-ask spreads on
the days when analysts publish reports with spin relative to days when analysts publish reports
without spin. These results are consistent with the theoretical work in Kim and Verrecchia (1991,
1994) that suggests analyst spin leads to, or is a reflection of, greater disagreement among
market participants in interpreting public signals with less precision.
The collective findings in this study contribute to several literatures. First, while the
literature on sell-side equity analysts is voluminous (e.g., Barron, Kim, Lim, and Stevens 1998;
Bradshaw 2011 and the papers cited therein), our study extends this literature by documenting an
important institutional phenomenon in which analysts interpret earnings news differently than
6
expected, which can be a reflection of their experience, knowledge, and effort, rather than
opportunism (e.g., Lin and McNichols 1998). Second, we highlight circumstances under which
an analyst would choose to convey seemingly inconsistent information signals in their reports.
As such, our results highlight a new source of investor disagreement in interpreting firms’
earnings and non-earnings news, as well as empirical evidence for the models in Kim and
Verrecchia (1991, 1994). Third, our focus on the analyst report title as an analyst’s qualitative
summary opinion, in conjunction with commonly-examined, quantitative summary assessments
of the report, contributes to the textual analysis literature (Asquith, Mikhail, and Au 2005; De
Franco, Vasvari, Vyas, and Wittenberg Moerman 2014; Huang, Zang, and Zheng 2014; De
Franco, Hope, Vyas, and Zhou 2015) by highlighting the determinants and market consequences
of an incongruence between the qualitative and quantitative report summary.
2. Literature Review and Hypothesis Development
2.1 Literature on Sell-Side Equity Analysts
Sell-side analysts have been studied extensively by academic researchers over the past
several decades. Facilitated by data availability, early research focused almost exclusively on
earnings forecasts contained in analyst reports (Bradshaw 2011). Compared to time-series
models, analyst forecasts were found to be more accurate (e.g., Fried and Givoly 1982; Brown,
Griffin, Hagerman, and Zmijewski 1987; O’Brien 1988; Clement 1999). Later research
examined the relation between analysts’ forecasting activities and stock prices (e.g., Philbrick
and Ricks 1991), the value-relevance of stock recommendations and price targets (Womack
1996; Brav and Lehavy 2003), and the incentives that influence potential analyst opportunism
(e.g., Francis and Philbrick 1993; McNichols and O’Brien 1997; Lin and McNichols 1998).
7
More recently, a few studies have looked beyond the quantitative measures in analyst
reports to investigate the qualitative content of the reports. Asquith et al. (2005) manually code
analysts’ stated arguments in their reports and find that the content is incrementally informative
to earnings forecasts, stock recommendations, and target prices. Using a naïve Bayes machine
learning approach, Huang et al. (2014) extract the textual opinions from analyst reports and find
that the market reacts to the textual opinions. De Franco et al. (2015) study the readability of
analyst reports and find that readability is positively associated with analyst ability and stock
trading volume. Our study extends this stream of literature by examining an institutionally
important phenomenon in which analysts appear to disagree with the direction of a firm’s
earnings news, and hence, recommend to investors to interpret the news in the opposite direction.
Our study focuses on answering the question of whether analyst spin is a reflection of analysts’
superior interpretation of firm news or represents mere “cheap talk” (Crawford and Sobel 1982;
Farrell and Rabin 1996).
2.2 Literature on the Influence of Titles
While we do not imply that the body of an analyst report is uninformative, we argue that
an analyst writes a report title to be the first and primary text to influence investors’ beliefs.5 A
document title is a natural focus of attention for readers as it is what they first see, interpret, and
internalize. A title is both front and summary matter and is therefore expected to convey a key
message. Mahoney (1991) states that “…headlines drive an idea, instead of simply identifying a
subject.” In the context of sell-side equity analysts, each research report they write provides them
the opportunity to emphasis an investment thesis or opinion with a report title. For example, on
January 4, 2007, the analyst for Wedbush Morgan Securities published a report on Intervoice, a
5
Our conversations with several sell-side equity analysts indicate that the lead analyst writes each report title, rather
than the junior analyst or an editor.
8
speech technology company, entitled: “Strong Backlog Growth Indicates Improving Business
Momentum; Reiterate Strong Buy and Raising Target from $10.50 to $11.00.” As such, we
conjecture that analysts purposely use report titles to guide investors’ interpretation or
impression of firm news events, and perhaps, reduce the need to read the entire report.
Research from psychology and behavioral economics helps to explain why analysts use
report titles to shape investors’ beliefs. The psychology literature has documented the importance
of primacy bias, or the serial position effect, in which people heavily weight and subsequently
remember most easily the first item in an ordered list (e.g., Asch 1946; Murdock 1960). Day
(1994) states that ‘‘first impressions are strong impressions’’ and Paiva, Lima, and Paiva (2012)
find that short titles presenting results or conclusions predict higher citations in medical research
journals. The behavioral economics literature has made similar arguments on the importance of
first impressions. For example, Rabin and Schrag (1999) propose that information received first
can potentially bias the interpretation of subsequent information.
In the accounting literature, studies have examined whether managers’ choice of words in
headlines influence investors’ reactions to disclosures. Bowen, Davis, and Matsumoto (2005)
find that managers can influence investors’ perceptions of earnings based on where earnings
metrics are placed in a press release (headline, body, or tables). Files, Swanson, and Tse (2009)
investigate earnings restatements and find that when managers place information about the
restatements in the headline of the press release, it generates a stronger reaction than when they
place the information in the body or footnote in the press release. Huang et al. (2013) examine
earnings press releases and find that managers that place greater salient information in the
headline experience a stronger earnings announcement stock reaction and a weaker post-earnings
9
announcement reversal. In this study, we draw on the aforementioned theories and empirical
findings to use the title of a report as the analyst’s overall qualitative summary opinion of a firm.
2.3 Hypothesis Development
We draw motivation from the analytic work of Fischer and Stocken (2010), who model
an information acquisition and communication game between an analyst and a decision-maker
(i.e., investor). 6 Fischer and Stocken (2010) provide several predictions about an analyst’s
reporting behavior as a function of the parameters featured in their model. Of their predictions,
one is particularly relevant to our inquiry: when the difference between state payoffs is large and
the precision of public information is low, an analyst uses more effort to gather, process, and
interpret public information. This scenario closely relates to our examination of analyst spin for
several reasons.
First, empirical evidence (Loh and Stulz 2011) suggests that the “analyst calls” that are
potentially most profitable (i.e., have the largest state payoff difference) for investors are those
that diverge from the conventional view or prevailing consensus. These stock calls are more
likely to arise when the precision of firm news is low and the range of possible interpretations of
the news is wide. Second, unconventional opinions are also potentially most beneficial for
analysts who seek to stand out from the crowd to gain visibility and recognition in the market.
However, standing out is not costless insofar as their stock calls that are realized to be
unprofitable ex post (i.e., “bad calls”) have the opposite effect (i.e., a loss of market influence for
the analyst). Third, as Fischer and Stocken (2010) demonstrate, an analyst’s ability to
Key features of their model that are relevant for our setting are: 1) the analyst’s credibility, 2) the precision of the
analyst’s private information, 3) the precision of publicly observed information (i.e., firm news), 4) the analyst’s
effort in interpreting the public information, 5) the analyst’s report to the investor, 6) the investor’s responsiveness
to the report, and 7) different state payoffs. These features map reasonably well into our institutional setting in
which there is a large sample of analysts, with varying degrees of credibility, who continually gather and interpret a
wide range of information for an even larger sample of firms in order to publish research reports intended for
sophisticated, institutional investors who may or may not react to the report.
6
10
communicate credibly (prior to the call) is an important consideration. Analysts able to
communicate credibly are more likely to persuade investors to follow an unconventional stock
call. Therefore, analysts who have yet to build sufficient credibility among investors will likely
not try to spin earnings news. Given the preceding discussion, our first hypothesis is as follows:
H1: Analysts with greater credibility are more likely to spin earnings news than
analysts with less credibility.
Prior work that documents most analysts herd in their forecasts and recommendations
(Trueman 1994; Clement and Tse 2005; Hong, Kubik, and Solomon 2000; Welch 2000), coupled
with our finding that analysts who spin earnings news tend to have forecasts that diverge from
the consensus, suggests that analysts who spin gain visibility (i.e., greater recognition) by
standing out from the herd. We conjecture that increased visibility or recognition for an analyst
leads to substantial benefits for the analyst and the employing brokerage house, including
increased trading commissions and investment banking business for the brokerage house
(Goldstein, Irvine, Kandel, and Wiener 2009), and positive career outcomes for analysts.
However, given that some of these benefits are unobservable (e.g., analyst compensation), we
focus on testing whether analyst spin is related to career advancement. Our second hypothesis is
as follows:
H2: Analyst spin is associated with analyst career advancement.
If analyst spin reflects “cheap talk” or some form of analyst opportunism that
sophisticated investors largely discount, then analyst reports with spin should be associated with
lower market reactions than reports without spin. Alternatively, considering that our definition of
analyst spin involves an inconsistency between an analyst’s qualitative summary assessment of
11
firm news and the analyst’s quantitative summary measures, we expect there to be inherently
increased uncertainty for investors who read the report. Thus, if analyst spin represents greater
investor disagreement following firm news, then the market reaction to analyst reports with spin
should be greater relative to analyst reports without spin. This prediction is supported by
analytical models of trading as a function of investor disagreement (Kim and Verrecchia 1991,
1994) and by empirical evidence (Bamber et al. 1997). Our third hypothesis is as follows:
H3: Analyst reports with spin are associated with a greater market reaction relative
to analyst reports without spin.
3. Data, Sample, and Descriptive Evidence
3.1 Sample Composition
We obtain data on research reports written by sell-side equity analysts from Thomson
One Banker. The data include the name of the firm for which the report is written, the analyst
who wrote the report, the publication date, the employing brokerage house, and the title of the
report. We collect this data for any research report published between January 1, 2005 and
December 31, 2010, excluding reports that are not from sell-side analysts (which Thomson calls
non-broker research). 7 Using the name of the analyst and firm on each report, we manually
match the data to the I/B/E/S detail history estimates and recommendations databases to obtain
analysts’ earnings-per-share (EPS) forecasts, actual reported EPS, and stock recommendations.
Our sample includes 396,188 reports written by 2,787 analysts from 202 brokerage houses
covering 2,005 firms.
