Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
Buy-Side Analysts and Earnings Conference Calls
Michael J. Jung Stern School of Business
New York University [email protected]
M.H. Franco Wong Rotman School of Management
University of Toronto [email protected]
X. Frank Zhang School of Management
Yale University [email protected]
October 18, 2015
We thank Larry Brown, Nathan Dong, Daehyun Kim, Stan Markov, Bill O’Dell, Elizabeth Hill Shah, Eugene Soltes, Nan Yang and workshop participants at New York University, University of Toronto, Hong Kong Polytechnic University, Southern Methodist University, George Washington University, and the Haskell & White Corporate Reporting & Governance Academic Conference for their helpful suggestions and comments. We gratefully acknowledge the contributions from Thomson Reuters for granting access to the StreetEvents conference call transcript database. We also acknowledge financial support from New York University, the CPA Professorship in Accounting, University of Toronto, and Yale University.
Buy-Side Analysts and Earnings Conference Calls
Abstract
Companies’ earnings conference calls are perceived to be venues for sell-side equity analysts to ask management questions. In this study, we examine another important conference call participant—the buy-side analyst—that has been underexplored in the literature due to data limitations. Using a large sample of transcripts, we identify 4,045 buy-side analysts from 724 institutional investment firms that participated (i.e., asked a question) on 13,373 conference calls to examine the determinants and implications of their participation. Buy-side analysts are more likely to participate when sell-side analyst coverage is low, dispersion in earnings forecasts is high, and when the hosting company reported bad news, consistent with buy-side analysts directly acquiring information when a company’s information environment is poor. Institutional investors trade more of a company’s stock after their buy-side analysts participate on the call, suggesting buy-side analysts update their stock recommendations after a call and their employers act on the updated research. Finally, we find evidence that the buy-side analysts we observe on the calls and the subsequent trading decisions by their employing institutions are reflective of other buy-side analysts and institutions we do not observe, resulting in company-level changes in stock prices, trading volume, institutional ownership, and short interest. Keywords: Buy-side analysts; institutional investors; sell-side analysts; earnings conference calls
1
1. Introduction
The role of sell-side equity analysts in the capital markets has been extensively
researched by academics over the past several decades (Bradshaw 2011). In contrast, due to data
limitations, there has been very little research on buy-side analysts. Buy-side analysts work for
institutional investment firms and have different incentives and responsibilities compared to their
sell-side counterparts working at brokerage firms (Groysberg, Healy, and Chapman, 2008),
which makes buy-side analysts not only worthy of study in their own right, but also makes the
inferences and conclusions from the sell-side analyst literature not generalizable to buy-side
analysts. While it is widely assumed that buy-side analysts conduct fundamental research and
make stock recommendations to their firms’ portfolio managers, little is known about how they
gather and process information because their research activities are not generally observable. In
this paper, we contribute to the literature on buy-side analysts by 1) identifying 4,045 buy-side
analysts from 724 institutional investment firms that participated (i.e., asked a question) in
13,373 earnings conference calls, 2) examining the economic determinants of their participation,
3) investigating the role of their participation in their investment firms’ trading of the companies’
stock, and 4) exploring the implications for future stock returns, trading volume, total
institutional ownership, and short interest of the company hosting the conference call.1
The few published papers on buy-side analysts have focused on the outputs of their
research: earnings forecasts and stock recommendations (Groysberg, Healy, and Chapman 2008;
Groysberg, Healy, Serafeim, and Shanthikumar 2013; Cheng, Liu, and Qian 2006; Rebello and
Wei 2014; Frey and Herbst 2014). Typically using proprietary data from a single institutional
investment firm, these papers conclude that research generated by buy-side analysts has value
1 Throughout this paper, we use the terms “firm” and “institution” when referring to an institutional investment firm that employs a buy-side analyst and the term “company” when referring to a company that hosts an earnings conference call.
2
and is associated with positive abnormal returns for the funds that use it. In contrast to studies
that focus on the outputs of buy-side research, Brown, Call, Clement and Sharp (2014) survey
344 buy-side analysts from 181 investment firms and conduct follow-up interviews with 16
analysts to gain insights about the inputs and incentives that shape buy-side research. They find
that recent 10-K/Qs are more useful than quarterly conference calls, management guidance, and
recent earnings performance in determining buy-side analysts’ stock recommendations.2
We examine participation in companies’ earnings conference calls to shed light on its
importance as one of the research activities performed by buy-side analysts and to answer several
unexplored research questions. In particular, we are interested in understanding the prevalence of
buy-side analysts in companies’ earnings conference calls and the reasons they participate in the
calls. We test predictions about whether buy-side analyst participation is related to a company’s
information environment, subsequent trading in the company’s stock by the employing
institution, and company-level price discovery.