3.2 Identification of Analyst Spin
7
The reports are downloadable as a Portable Document Format (PDF) file. However, the Thomson One Banker
interface allows users to download a maximum of 50 reports at a time, which limits our ability to collect the entire
history of reports.
12
To identify cases of analyst spin, we look for reports in which the direction of the
analyst’s qualitative summary opinion of the firm is not consistent with the analyst’s quantitative
summary assessment of the firm’s earnings. This identification procedure involves two steps.
First, we identify reports that contain quantifiable earnings surprises or forecast revisions. Of the
396,188 analyst reports in our sample, there are 150,686 reports (38%) that are considered
reviews of earnings announcements because they are published on the same day or one day after
a firm’s earnings announcement. For these reports, we compute the analyst’s forecast error (FE),
defined as the actual reported EPS minus the analyst’s latest forecasted EPS for the reported
quarter. In addition, for all reports (whether they are reviews of earnings announcements or not),
we compute revisions (if any) to the analyst’s forecasted EPS for next quarter (NQESTCHG).
Thus, we can identify any report in which there is either a non-zero forecast error or change in
next quarter’s EPS estimate. We then classify an analyst report as having a positive quantitative
earnings news effect if either the analyst’s forecast error is positive or the analyst’s revision to
next quarter’s EPS estimate is positive (i.e., FE>0 or NQESTCHG>0). Similarly, the quantitative
effect is negative if either the forecast error or revision is negative.8
In the second step of our identification procedure, we assess the analyst’s qualitative
summary opinion of the firm by identifying the number of positive and negative words used by
the analyst in the report’s title. Because words can be considered positive or negative depending
on the context in which they are used and by the type of user (Henry 2006, 2008), we perform a
8
As a robustness check to our measure of the direction earnings news, we consider cases in which a firm exceeded
an analyst’s estimate for the reported quarter (FE>0) while the analyst reduced his or her estimate for the next
quarter (NQESTCHG<0) and vice versa (FE<0 and NQESTCHG>0). These cases represent 3.9% of our overall
sample of 396,188 analyst reports. We argue that the precision of earnings information in these cases is especially
low, and that an analyst’s qualitative summary opinion becomes even more important in shaping investors’
interpretation of the firm’s news. For this reason, we believe these cases should be included in our sample of reports
with spin. However, if we were to exclude these cases, our sample of analyst reports with positive or negative spin
would be reduced by 35% while the inferences from the results presented in Tables 2 through 6 would remain
unaffected.
13
detailed examination of the most frequent phrases used by analysts in their report titles. We
compile the most frequent two-, three-, and four-word combinations, read the context in which
these phrases are used, and create a positive and negative word list customized for sell-side
analyst vernacular. For example, we find that one of the most frequently used words in report
titles is “raising,” used 21,726 times in positive phrases such as “raising price target…,” “raising
estimates…,” and “raising rating…” In contrast, we do not find any instances of the word used in
negative phrases such as “raising concern…,” raising worry…,” or “raising risk…” As a result,
we classify “raising” and its variants (raise, raised, and raises) as positive words. Repeating this
detailed analysis for thousands of commonly-used words and phrases in report titles leads us to
create a custom analyst-focused list of 452 positive and 300 negative words, which are shown in
Appendix A. 9 Using our custom word list, we classify an analyst’s qualitative summary
interpretation of the news about the firm as positive if the analyst uses more positive words than
negative words in the report’s title, and vice versa.10
After measuring the analyst’s qualitative summary opinion of firm news and the
quantitative summary assessment that the news has on the analyst’s earnings forecasts, we
classify analyst reports to have spin if the qualitative opinion is positive while the quantitative
assessment is negative or if the qualitative opinion is negative while the quantitative assessment
is positive. We consider the first scenario to be “positive spin” and the second scenario to be
“negative spin.” To illustrate, we include in Appendix B several examples of reports that we
have classified as having spin. Panel A shows four reports of positive spin in which either the
Additional examples of words frequently used by analysts that we classify as positive are “strong,” “solid,”
“upside,” “beat,” “outperform,” and “highlights.” Examples of words frequently used by analysts that we classify as
negative are “lowering,” “weak,” “below,” “tough,” “disappointing,” and “concerns.” Analysts also tend to use
abbreviations such as “ow” and “uw,” which stand for “overweight” and “underweight,” respectively. We
emphasize that words are classified as positive or negative based on their usage in phrases and not in isolation.
10
Our analyst context-specific word list has a sizeable overlap with commonly-used sentiment or tonal dictionaries.
For example, over half of our positive words can be found in Loughran and McDonald’s (2011) positive tone
dictionary and over one quarter of our negative words are in their negative tone dictionary.
9
14
analyst’s earnings forecast error or revision is negative while the report title expresses a positive
opinion. Panel B shows four reports of negative spin in which either the analyst’s earnings
forecast error or revision is positive while the report title expresses a negative opinion.
Descriptive statistics of our entire sample of analyst report titles are provided in Table 1.
Panel A shows the number of reports by year and calendar quarter. The number of reports has
declined steadily from 75,190 in 2005 to 57,717 in 2010, consistent with the documented recent
decline in the sell-side analyst profession (Mola, Rau, and Khorana 2013). Panels B and C show
the number and percentage of analyst reports that contain positive and negative spin,
respectively. We find that approximately 8% of analyst reports contain positive spin and 3%
contain negative spin. Panel D shows the number and percentage of analyst reports from the top
40 brokerage houses, ranked by the number of reports contained in the Thomson One Banker
database. More than half of the reports in our sample come from the top ten brokerage houses,
with J.P. Morgan, Credit Suisse, Deutsche Bank and Morgan Stanley each accounting for more
than 5% of all reports. Also shown is the number and percentage of analyst reports that contain
spin (positive or negative) from each brokerage house. Compared to the overall average of
11.1% of analyst reports with spin, a slightly smaller percentage of reports from the top five
brokerage houses contain spin, which suggests that analysts who spin earnings news either
positively or negatively tend to work for brokerages houses below the top five. Panel E shows
the number and percentage of analyst reports by industry for firms with non-missing two-digit
SIC codes. The highest number of reports is written about firms in the chemical, business
services, electronic, and industrial industries, and a higher percentage of reports for firms in the
oil, building, and financial industries contain analyst spin.11 Lastly, Panel F shows the number of
11
In the regression analyses discussed in Sections 4 and 5, we include year fixed effects, industry fixed effects, or
both, to control for temporal or industry trends that may be associated with analyst spin.
15
unique analysts in our sample (2,787), categorized by whether they have ever written a report
with positive or negative spin or never written such a report, along with the number of total
reports and reports with spin they have written. The results indicate that most of the analysts
(58%) have written both a report with positive spin and negative spin, while a smaller number of
analysts have either never written a report with spin or written only one direction of spin but not
the other.
3.3 Descriptive Evidence of Analyst Spin and Stock Rating, Forecast Error, and Dispersion
To motivate our inquiry, we first attempt to gain insight into how analyst spin relates to
constructs commonly used in prior studies on analysts. Namely, we examine how analyst spin
relates to analyst stock ratings (i.e., recommendations), forecast errors, and forecast dispersion.
To provide descriptive evidence on the relation between analyst spin and stock ratings, we
tabulate the number and percentage of reports with analyst spin by stock rating. Based on data
from the I/B/E/S detailed recommendations database, we identify the stock rating that an analyst
had for a firm in 39,794 of the reports that we classify as having spin. The results are presented
in Table 2, Panel A. Of the 29,207 analyst reports with positive spin, 63% have a Buy or Strong
Buy rating, 34% have a Hold rating, and 4% have a Sell or Strong Sell rating. This result
suggests that if an analyst spins negative earnings news in a positive light, then the analyst does
so to reiterate or defend a positive stock rating. By comparison, only 52% of the reports without
spin have a Buy or Strong Buy rating. Of the 10,587 analyst reports with negative spin, 15%
have a Sell or Strong Sell rating, 59% have a Hold rating, and 25% have a Buy or Strong Buy
rating. While this result does not show an obvious correlation between negative spin and a
negative stock rating, if one considers a Hold rating to be somewhat negative, as practitioners
16
often do, then 74% of the reports with negative spin have a negative rating (compared to 48% for
reports without spin). Overall, the results in Panel A provide some descriptive evidence that
analyst’s direction of spin is related to an analyst’s rating on a firm’s stock. For this reason, we
include indicator variables for stock rating as controls in subsequent regression analyses
(discussed later in Section 4).
Next, we examine the relation between analyst spin and analyst forecast errors. Individual
analyst forecast errors can vary across analysts because each analyst can have a different
expectation of a firm’s earnings prior to the earnings announcement. We investigate whether
analysts who spin earnings news tend to have larger, smaller, or similar forecast errors as
analysts who do not spin earnings news. If analyst spin is related to the magnitude of forecast
errors, then we can gain insight into how an analyst’s decision to spin earnings news may be
related to that analyst’s expectation about the firm prior to the earnings announcement. That is,
analysts who spin earnings news may have higher or lower expectations of a firm’s earnings
relative to analysts who do not spin earnings news. To investigate our conjecture, we compute
the average absolute forecast error (ABSFE) for the analyst reports with spin and compare it to
the average absolute forecast error for the analyst reports without spin.
Panel B of Table 2 presents the results of this analysis using the subsample of analyst
reports that were published on the same day or one day after a firm’s earnings announcement.
We find that the mean ABSFE for the analyst reports with spin is $0.076, which is higher than
the $0.071 mean ABSFE for the analyst reports without spin, and the difference is significant at
the 1% level based on a two-sided t-test. The median value of ABSFE is $0.030 for each group,
but a closer investigation of the distribution of the values indicates that a greater proportion of
the absolute forecast errors for the analyst reports with spin are one cent or higher, which drives
17
the significant difference in the Wilcoxon signed rank test. These findings indicate that analysts
who spin earnings news tend to have forecasts that are farther from the eventual reported EPS,
which is indicative of those analysts having higher or lower expectations about a firm’s
performance relative to analysts who do not spin earnings news. Our results are consistent with
prior studies that show analyst compensation (Groysberg, Healy, and Maber 2011) and career
advancement (Hilary and Hsu 2013) are not necessarily tied to forecast accuracy.
Lastly, we examine the relation between analyst spin and analyst forecast dispersion.
Dispersion represents disagreement among analysts about a firm’s future earnings and is
measured at the firm level, typically each quarter. Therefore, we examine how analysts who spin
earnings news affect the overall dispersion of forecasts for a firm in a given quarter. If analysts
who spin earnings news tend to have more extreme earnings forecasts relative to analysts who do
not spin, as indicated by our previous results on absolute forecast error (ABSFE), then the
analysts who spin should contribute to higher overall dispersion. We compute dispersion as the
standard deviation of all analysts’ EPS forecasts for a firm in a given quarter. The results are
presented in Table 2, Panel C. The mean and median dispersion for all firm-quarters in our
sample is $0.041 and $0.017, respectively. When we exclude all analysts who spin earnings news
at least once during the quarter for a given firm, the mean and median dispersion is $0.036 and
$0.014, respectively. The differences in the mean and median values are significant at the 1%
level. These results related to analyst spin and analyst forecast dispersion, coupled with the
previous results for analyst forecast error, indicate that analysts who spin earnings news tend to
have forecasts that are farther from the eventual reported EPS and consensus EPS forecast. This
evidence also suggests one explanation for analyst spin: standing out from the crowd. That is, an
analyst spins earnings news (or any firm news) to persuade investors in favor of the analyst’s
18
contrarian view (rating) of the firm and its stock. Our finding is consistent with Bernhardt,
Campello, and Kutsoati (2006), who focus only on earnings forecasts but find that analysts
systematically anti-herd.
4. Determinants of Analyst Spin
Our first hypothesis is that analysts with higher credibility are more likely to spin
earnings news than analysts that have less credibility. We test this hypothesis by running a
logistic regression in which the dependent variable is an indicator variable for whether an analyst
report contains positive or negative spin, the independent variables of interest are proxies for
analyst credibility, and other variables control for firm and brokerage characteristics. For each
report in our sample written by analyst j for firm i at time t, we estimate the following regression
equation:
P(POSSPIN=1 or NEGSPIN=1)j,i,t = β0 + β1(EXPERIENCEj,t) + β2(EXPWITHFIRMj,i,t)
+ β3(COVERAGESIZEj,t) + β4(COVERAGEFOCUSj,t) + β5(FORECASTFREQj,i,t) +
βk(CONTROLSj,i,t) + Year and Industry Fixed Effects
(1)
4.1 Definition of Variables
Following the identification procedure outlined in Section 3.2, we define two indicator
variables to be used as dependent variables, POSSPIN and NEGSPIN, set to 1 (0 otherwise) if an
analyst report contains positive spin and negative spin, respectively. Considering that our
construct of interest, analyst credibility, is unobservable and time-varying, we use five different
analyst-specific or analyst-firm-specific measures to proxy for the level of credibility that an
analyst has in the year of the report. EXPERIENCE is the number of years of experience that an
analyst has accumulated, as measured by the number of years that an analyst has been in the
I/B/E/S database by the year of the report. EXPWITHFIRM is the number of years of experience
19
that an analyst has covered the firm for which the report is written, as measured by the number of
years that an analyst has published forecasts for that firm.12 We expect analyst credibility to be
positively correlated with EXPERIENCE and EXPWITHFIRM, which implies a positive
coefficient for each variable (β1 > 0, β2 > 0) in equation (1). To capture an analyst’s breadth and
depth of industry knowledge, which we expect to be positively correlated with credibility, we
measure the number of distinct firms that an analyst covers (COVERAGESIZE) and the degree
to which those firms are in the same industry (COVERAGEFOCUS). COVERAGEFOCUS is
the number of distinct three-digit Standard Industrial Classification (SIC) codes that classify the
firms in an analyst’s coverage set in the year of the report, multiplied by −1 to ease interpretation
of its coefficient. We expect analyst credibility to be positively correlated with
COVERAGESIZE and COVERAGEFOCUS, which again implies a positive coefficient for each
variable (β3 > 0, β4 > 0). The last variable used to proxy for credibility is FORECASTFREQ,
defined as the number of dates in which an analyst issues earnings forecasts for a firm in the year
of the report, and it is intended to capture an analyst’s effort in covering a firm. Consistent with
the other proxies for analyst credibility, we predict a positive coefficient for FORECASTFREQ
(β5 > 0). These and other variables are summarized in Appendix C.
We include several variables in equation (1) to control for other factors that may be
associated with an analyst’s decision to spin earnings news. As our descriptive evidence
discussed in Section 3.3 shows, analyst spin may be related to the analyst’s stock rating.
Accordingly, we define five indicator variables set to 1 (0 otherwise) to capture an analyst’s
stock rating (recommendation) for the firm that is the subject of the report: STRONGBUY,
Brown, Call, Clement, and Sharp (2015) find that a sell-side analyst’s experience in covering a specific firm is the
number one attribute that buy-side analysts consider in deciding whether to use the sell-side analyst’s information.
12
20
BUY, HOLD, SELL, and STRONGSELL.13 HOLD is used as the base case and is therefore
excluded from the regression. Because analyst credibility may be correlated with the credibility
of the brokerage house that employs the analyst, we include TOPBROKER as a control, 14
defined as an indicator variable set to 1 (0 otherwise) if a brokerage house has ever been among
the top ten employers of ranked analysts in Institutional Investor’s All-American sell-side equity
research survey from 2005 to 2010.15
The last set of control variables we define are firm-specific characteristics. FIRMSIZE is
the log of the firm’s market capitalization, MTOB is the market-to-book ratio, LEVERAGE is
the debt-to-equity ratio, EARNGROWTH is the seasonal difference in earnings before
extraordinary items deflated by total assets, and ROA is earnings before extraordinary items
deflated by total assets. Each of these variables are computed using firms’ quarterly financial
data from Compustat and are measured for the most recent fiscal quarter ended prior to an
analyst’s report date. We also proxy for a firm’s stock volatility, measured as the standard
deviation of monthly stock returns during the twelve months prior to an analyst’s report date
(STD_PRIOR1Y_RET) using data from CRSP.
Table 3, Panel A shows the distribution of the variables used in regression equation (1).
All continuous variables have been winsorized at the 1st and 99th percentiles to reduce the
influence of outliers and observations with negative market-to-ratios have been excluded. For
indicator variables, we only report the mean. Analysts have, on average, 8.3 years of general
experience and 4.5 years of experience covering specific firms. The mean number of firms in an
13
When a stock rating cannot be identified for an analyst report due to missing data from I/B/E/S, we set the
indicator variables to missing values.
14
While TOPBROKER can also proxy for analyst credibility in most studies in the analyst literature, in this study, it
proxies for an analyst’s current career standing.
15
Our sample contains seven of twelve brokerage houses that have ever been in the top ten between 2005 and 2010,
with only Lehman Brothers, Merrill Lynch, Goldman Sachs, Bank of America Securities, and UBS missing because
they are not contained in the Thomson One Banker database. The seven brokerage houses account for 34.5% of the
reports in our sample.
21
analyst’s coverage set is 17 firms, spanning 6 distinct three-digit SIC industries, on average. The
average analyst issues forecasts for firms seven times per year, and approximately one-third of
the reports in our sample come from top tier brokerage firms.
4.2 Regression Results
The results from estimating regression equation (1) are presented in Table 3, Panel B.
Year and industry fixed effects are included and standard errors are clustered by analyst. 16
Column (1) presents the results for when the dependent variable is POSSPIN, while Column (2)
presents the results for NEGSPIN. In both columns, the coefficient for EXPERIENCE is
insignificant, indicating that an analyst’s general years of experience is not a determinant of
analyst spin. In Column (1), the coefficients for COVERAGESIZE, COVERAGEFOCUS, and
FORECASTFREQ are significantly positive at the 1% level. To assess the economic
significance of the coefficients, we multiply each by the inter-quartile range of the corresponding
variable and compute the exponential function (less one) to calculate the increase in the odds
ratio. An inter-quartile shift in the values for COVERAGESIZE, COVERAGEFOCUS, and
FORECASTFREQ increase the odds by 7.5%, 11.1%, and 7.0%, respectively. In Column (2),
the coefficients for EXPWITHFIRM and FORECASTFREQ are significant, and an inter-quartile
shift in the values for these variables increase the odds by 9.6% and 4.9%, respectively. The
other variables of interest are insignificant, however, we note that there is less statistical power
when the dependent variable is NEGSPIN because only 2.9% of the analyst reports in our sample
exhibit negative spin.
16
For the industry fixed effects, we include six indicator variables to capture the Fama-French 5 industries
(consumer goods, manufacturing, high-tech, healthcare, and others) and the financial industry.
22
In terms of control variables, TOPBROKER has a significantly negative coefficient in
Column (1), which suggests that analysts who spin earnings news positively tend to work for less
prestigious brokerage houses. We conjecture that these analysts attempt to stand out from the
crowd to gain recognition and opportunities to advance their careers to one of the top-tier
brokerage houses, which we formally test in the next section. In both columns, there is high
explanatory power for the indicator variables BUY, STRONGBUY, SELL, and STRONGSELL,
consistent with the descriptive evidence in Table 2, Panel A. Analysts are more likely to spin
earnings news positively (negatively) when they have a positive (negative) rating on a firm’s
stock (relative to HOLD) and when a firm is smaller, has lower growth and higher leverage, and
when the firm’s stock has lower volatility. In summary, the results of our regression analysis
provide some evidence consistent with our first hypothesis insofar as analysts with more
credibility, as proxied by analysts with more firm-specific experience, greater breadth and depth
of coverage, and higher frequency of coverage forecasting activity, are more likely to spin
earnings news than analysts with less credibility.
5. Analyst Spin and Analyst Career Outcomes
Our second hypothesis is that analyst spin is associated with analyst career advancement.
While prior studies show most analysts tend to herd in their forecasts, our evidence thus far
shows that analysts who spin earnings news tend to have forecasts that diverge from the
consensus. If standing out from the crowd increases visibility and recognition for the analysts
who spin, then we expect them to be more likely to experience positive career outcomes.
Alternatively, if analyst spin represents “cheap talk” or some form of opportunism, then analysts
who spin should not experience better career outcomes or experience negative career outcomes.
23
Since we cannot observe changes in compensation for analysts or their promotions within
a brokerage house, we focus on analysts who changed employers. Our key assumption is that
moving to a top-tier brokerage house from a lower-tier brokerage house is career advancement,
while the opposite is career decline. We define a brokerage house to be in the top tier if it was
ever named among the top ten employers of ranked analysts in Institutional Investor’s AllAmerican sell-side equity research survey from 2005 to 2010.17 We consider movements within
the top tier and within the lower tier as neither advancement nor decline. In addition, we consider
an exit from the profession as a negative career outcome, although there is more measurement
error in this case because an exit could be the result of retirement or more attractive career
opportunities for an analyst.
Out of the 2,787 unique analysts in our sample, 913 of them experienced career
movements of any kind during our sample period, based on I/B/E/S data and the year-over-year
changes in analysts’ affiliations. Among the career movers, we identify 78 cases of career
advancement, 170 cases of career decline, 106 cases of movement within top-tier brokerage
houses, 962 cases of movement within lower-tier brokerage houses, and 1,435 cases of career
exit. We define three indicator variables set to 1 (0 otherwise) to capture the aforemented career
movements across brokerage houses: ADVANCE, DECLINE, and EXIT. For each analyst in our
sample, we measure the indicator variables for each calendar year, starting in 2006 and ending in
2011.
We test our second hypothesis by running a logistic regression in which the dependent
variable is ADVANCE, DECLINE, or EXIT, and the independent variables are indicator
variables for whether an analyst wrote a report with positive spin or negative spin in the prior
17
There are twelve brokerage firms that were ever in the top ten between 2005 and 2010: Lehman Brothers, JP
Morgan, Merrill Lynch (became part of Bank of America in 2009), UBS, Credit Suisse, Sanford C. Bernstein,
Citigroup, Goldman Sachs, Morgan Stanley, Deutsche Bank, Bank of America Securities, and Bear Stearns.
24
calendar year and standard errors are clustered by analyst. We include year fixed effects in case
analyst job changes were more prevalent in certain years. For each analyst j and year t, we
estimate the following regression equation:
P(ADVANCE=1 or DECLINE=1 or EXIT=1)j,t = β0 + β1(POSSPINj,t-1) + β2(NEGSPINj,t(2)
1) + Year Fixed Effects
We estimate this regression on two samples. The first sample includes all analysts; i.e., we
include those who did not change employers during our sample period, resulting in 8,837
analyst-years for the regression. The second sample only includes years when analysts changed
employers or exited the profession, resulting in 2,154 analyst-years.
The results from estimating regression equation (2) are presented in Table 4. Panel A
shows the results from regressions run on the first sample using all analysts. Column (1) shows
the result when the dependent variable is ADVANCE; the coefficient on prior year spin
(POSSPINt-1) is significantly positive at the 5% level, indicating that analysts who spin earnings
news in one year are more likely to move to a top-tier brokerage house in the following year. In
Column (2), where the dependent variables is DECLINE, the coefficient on POSSPINt-1 is
insignificant, suggesting that there is no association between positive analyst spin and career
declines. In Column (3), where the dependent variable is EXIT, the coefficient on POSSPINt-1 is
significantly negative at the 1% level, indicating that analysts who spin earnings news positively
are less likely to exit the profession the following year. In terms of economic significance,
compared to analysts who did not spin, analysts who spin positively in the prior year have 3.36%
higher odds to advance to a top-tier brokerage house and 23.7% lower odds to exit the
professional in the current year. Columns (4) through (6) present the results when we re-run the
25
regressions with prior year negative spin (NEGSPINt-1) as the independent variable; the results
are similar to those presented in Columns (1) through (3).
Panel B shows the results of the regression run on the second sample which only includes
years when analysts changed employers or exited the profession. Despite the loss of power, the
results are similar to those presented in Panel A. We find that analysts who spin positively or
negatively tend to advance in their careers by moving to top-tier brokerage houses, while we do
not find that they decline in their careers. We also find that analysts who spin are less likely to
exit the profession. Overall, we interpret the results in Table 4 to be supportive of our second
hypothesis that analyst spin is associated with analyst career advancement.18
6. Analyst Spin and Market Reactions
6.1 Short-Term Market Reaction
Our third hypothesis is that analyst reports with spin are associated with a greater market
reaction relative to analyst reports without spin. If analyst reports with spin lead to greater
disagreement among investors, then the market reaction to analyst reports with spin should be
greater relative to analyst reports without spin (Kim and Verrecchia 1991, 1994; Bamber et al.
1997). In contrast, if analyst spin reflects inferior interpretation of firm news that sophisticated
investors largely discount, then analyst reports with spin should be associated with lower market
reactions than reports without spin.
18
We note that the results in Table 4, along with earlier results on forecast errors in Table 2, Panel B, do not suggest
that analysts who are more inaccurate (have higher forecast errors) are more likely to advance in their careers. Our
results highlight that analysts who spin earnings news tend to advance in their careers, and that this subsample of
analysts tend to have slightly larger forecast errors than analysts who do not spin. Further, our results are not
attributable to all-star analysts because, unlike the vast majority of all-stars, analysts who spin tend to work for less
prestigious brokerages.
26
We test our third hypothesis by computing three market measures—absolute return, share
turnover, and bid-ask spread—for firms on the dates in which any analyst report was published
about the firm. Absolute return (ABSRET) is defined as the absolute value of the stock return,
share turnover (TURNOVER) is the trading volume divided by shares outstanding, and bid-ask
spread (SPREAD) is the ask price minus the bid price divided by the mean of the ask and bid
prices. Each of these variables are computed for the analyst report date using daily stock data
from CRSP. Then we compare the average value of the market measures for reports with spin to
the reports without spin. In addition, since analyst reports tend to cluster on the days immediately
following firms’ earnings announcements, we repeat the analysis excluding reports that were
published on a firm’s earnings announcement date or one day afterwards.
The results of our analysis are presented in Table 5. Panel A shows the mean and median
values of ABSRET, TURNOVER, and SPREAD for all report-firm-days. Tests of differences in
means are based on two-sided t-tests and differences in medians are based on Wilcoxon signed
rank tests. For analyst reports with spin, the mean (median) ABSRET is 3.89% (2.45%), which is
significantly greater (at the 1% level) than the 2.57% (1.53%) for reports without spin.
Regarding TURNOVER, the mean (median) value of 2.66% (1.77%) for analyst reports with
spin is significantly greater than the 1.77% (1.15%) for reports without spin. Finally, for
SPREAD, the mean (median) value of 0.16% (0.09%) for analyst reports with spin is
significantly greater than the 0.14% (0.08%) for reports without spin. For each market measure,
we find the mean and median market reaction is greater for analyst reports with spin, consistent
with our third hypothesis.
Panel B shows the results excluding analyst reports published on or one day after a firm’s
earnings announcement. As one would expect, all the mean and median market reaction
27
variables are smaller than the comparable values in Panel A because the reports are not tied to a
specific earnings announcement. Despite the reduction in the number of firm-days in the analysis
and the lower average market reactions, we again find that the mean and median ABSRET,
TURNOVER, and SPREAD are greater for analyst reports with spin, compared to analyst
reports without spin. In summary, we interpret the results in Table 5 to be consistent with analyst
spin being associated with greater disagreement across investors and therefore supportive of our
third hypothesis.
6.2 Stock Predictions of Analysts who Spin
Having documented the career outcomes of analysts who spin and the short-term stock
market reactions of analyst reports with spin, a natural question arises as to whether or not
analysts who spin are correct in their stock calls. On the one hand, if spin reflects superior
interpretation of firm news, then analysts who spin earnings news should predict stock
performance better than analysts who interpret similar earnings news but do not spin the news.
On the other hand, if spin represents “cheap talk” or opportunism, then analysts who spin should
not have better stock predictions relative to analysts who do not spin. As such, we conclude by
examining the short- and long-term stock predictions of analyst who spin. Specifically, we
compare firms’ stock returns in the 3- and 12-month period following the release of analyst
reports both with and without spin.
To obtain a control sample of analyst reports about other firms without spin, we use
propensity score matching based on the determinants model found in Table 3, Panel B. The
intent of our matching procedure is to compare the stock prediction of an analyst who spins
28
earnings news about a given firm to another analyst who does not spin earnings news about a
different firm but had a similar propensity to spin the news.19
The results of our analysis are presented in Table 6. Panel A shows results of a
comparison between analyst reports with positive spin and a control sample of reports without
spin. We find that analysts who spin negative earnings news in a positive light tend to predict
higher positive returns than analysts who do not spin the news (i.e., they interpret negative
earnings news in a negative light). The mean 3-month (12-month) return following analyst
reports with positive spin is 3.0% (11.2%), compared to 1.6% (8.0%) for analyst reports without
spin, and the difference is significant at the 1% (5%) level. Panel B shows results of a
comparison between analyst reports with negative spin and a control sample of analyst reports
without spin. We find that analysts who spin positive earnings news in a negative light tend to
predict lower positive returns than analysts who do not spin the news (i.e., they interpret positive
earnings news in a positive light). The mean 3-month (12-month) return following analyst
reports with negative spin is 2.3% (10.2%), compared to 3.6% (15.1%) for analyst reports
without spin, and the difference is significant at the 1% (5%) level. In conclusion, we interpret
the results in Table 6 to be consistent with analyst spin being reflective of analysts’ superior
beliefs about future stock performance and therefore provide further support for our third and
final hypothesis.
19
Separately for each direction of spin (positive or negative), we match each analyst report with spin with another
analyst report without spin (about a different firm) that (a) has the same directional quantitative earnings news, (b) is
not followed by any spin report during the return accumulation period, and (c) has the closest likelihood of being a
report with (positive or negative, as appropriate) spin. We require the difference in the likelihood of being a spin
report between the treatment and control observation to be no larger than 0.01. An analysis of covariates reveals
that the treatment and control samples are well-balanced.
29
7. Conclusion
In this study, we examine an interesting and important institutional phenomenon in which
sell-side equity analysts interpret earnings news in a direction that appears to contradict
conventional wisdom. We call this phenomenon “analyst spin” and find that analysts with higher
credibility tend to spin earnings news more so than analysts with lower credibility. We also
provide evidence that their spin reflects superior interpretation of imprecise firm news, as
analysts who spin are more likely to move up to a top-tier brokerage in the following year and
also less likely to exit the profession. Further, we corroborate these results by showing that
analyst spin reflects analysts’ superior beliefs about future stock performance. In contrast, we do
not find that analysts who spin are more likely to decline in their careers. Finally, we show that
analyst spin is associated with greater market reactions. Our study extends the analyst literature
by documenting a subtle and nuanced aspect of analyst research—disagreement with the
apparent earnings news—which likely reflects analyst experience, knowledge, and effort rather
than opportunism.
30
References
Asch, S. E. 1946. Forming Impressions of Personality. Journal of Abnormal and Social
Psychology 41 (3): 1230-1240.
Asquith, P., M. Mikhail, and A. Au. 2005. Information content of equity analyst reports. Journal
of Financial Economics 75 (2): 245-282.
Bamber, L., O. Barron., and T. Stober. 1997. Trading volume and different aspects of
disagreement coincident with earnings announcements. The Accounting Review 72 (4): 575-597.
Barron, O., O. Kim., S. Lim., and D. Stevens. 1998. Using analysts' forecasts to measure
properties of analysts' information environment. The Accounting Review 73 (4): 421−433.
Bernhardt, D., M. Campello, and E. Kutsoati. 2006. Who Herds? Journal of Financial
Economics 80: 657–675.
Bowen, R. M., A. K. Davis, and D. A. Matsumoto. 2005. Emphasis on pro forma versus GAAP
earnings in quarterly press releases: Determinants, SEC intervention, and market reactions. The
Accounting Review 80 (4): 1011–1038.
Bradshaw, M. T. 2011. Analysts’ forecasts: What do we know after decades of work? Working
paper.
Brav, A., and R. Lehavy. 2003. An empirical analysis of analysts' target prices: Short-term
informativeness and long-term dynamics. The Journal of Finance 58 (5): 1933-1968.
Brown, L.D., P. Griffin, R. Hagerman., and M. Zmijewski. 1987. Security analyst superiority
relative to univariate time-series models in forecasting quarterly earnings. Journal of Accounting
and Economics 9 (1): 61-87.
Brown, L.D., A.C. Call, M.B. Clement, N.Y. Sharp. 2015. Skin in the Game: The Inputs and
Incentives that Shape Buy-Side Analysts’ Stock Recommendations. Working Paper.
Clement, M. 1999. Analyst forecast accuracy: Do ability, resources and portfolio complexity
matter? Journal of Accounting and Economics 27 (3): 285-303.
Clement, M., and S. Tse. 2005. Financial analyst characteristics and herding behavior in
forecasting. Journal of Finance 60 (1): 307−341.
Crawford, V. and J. Sobel. 1982. Strategic Information Transmission. Econometrica 50(6):
1431-1451.
Day, R. 1994. How to write and publish scientific papers. Fourth Edition, Oryx Press, Phoenix.
31
De Franco, G., F.P. Vasvari, D. Vyas., and R. Wittenberg Moerman. 2014. Debt analysts’ view
of debt-equity conflicts of interest. The Accounting Review 89 (2): 571-604.
De Franco, G., O.K. Hope, D. Vyas, and Y. Zhou. 2015. Analyst Report Readability.
Contemporary Accounting Research 32 (1): 76-104.
Farrell, J. and M. Rabin. 1996. The Journal of Economic Perspectives 10(3): 103-118.
Files, R., E. Swanson., and S. Tse. 2009. Stealth Disclosure of Accounting Restatements. The
Accounting Review 84 (5): 1495-1520.
Fischer, P., and P. Stocken. 2010. Analyst information acquisition and communication. The
Accounting Review 85 (6):1985-2009.
Francis, J., and D. Philbrick. 1993. Analysts’ decisions as products of a multi-task environment.
Journal of Accounting Research 31 (2): 216-230.
Fried, D., and D. Givoly. 1982. Financial analysts’ forecasts of earnings: a better surrogate for
market expectations. Journal of Accounting and Economics 4 (2): 85-107.
Goldstein, M.A., P.J. Irvine, E. Kandel., and Z. Wiener. 2009. Brokerage commissions and
institutional trading patterns. Review of Financial Studies 22 (12): 5175-5212.
Groysberg, B., P.M. Healy, and D.A. Maber. 2011. What Drives Sell-Side Analyst
Compensation at High-Status Investment Banks? Journal of Accounting Research 49 (4): 9691000.
Henry, E. 2006. Market reaction to verbal components of earnings press releases: Event study
using a predictive algorithm. Journal of Emerging Technologies in Accounting 3 (1): 1-19.
Henry, E. 2008. Are investors influenced by how earnings press releases are written? Journal of
Business Communication 45 (4): 363-407.
Hilary, G., and C. Hsu. 2013. Analyst Forecast Consistency. Journal of Finance 68 (1): 271-297.
Hirshleifer, D., and S.H.Teoh. 2003. Limited attention, information disclosure, and financial
reporting. Journal of Accounting & Economics, 36 (1-3): 337-386.
Hirshleifer, D., S.S. Lim, and S.H. Teoh. 2009. Driven to Distraction: Extraneous Events and
Underreaction to Earnings News. Journal of Finance, 64 (5): 2289-2325.
Hong, H., J. Kubik, and A. Solomon. 2000. Security Analysts’ Career Concerns and Herding of
Earnings Forecasts. The RAND Journal of Economics 31(1): 121-144.
32
Huang, X., A. Nekrasov., and S. H. Teoh. 2013. Headline Salience and Over- and Underreactions
to Earnings. Working paper.
Huang, A.H., A.Y. Zang, and R. Zheng. 2014. Evidence on the information content of text in
analyst reports. The Accounting Review 89 (6): 2151-2180.
Kim, O., and R. E. Verrecchia. 1991. Trading volume and price reactions to public
announcements. Journal of Accounting Research 29 (2): 302-321.
Kim, O., and R. E. Verrecchia. 1994. Market liquidity and volume around earnings
announcements. Journal of Accounting and Economics 17 (1-2): 41-67.
Lin, H., and M. McNichols. 1998. Underwriting relationships, analysts’ earnings forecasts and
investment recommendations. Journal of Accounting and Economics 25 (1): 101-127.
Loh, R.K., and R.M. Stulz. 2011. When Are Analyst Recommendation Changes Influential? The
Review of Financial Studies 24 (2): 593-627.
Loughran, T. and B. McDonald. 2011. When is a Liability Not a Liability? Textual Analysis,
Dictionaries, and 10-Ks. Journal of Finance 56(1): 35-65.
Mahoney, W. F. 1991. Investor Relations: The Professional’s Guide to Financial Marketing and
Communications. New York, NY: New York Institute of Finance.
McNichols, M., and P. O’Brien. 1997. Self-selection and analyst coverage. Journal of
Accounting Research 35 (Supplement): 167-199.
Murdock Jr., B.B. 1960. The distinctiveness of stimuli. Psychological Review 67 (1): 16–31.
Mola, S., P. Rau., and A. Khorana. 2013. Is there life after the complete loss of analyst coverage?
The Accounting Review 88 (2):667–705
O’Brien, P. 1988. Analysts' forecasts as earnings expectations. Journal of Accounting and
Economics 10 (1): 53-83.
Paiva, C.E., J.P.S.N. Lima, B.S.R. Paiva. 2012. Articles with short titles describing the results
are cited more often. Clinics 67 (5): 509-513.
Philbrick, D., and W. Ricks. 1991. Using Value Line and IBES analyst forecasts in accounting
research. Journal of Accounting Research 29 (2): 397-417.
Rabin, M., and J. Schrag. 1999. First Impressions Matter: A Model of Confirmatory Bias.
Quarterly Journal of Economics 114 (1): 37-82.
Trueman, B. 1994. Analyst forecast and herding behavior. Review of Financial Studies 7 (1):
97−124.
33
Welch, I. 2000. Herding among Security Analysts. Journal of Financial Economics 58 (3): 369–
96.
Womack, K. 1996. Do brokerage analysts' recommendations have investment value? The
Journal of Finance 51 (1): 137-167.
34
Appendix A: List of positive and negative words used by analysts
Panel A: Positive words
above
accelerate
accelerated
accelerates
accelerating
acceleration
accretion
accretive
add
added
adding
adds
affirm
affirmed
affirming
affirms
aggressive
aggressively
ahead
answer
answered
answering
answers
approvable
approval
approve
approved
approves
approving
attractive
attractiveness
beat
beating
beats
began
begin
beginning
begins
begun
beneficial
beneficially
benefit
benefited
benefiting
benefits
benefitted
benefitting
best
better
big
bigger
biggest
bolster
bolstered
bolstering
bolsters
boost
boosted
boosting
boosts
bright
brighter
brightest
build
building
builds
built
bullish
bullishness
buy
buyback
buybacks
buyer
buyers
buying
buys
cheap
cheaper
cheapest
clarity
clean
cleaner
cleanest
clear
clearer
clearest
clearing
clearly
clears
comfort
comforted
comforting
comforts
compelling
competitive
competitively
competitiveness
complete
completed
completes
completing
confidence
confident
create
created
creates
creating
creative
creatively
creativeness
creativity
cylinders
decent
deep
deeper
deepest
deeply
deliver
delivered
delivering
delivers
discipline
disciplines
drive
driven
driver
drivers
drives
driving
drove
easier
easiest
easily
easy
efficiency
efficient
efficiently
emerge
emerged
emerges
emerging
encourage
encouraged
encouragement
encourages
encouraging
endeavor
endeavored
endeavoring
endeavors
exceed
exceeded
exceeding
exceeds
excellence
excellent
expand
expanded
expanding
expands
expansion
fast
faster
fastest
favor
favorable
favorably
favored
favoring
favorite
favorites
favors
firing
focus
focused
focuses
focusing
gain
gained
gaining
gains
generate
generated
generates
generating
good
great
greater
greatest
greatly
greatness
grew
grow
growing
grown
grows
growth
happier
happiest
happiness
happy
healthier
healthiest
healthy
help
helped
helping
helps
high
higher
highest
highlight
highlighted
highlighting
highlights
highly
hit
hits
hitting
hot
impress
impressed
impresses
impressing
impressive
impressively
improve
improved
improvement
improvements
improves
improving
increase
increased
increases
increasing
inexpensive
inflection
inflections
intact
interesting
king
launch
launched
launches
launching
lead
leader
leadership
leading
leads
led
lift
lifted
lifting
lifts
like
likes
love
loves
meet
meets
met
momentum
more
nice
nicer
nicest
opportunities
opportunity
optimistic
outperform
outperformance
outperformed
outperforming
outperforms
overblown
overdone
overweight
ow
patience
patient
plus
poise
poised
poises
poising
positive
positively
positives
power
powerful
premium
pretty
profitability
profitable
profitably
progress
progresses
progressing
promising
prospect
prospecting
prospects
pullback
pullbacks
qualities
quality
raise
raised
raises
raising
ramp
reaccelerate
reaccelerated
reaccelerates
reaccelerating
reacceleration
reaffirm
reaffirmed
reaffirming
reaffirms
reasonable
reassure
reassured
reassures
reassuring
rebound
rebounded
rebounding
rebounds
record
reinforce
reinforced
reinforcement
reinforces
reinforcing
relief
reliefs
repurchase
repurchased
repurchases
repurchasing
resolution
resolutions
respectable
respectably
reward
rewarded
rewarding
rewards
right
rise
rises
rising
robust
robustness
rolling
savings
snippet
snippets
solid
solidly
solidness
solution
solutions
stability
stabilization
stabilizations
stabilize
stabilized
stabilizes
stabilizing
stable
start
started
starting
starts
stellar
strength
strengthen
strengthened
strengthening
strengthens
strengths
strong
stronger
strongest
strongly
strongness
succeed
succeeded
succeeding
succeeds
success
successes
successful
successfully
superior
surge
surged
surges
surging
sweet
synergies
synergy
tidbit
tidbits
track
tracked
tracking
tracks
traction
transform
transformation
transforming
transforms
trough
turnaround
underappreciate
underappreciated
underappreciates
underappreciating
undervalue
undervalued
up
upbeat
upgrade
upgraded
upgrades
upgrading
upping
ups
upside
upturn
upturned
upturning
upturns
upward
valuable
value
values
victory
warming
well
win
winner
winners
winning
wins
won
working
35
Appendix A: List of positive and negative words used by analysts (continued)
Panel B: Negative words
bad
behind
below
caution
cautionary
cautioned
cautioning
cautions
cautious
challenge
challenged
challenges
challenging
cloud
clouded
clouding
clouds
cloudy
cold
competition
compress
compressed
compresses
compressing
compression
concern
concerns
cool
cooler
coolest
cools
cut
cuts
cutting
damage
damaged
damages
damaging
decelerate
decelerated
decelerates
decelerating
deceleration
decline
declined
declines
declining
decrease
decreased
decreases
decreasing
delay
delayed
delaying
delays
difficult
difficulties
difficultly
difficulty
dilute
diluted
dilutes
diluting
dilution
disappoint
disappointed
disappointing
disappointingly
disappointment
disappoints
doubt
doubted
doubtful
doubts
down
downgrade
downgraded
downgrades
downgrading
downside
downward
downwards
drag
drags
drastic
drastically
drop
dropped
drops
elusive
expensive
fail
failed
failing
failings
fails
failure
failures
fall
fallen
falling
falls
fell
headwind
headwinds
hurt
hurting
hurts
issue
issues
lack
lacked
lacking
lackluster
lacks
lag
lagged
lagging
lags
late
less
limit
limitation
limited
limits
lose
loses
losing
loss
losses
lost
low
lower
lowered
lowering
lowers
mess
messy
miss
missed
misses
missing
moderate
moderated
moderates
moderating
moderation
mute
muted
negative
negatively
negatives
noise
noisy
overhang
overhanging
overhangs
overhung
overshadow
overshadowed
overshadowing
overshadows
pain
painful
pains
pause
paused
pauses
pausing
persist
persisted
persistence
persistent
persistently
persisting
persists
poor
poorly
pressure
pressures
problem
problems
question
questions
reduce
reduced
reduces
reducing
reduction
remove
removed
removes
removing
restate
restated
restatement
restatements
restates
restating
rich
risk
riskier
riskiest
risks
risky
sell
selling
sells
setback
setbacks
shortfall
shortfalls
sideline
sidelines
slow
slowdown
slowdowns
slowed
slower
slowest
slowing
slowly
slowness
slows
sluggish
sluggishly
sluggishness
soft
soften
softening
softens
softer
softest
softness
sold
storm
struggle
struggled
struggles
struggling
threat
threaten
threatened
threatening
threats
tough
tougher
toughest
trim
trimmed
trimming
trims
ugly
unanswered
uncertain
uncertainty
unclear
underperform
underperformance
underperformed
underperforming
underperforms
underweight
underwhelm
underwhelmed
underwhelming
underwhelms
uneventful
uw
volatile
volatility
warn
warned
warning
warnings
warns
weak
weaken
weakened
weakening
weakens
weaker
weakest
weakly
weakness
weaknesses
weigh
weighed
weighing
weighs
worried
worries
worry
worrying
worse
worsen
worsened
worsening
worsens
worst
36
Appendix B: Examples of Analyst Spin
Panel A: Examples of positive spin on negative earnings forecast error or revision
Company
Associated
Banc-Corp
United
Technologies
Amazon
Anadarko
Petroleum Co.
Brokerage
RBC
Capital Markets
Morgan Stanley
Bernstein
Research
Oppenheimer
& Co.
Date of
Report
Analyst’s
Forecasted
EPS
Actual
Reported
EPS
Analyst
Forecast
Error
Revision
to next qtr
forecasted
EPS
4/23/2010
$0.08
−$0.20
−$0.28
−$0.22
1/25/2006
$0.73
$0.71
−$0.02
None
11/6/2009
n/a
n/a
n/a
−$0.03
2/1/2005
n/a
n/a
n/a
−$0.07
Report Title (with positive tone;
positive words in italics)
Resetting the bar and aggressively focusing on
improving asset quality.
United Technologies high quality Results Bolster
Confidence In Above Consensus 2006 EPS
It’s the 5th inning not the 9th; still more growth and
margin expansion to come.
Higher prices boost earnings; growth strategy on
track.
Panel C: Examples of negative spin on positive earnings forecast error or revision
Company
Monsonto
Brokerage
Morgan
Stanley
Wedbush
Morgan
Jefferies
& Co.
Cardinal Health
Credit Suisse
Alcoa Inc.
Arctic Cat Inc.
Date of
Report
Analyst’s
Forecasted
EPS
Actual
Reported
EPS
Analyst
Forecast
Error
Revision
to next qtr
forecasted
EPS
4/7/2005
$0.36
$0.40
$0.04
$0.03
1/24/2007
$0.35
$0.43
$0.08
None
2/24/2010
n/a
n/a
n/a
$0.15
1/8/2009
n/a
n/a
n/a
$0.17
Report Title (with negative tone;
negative words in italics)
Cost headwinds and weak dollar continue to restrict
earnings.
Difficult outlook for snowmobiles given challenging
weather conditions.
Q2 looks weaker than expected Glyphosate
inventory issues persist
Weaker hospital capex environment still weighs on
CAH.
37
Appendix C: Variable Definitions
Variable
Definition
FE
Forecast error of the analyst issuing a report immediately after a company’s earnings
announcement (on the same day or one day after); FE is measured as actual reported
EPS minus the analyst’s last forecast before the earnings announcement.
ABSFE
Absolute forecast error of the analyst issuing a report immediately after a company’s
earnings announcement; ABSFE measured as the absolute value of actual reported
EPS minus the analyst’s last forecast before the earnings announcement.
NQESTCHG
Revision of an analyst's forecast for the next fiscal quarter on the report. It equals 0
if the analyst did not revise his/her forecast for the next fiscal quarter on the report.
Standard deviation of all analyst EPS forecasts for a company in a given fiscal
quarter.
DISPERSION
POSSPIN
Indicator variable equal to 1 (0 otherwise) if FE<0 or NQESTCHG<0 and the
analyst used more positive words than negative words in the report title.
NEGSPIN
Indicator variable equal to 1 (0 otherwise) if FE>0 or NQESTCHG>0 and the
analyst used more negative words than positive words in the report title.
Indicator variable equal to 1 (0 otherwise) if the stock recommendation on an analyst
report is Strong Buy.
Indicator variable equal to 1 (0 otherwise) if the stock recommendation on an analyst
report is Buy.
Indicator variable equal to 1 (0 otherwise) if the stock recommendation on an analyst
report is Hold.
Indicator variable equal to 1 (0 otherwise) if the stock recommendation on an analyst
report is Sell.
Indicator variable equal to 1 (0 otherwise) if the stock recommendation on an analyst
report is Strong Sell.
Log of the market capitalization of the company at the end of the most recent
reported quarter prior to the report.
The market value of equity divided by stockholders’ equity at the end of the most
recent reported quarter prior to the report. Firms with negative MTOB are excluded.
The debt to equity ratio of the company at the end of the most recent reported
quarter prior to the report.
Change in earnings before extraordinary items for the most recent reported quarter
prior to the report, relative to earnings before extraordinary items for the same fiscal
quarter in the prior year, divided by the total assets of the company.
Earnings before extraordinary items divided by total asset at the end of the most
recent reported quarter prior to the report.
STRONGBUY
BUY
HOLD
SELL
STRONGSELL
FIRMSIZE
MTOB
LEVERAGE
EARNGROWTH
ROA
38
Appendix C: Variable Definitions (continued)
Variable
Definition
TOPBROKER
Indicator variable set to 1 (0 otherwise) if a brokerage house has ever been among
the top ten employers of ranked analysts in Institutional Investor’s All-American
sell-side equity research survey from 2005 to 2010.
Number of years an analyst is contained in the I/B/E/S EPS forecast database.
EXPERIENCE
EXPWITHFIRM
COVERAGESIZE
COVERAGEFOCUS
FORECASTFREQ
ADVANCE
DECLINE
EXIT
ABSRET
Number of years an analyst has published forecasts for a specific company in the
I/B/E/S EPS forecast database.
Number of companies an analyst covers in the year of a report.
Negative of the number of distinct three-digit SIC codes an analyst covers in the
year of the report.
Number of forecasts that an analyst issues for the company in the year of a report.
Indicator variable set to 1 (0 otherwise) if an analyst moves to a top-tier brokerage
house from a lower-tier brokerage house during the year. We define a brokerage
house to be in the top tier if it has ever been among the top ten employers of ranked
analysts in Institutional Investor’s All-American sell-side equity research survey
from 2005 to 2010.
Indicator variable set to 1 (0 otherwise0 if an analyst moves to a lower-tier
brokerage house from a top-tier brokerage house during the year. We define a
brokerage house to be in the top tier if it has ever been among the top ten employers
of ranked analysts in Institutional Investor’s All-American sell-side equity research
survey from 2005 to 2010.
Indicator variable set to 1 (0 otherwise) if an analyst exits from the profession during
the year. We define an analyst to be exit the profession on a day when IBES forecast
file contains no forecasts issued by him/her in the following twelve months.
Absolute value of a company’s stock return on the date of the analyst report.
TURNOVER
Trading volume of a company’s stock divided by shares outstanding on the date of
the analyst report.
SPREAD
Ask price minus bid price, divided by the mean of the bid and ask price, on the date
of the analyst report.
39
Table 1: Sample Composition
Panel A: Number of analyst reports by year and quarter
Year
1st Qtr. 2nd Qtr.
3rd Qtr.
4th Qtr.
2005
18,364
18,664
18,708
19,454
2006
18,495
18,041
16,745
18,003
2007
19,311
16,311
15,577
16,878
2008
16,860
15,851
15,010
15,743
2009
14,873
13,828
15,582
16,173
2010
16,465
14,674
9,603
16,975
Total
104,368
97,369
91,225
103,226
Total
75,190
71,284
68,077
63,464
60,456
57,717
396,188
Panel B: Number and percentage of analyst reports with positive spin by year and quarter
Year
1st Qtr. 2nd Qtr.
3rd Qtr.
4th Qtr.
Total
2005
1,207
1,408
1,443
1,203
5,261
7%
8%
8%
6%
7%
2006
1,416
1,531
1,423
1,065
5,435
8%
8%
8%
6%
8%
2007
1,415
1,362
1,476
1,218
5,471
7%
8%
9%
7%
8%
2008
1,385
1,497
1,555
1,256
5,693
8%
9%
10%
8%
9%
2009
1,401
1,245
1,456
1,178
5,280
9%
9%
9%
7%
9%
2010
1,485
1,456
878
1,270
5,089
9%
10%
9%
7%
9%
Total
8,309
8,499
8,231
7,190
32,229
8%
9%
9%
7%
8%
Panel C: Number and percentage of analyst reports with negative spin by year and quarter
Year
1st Qtr. 2nd Qtr.
3rd Qtr.
4th Qtr.
Total
2005
399
434
500
474
1,807
2%
2%
3%
2%
2%
2006
440
447
482
532
1,901
2%
2%
3%
3%
3%
2007
535
402
437
493
1,867
3%
2%
3%
3%
3%
2008
527
553
586
703
2,369
3%
3%
4%
4%
4%
2009
696
471
530
462
2,159
5%
3%
3%
3%
4%
2010
437
354
206
500
1,497
3%
2%
2%
3%
3%
Total
3,034
2,661
2,741
3,164
11,600
3%
3%
3%
3%
3%
40
Table 1: Sample Composition (continued)
Panel D: Number and percentage of total analyst reports, and reports with analyst spin, by brokerage house
Top 40 Brokerage Firms
(by Number of Reports)
JP Morgan
Credit Suisse
Deutsche Bank
Morgan Stanley
Wells Fargo Securities
RBC Capital Markets
Jefferies & Co.
Oppenheimer & Co.
Bear Stearns
William Blair & Co.
Keybanc Capital Markets
Prudential Equity Group
Buckingham Research Group
Suntrust Robinson Humphrey
CIBC World Markets
Morgan Keegan & Co.
A.G. Edwards & Sons
Wedbush Morgan
Citigroup
Canaccord
Wall Street Strategies
Sterne Agee & Leach
Thinkequity
Janney Montgomery Scott
Stanford Financial Group
Caris & Company
Roth Capital Partners
Fox-Pitt Kelton
D.A. Davidson & Co.
Kaufman Brothers
Macquarie Research
Brean Murray
CL King And Associates
Collins Stewart
Argus Institutional Partners
Susquehanna Financial Group
Rodman & Renshaw
Craig Hallum Capital
Davenport & Company
Pacific Growth
All Others
Total
Total #
of reports
41,778
27,517
25,969
22,858
18,899
18,082
17,396
12,482
11,100
10,901
9,956
9,700
9,404
9,395
8,113
7,984
7,408
7,132
5,870
5,783
5,445
4,890
4,702
4,694
4,131
3,745
3,634
3,513
3,348
3,293
3,167
2,779
2,596
2,393
2,195
2,174
2,016
1,978
1,932
1,926
43,910
396,188
%. of
total
10.5%
6.9%
6.6%
5.8%
4.8%
4.6%
4.4%
3.2%
2.8%
2.8%
2.5%
2.4%
2.4%
2.4%
2.0%
2.0%
1.9%
1.8%
1.5%
1.5%
1.4%
1.2%
1.2%
1.2%
1.0%
0.9%
0.9%
0.9%
0.8%
0.8%
0.8%
0.7%
0.7%
0.6%
0.6%
0.5%
0.5%
0.5%
0.5%
0.5%
11.1%
100.0%
Reports w/
analyst spin
4,487
2,446
2,702
2,504
2,026
2,418
2,341
1,782
1,027
1,018
1,299
888
977
1,100
1,097
939
819
833
581
738
16
729
671
623
571
421
446
446
261
435
364
382
224
298
0
265
185
371
193
225
4,681
43,829
% of Brokerage
reports w/ spin
10.7%
8.9%
10.4%
11.0%
10.7%
13.4%
13.5%
14.3%
9.3%
9.3%
13.0%
9.2%
10.4%
11.7%
13.5%
11.8%
11.1%
11.7%
9.9%
12.8%
0.3%
14.9%
14.3%
13.3%
13.8%
11.2%
12.3%
12.7%
7.8%
13.2%
11.5%
13.7%
8.6%
12.5%
0.0%
12.2%
9.2%
18.8%
10.0%
11.7%
10.7%
11.1%
41
Table 1: Sample Composition (continued)
Panel E: Number and percentage of total analyst reports, and reports with analyst spin, by industry (2-digit SIC)
2-digit
SIC
28
73
36
35
38
60
13
49
67
63
48
56
20
59
37
58
53
62
80
87
50
23
61
79
57
27
26
33
29
55
51
34
42
15
39
40
52
25
82
30
Top 40 Industries
(by Number of Reports)
Chemicals & Allied Products
Business Services
Electronic Equip. & Comp., Except Computers
Industrial & Commercial Machinery & Computers
Control Instruments; Photo., Medical & Optical
Depository Institutions
Oil & Gas Extraction
Electric, Gas, & Sanitary Services
Holding & Other Investment Offices
Insurance Carriers
Communications
Apparel & Accessory Stores
Food & Kindred Products
Miscellaneous Retail
Transportation Equipment
Eating & Drinking Places
General Merchandise Stores
Security/Commodity Brokers, Dealers & Exchanges
Health Services
Engineering, Accounting, Research & Management
Wholesale Trade-durable Goods
Apparel & Finished Products Made From Fabrics
Non-depository Credit Institutions
Amusement & Recreation Services
Home Furniture, Furnishings, & Equipment Stores
Printing, Publishing, & Allied Industries
Paper & Allied Products
Primary Metal Industries
Petroleum Refining & Related Industries
Automotive Dealers & Gasoline Service Stations
Wholesale Trade-non-durable Goods
Fabr. Metal Products, Except Mach. & Transport
Motor Freight Transportation & Warehousing
Building Construction General Contractors
Miscellaneous Manufacturing Industries
Railroad Transportation
Building Materials, Hardware, & Garden Supply
Furniture & Fixtures
Educational Services
Rubber & Miscellaneous Plastics Products
All Others
Total
Total #
% of
Reports w/ % of Reports
of reports total analyst spin
w/ spin
43,034 11.2%
4,096
9.5%
41,558 10.8%
4,737
11.4%
34,081
8.8%
3,995
11.7%
22,108
5.7%
2,406
10.9%
19,829
5.1%
2,494
12.6%
17,819
4.6%
2,398
13.5%
14,738
3.8%
2,502
17.0%
12,173
3.2%
1,253
10.3%
11,988
3.1%
921
7.7%
11,656
3.0%
1,253
10.7%
11,229
2.9%
1,468
13.1%
10,425
2.7%
808
7.8%
9,308
2.4%
993
10.7%
8,459
2.2%
951
11.2%
8,388
2.2%
862
10.3%
8,076
2.1%
956
11.8%
5,909
1.5%
492
8.3%
5,737
1.5%
580
10.1%
5,604
1.5%
584
10.4%
5,001
1.3%
630
12.6%
4,881
1.3%
537
11.0%
4,396
1.1%
458
10.4%
4,077
1.1%
388
9.5%
3,674
1.0%
480
13.1%
3,627
0.9%
406
11.2%
3,323
0.9%
270
8.1%
3,170
0.8%
364
11.5%
3,051
0.8%
380
12.5%
3,032
0.8%
371
12.2%
2,938
0.8%
347
11.8%
2,937
0.8%
295
10.0%
2,576
0.7%
344
13.4%
2,501
0.6%
252
10.1%
2,442
0.6%
341
14.0%
2,354
0.6%
276
11.7%
2,156
0.6%
132
6.1%
2,094
0.5%
211
10.1%
1,838
0.5%
206
11.2%
1,757
0.5%
200
11.4%
1,727
0.4%
232
13.4%
19,536
5.1%
2,254
11.5%
385,207 100.0%
43,123
11.2%
42
Table 1: Sample Composition (continued)
Panel F: Breakdown of analysts who spin and the number of their reports
Positive spin only
Negative spin only
Both directions of spin
No spin
Total analysts
Analysts
494
71
1,612
610
2,787
% of
Analysts
18%
3%
58%
22%
100%
Reports
22,402
2,359
355,771
15,656
396,188
Reports
w/ spin
1,844
98
41,887
43,829
% of Reports
w/ spin
8%
4%
12%
0%
11%
Table 1 describes sample composition. Panel A shows the number of reports by year and calendar quarter. Panels B
and C show the number and percentage of analyst reports that contain positive and negative spin, respectively. Panel
D shows the number and percentage of analyst reports from the top 40 brokerage houses, ranked by the number of
reports contained in the Thomson One Banker database. Panel E shows the number and percentage of analyst reports
by industry for firms with non-missing two-digit SIC codes. Panel F shows a breakdown of analysts who spin and
the number of their reports.
43
Table 2: Sample Descriptive Statistics
Panel A: Analyst spin and stock rating
Number of reports with positive spin
Percentage of total
Number of reports without spin
Percentage of total
Number of reports with negative spin
Percentage of total
Analyst rating / recommendation on a firm's stock
Strong Sell
Sell
Hold
Buy
Strong Buy
164
806
9,863
10,878
7,496
1%
3%
34%
37%
26%
3,947
1%
14,669
5%
124,185
42%
95,024
32%
59,580
20%
426
4%
1,206
11%
6,251
59%
1,742
16%
962
9%
Total
29,207
297,405
10,587
Panel B: Average ABSFE of analyst with and without spin
Analyst reports with Spin
Analyst reports without Spin
Difference
Absolute Forecast Error (ABSFE)
Reports
Mean
Median
20,189 $0.076
$0.030
112,352 $0.071
$0.030
132,541 $0.005 ***
$0.000 ***
Panel C: Average DISPERSION when analysts with spin are included or excluded
All analysts included
Spin analysts excluded
Difference
Firm-quarters
23,835
18,905
42,740
Dispersion
Mean
$0.041
$0.036
$0.004 ***
Median
$0.017
$0.014
$0.003 ***
Table 2 presents sample descriptive statistics. Panel A shows the number and percentage of reports with analyst spin
by stock rating. Based on data from the I/B/E/S detailed recommendations database, we identify the stock rating that
an analyst had for a firm in 39,794 of the reports that we classify as having spin. Panel B compares the average
absolute forecast error (ABSFE) for the analyst reports with spin to the average absolute forecast error for the
analyst reports without spin. Panel C presents the average dispersion of forecasts for firm quarters including or
excluding analysts who spin earnings news at least once during the quarter for a given firm. *, **, *** indicate
significantly different from zero at the 0.10, 0.05, and 0.01 level, respectively, using a two-tailed t-test for means
and a Wilcoxon signed rank test for medians.
44
Table 3: Determinants of Analyst Spin
Panel A: Distribution of variables
Variable
POSSPIN
NEGSPIN
EXPERIENCE
EXPWITHFIRM
COVERAGESIZE
COVERAGEFOCUS
FORECASTFREQ
TOPBROKER
STRONGBUY
BUY
HOLD
SELL
STRONGSELL
FIRMSIZE
MTOB
LEVERAGE
EARNGROWTH
ROA
STD_PRIOR1Y_RET
N
396,188
396,188
376,102
376,102
376,102
376,102
376,102
396,188
396,188
396,188
396,188
396,188
396,188
382,660
378,358
382,648
382,351
382,937
383,689
Mean
0.081
0.029
8.332
4.535
17.016
−6.188
6.939
0.345
0.172
0.272
0.354
0.042
0.011
8.249
3.590
0.919
0.001
0.011
0.104
1st Pctl
25th Pctl
Median
75th Pctl
99th Pctl
1.000
1.000
3.000
−18.000
1.000
4.000
2.000
13.000
−8.000
5.000
7.000
4.000
16.000
−5.000
6.000
11.000
6.000
21.000
−3.000
9.000
25.000
18.000
40.000
−1.000
19.000
4.779
0.545
−0.688
−0.106
−0.148
0.027
7.012
1.717
0.060
−0.004
0.003
0.064
8.119
2.643
0.412
0.001
0.013
0.091
9.475
4.103
0.990
0.007
0.025
0.129
12.147
23.102
11.872
0.108
0.081
0.350
Table 3 reports the determinants of analyst spin. Panel A shows the distribution of the variables used in our test of
determinants of analyst spin, equation (1). All continuous variables have been winsorized at the 1st and 99th
percentiles to reduce the influence of outliers and observations with negative market-to-book ratio have been
excluded. For indicator variables, we only report the mean. Panel B presents results of logistic regressions in which
the dependent variable is an indicator variable of analyst reports that contains positive spin or negative spin, and the
independent variables are proxy variables for analyst credibility and control variables. T-statistics are shown in
parentheses. All variables are defined in Appendix C. *, **, *** Significantly different from zero at the 0.10, 0.05,
and 0.01 level, respectively, using a two-tailed t-test and standard errors clustered by analysts.
45
Table 3: Determinants of Analyst Spin (continued)
Panel B: Logistic regression of analyst spin on proxies for analyst credibility and control variables
EXPERIENCE
H1
Pred. Sign
+
EXPWITHFIRM
+
COVERAGESIZE
+
COVERAGEFOCUS
+
FORECASTFREQ
+
STRONGBUY
BUY
SELL
STRONGSELL
TOPBROKER
FIRMSIZE
MTOB
LEVERAGE
EARNGROWTH
ROA
STD_PRIOR1Y_RET
INTERCEPT
Year and Industry F.E.
N
Pseudo R2
Positive Spin Negative Spin
POSSPIN
NEGSPIN
(1)
(2)
-0.001
-0.003
(-0.20)
(-0.62)
0.005
0.023 ***
(1.11)
(4.89)
0.009 ***
0.002
(2.81)
(0.48)
0.021 ***
0.006
(3.78)
(1.05)
0.017 ***
0.012 ***
(3.30)
(2.81)
0.438 ***
-1.103 ***
(13.76)
(-24.78)
0.420 ***
-0.934 ***
(17.53)
(-26.53)
-0.411 ***
0.609 ***
(-8.52)
(13.97)
-0.793 ***
0.878 ***
(-8.10)
(13.72)
-0.140 ***
-0.020
(-3.86)
(-0.51)
-0.098 ***
-0.065 ***
(-12.51)
(-6.69)
0.003
-0.015 ***
(0.93)
(-3.46)
0.012 *
0.012
(1.88)
(1.48)
-0.463
-0.737
(-1.37)
(-1.47)
-1.411 ***
0.632
(-4.05)
(1.20)
-0.691 ***
-0.059
(-3.37)
(-0.23)
-1.925 ***
-2.927 ***
(-18.20)
(-25.93)
Yes
Yes
358,311
358,311
0.016
0.039
46
Table 4: Analyst Career Outcomes
Panel A: Full sample of analyst-years
Dependent Variable: ADVANCE=1
(1)
POSSPINt-1
1.446 **
(3.61)
NEGSPINt-1
INTERCEPT
Year Fixed Effects
N
Pseudo R2
−6.075 ***
(−13.18)
Yes
8,837
0.040
DECLINE=1
(2)
0.325
(1.48)
−5.329 ***
(−12.95)
Yes
8,837
0.030
EXIT=1
(3)
−0.313 ***
(−4.85)
−1.457 ***
(−16.29)
Yes
8,837
0.015
ADVANCE=1
(4)
DECLINE=1
(5)
EXIT=1
(6)
0.569 **
(2.05)
−5.199 ***
(−13.76)
Yes
8,837
0.020
0.123
(0.60)
−5.148 ***
(−13.62)
Yes
8,837
0.029
−0.281 ***
(−4.37)
−1.537 ***
(−18.73)
Yes
8,837
0.014
DECLINE=1
(5)
EXIT=1
(6)
0.155
(0.73)
−3.820 ***
(−10.03)
Yes
2,154
0.038
−0.433 ***
(−4.70)
0.593 ***
(4.81)
Yes
2,154
0.022
Panel B: Sample of analyst-years in which an analyst changed employers
EXIT=1
ADVANCE=1
Dependent Variable: ADVANCE=1 DECLINE=1
(1)
(2)
(3)
(4)
POSSPINt-1
1.514 ***
0.325
−0.685 ***
(3.73)
(1.45)
(−7.23)
NEGSPINt-1
0.688 ***
(2.44)
INTERCEPT
−4.759 ***
−3.978 ***
0.861 ***
−3.919 ***
(−10.30)
(−9.66)
(6.45)
(−10.20)
Year Fixed Effects
Yes
Yes
Yes
Yes
N
2,154
2,154
2,154
2,154
Pseudo R2
0.057
0.040
0.033
0.032
Table 4 presents results of logistic regressions in which the dependent variable is an indicator variable of career advancement, career decline, or exit from the
profession, and the independent variables are indicator variables for whether an analyst wrote a report with positive spin or negative spin in the prior calendar
year. Year fixed effects are included. Panel A shows results using a sample that includes all analysts, and Panel B shows results using a sample that includes only
years when analysts changed employers or exited the profession. T-statistics are shown in parentheses. All variables are defined in the Appendix C. *, **, ***
Significantly different from zero at the 0.10, 0.05, and 0.01 level, respectively, using a two-tailed test and standard errors clustered by analysts.
47
Table 5: Market Reactions to Analysts Reports with and without spin
Panel A: Market reactions including all firm-days with analyst reports
With analyst spin
Without analyst spin
Difference
Absolute Return (ABSRET)
Share Turnover (TURNOVER)
Bid-Ask Spread (SPREAD)
Firm-days Mean
Median
Firm-days Mean
Median
Firm-days Mean
Median
31,250 3.89%
2.45%
31,250 2.66%
1.77%
31,245 0.16%
0.09%
190,271 2.57%
1.53%
190,278 1.77%
1.15%
190,255 0.14%
0.08%
1.32% ***
0.92% ***
0.89% *** 0.62% ***
0.01% ***
0.01% ***
Panel B: Market reactions excluding days on or after a firm’s earnings announcement
With analyst spin
Without analyst spin
Difference
Absolute Return (ABSRET)
Share Turnover (TURNOVER)
Bid-Ask Spread (SPREAD)
Firm-days Mean
Median
Firm-days Mean
Median
Firm-days Mean
Median
9,485 2.91%
1.81%
9,485 2.20%
1.45%
9,485 0.14%
0.08%
157,989 2.29%
1.41%
157,994 1.64%
1.09%
157,982 0.13%
0.07%
0.62% ***
0.40% ***
0.56% *** 0.36% ***
0.01% ***
0.00% ***
Table 5 compares absolute return, share turnover, and bid-ask spread for firms on the dates of analyst reports with and without spin. Panel A shows results
including all firm days with an analyst report, and Panel B shows results excluding firm days on or after firms’ earnings announcements. *, **, *** indicate
significantly different from zero at the 0.10, 0.05, and 0.01 level, respectively, using a two-tailed test.
48
Table 6: Stock Performance following Analysts Spin Reports
Panel A: Analyst reports with positive spin, compared to analyst reports without spin
3-month returns
# of reports
Mean
12-month returns
# of reports
Mean
Analyst reports with positive spin
10,439 3.0%
2,306 11.2%
Control sample of reports without spin
10,439 1.6%
2,306
Difference
1.4% ***
8.0%
3.2% **
Panel B: Analyst reports with negative spin, compared to analyst reports without spin
3-month returns
# of reports
Mean
Analyst reports with negative spin
4,027 2.3%
Control sample of reports without spin
4,027 3.6%
Difference
−1.3% ***
12-month returns
# of reports
Mean
819 10.2%
819 15.1%
−4.9% **
Table 6 compares firms’ stock returns in the 3 and 12 months following the release of analyst reports both with and
without spin. Panel A shows results of a comparison between analyst reports with positive spin and a control sample
of reports without spin. Panel B shows results of a comparison between analyst reports with negative spin and a
control sample of analyst reports without spin. *, **, *** indicate significantly different from zero at the 0.10, 0.05,
and 0.01 level, respectively, using a two-tailed t-test.
49
Download