First, using a sample of 56,285 conference call transcripts from the second quarter of
2002 through the first quarter of 2009, we identify 4,045 buy-side analysts from 724 institutional
investment firms who asked at least one question on 13,373 earnings conference calls. Our
sample includes some of the largest investment firms in the U.S. (e.g., Barclays, Fidelity,
Wellington, and T. Rowe Price) and even several of the buy-side analysts named in Institutional
Investor magazine’s annual “Best of the Buy-Side” rankings, as voted by hundreds of sell-side
2 They argue that buy-side analysts do not value participating in companies’ earnings conference calls because they believe their private information (e.g., ideas, thoughts, opinions, etc.) will be publicly revealed. While this proprietary cost explanation makes sense for some buy-side analysts, it is likely not a belief shared by all buy-side analysts since we observe thousands of them in our sample of conference call transcripts. There are likely other explanations for why buy-side analysts do or do not participate in conference calls. We view our study as complementary to Brown et al. (2014) in trying to better understand the process in which buy-side analysts gather information.
3
analysts each year.3 The participation by these highly-ranked buy-side analysts suggests that
asking a question on a conference call is an important aspect of their research and due diligence.
Buy-side analysts ask questions in 24% of all earnings conference calls, over 3,000 conference
calls have two or more buy-side analysts asking questions, and buy-side analysts represent 5% of
all questioners. Thus, while the vast majority of conference call participants are sell-side equity
analysts, participation by buy-side analysts in earnings conference calls is not rare.
Secondly, we examine why these buy-side analysts participate in conference calls despite
the common perception that they are venues for sell-side analysts and company managers to
interact. Unlike sell-side analysts, who act as information intermediaries gathering and
disseminating information about companies, buy-side analysts are typically private in their
information gathering activities, so their observable participation in conference calls makes for a
unique research setting that has been previously unexplored. We posit that buy-side analysts
take a more active role in acquiring information (i.e., asking questions) when the company’s
information environment is poor and its future performance is uncertain. Consistent with this
prediction, we find that buy-side analysts are more likely to participate when a company has
lower sell-side analyst coverage, when there is greater uncertainty in earnings forecasts made by
sell-side analysts, and where a company misses consensus earnings forecast. Companies with
buy-side analysts in the conference call also tend to be older, have higher book-to-market ratios,
and experience more negative earnings surprises.
In our next analysis, we investigate whether conference call participation by buy-side
analysts is indicative of their investment firms’ trading of the shares of the company hosting the
3 There were 35 buy-side analysts from 17 investment firms voted as the “Best of the Buy-Side” between 2003 and 2008. We find that eight of these analysts are in our sample of earnings conference call transcripts. One of them is described as having little time to waste because he covers 55 companies; the time and effort he allocates to listen and participate in a company’s earnings conference call suggests that conference call participation is not a trivial task (Martin 2005).
4
conference call. Using pre- and post-conference call ownership data, we examine changes in
quarterly institutional ownership to better understand whether investment firms tend to change
their shareholdings after their buy-side analysts participate in the conference calls. We use a
difference-in-differences approach with a control sample of institutions that did not have buy-
side analysts participating in a conference call to examine if conference call participation leads to
a greater degree of trading after the call. This design allows us to examine differences in
ownership changes between the same set of institutions and companies across quarters in which
the key difference was participation in the conference call.
We find that of the nearly 18,000 instances in which a buy-side analyst asks a question on
a conference call, 55% of the time the institution employing the buy-side analyst does not own
the company’s stock as of the prior calendar quarter (i.e., a “non-owning institution”) and the
remaining 45% of participating buy-side analysts are from owning institutions. In either case, we
find that institutions with participating analysts on the conference call are more likely to change
their ownership compared to the control group of institutions without buy-side analyst
participation. In addition, institutions with buy-side analyst participation increased or decreased
their shareholdings to a greater degree compared to the control group. These results hold in both
univariate tests and in regressions in which we control for other factors that may be associated
with changes in institutional ownership. Overall, we interpret our results as suggesting that
institutions and their buy-side analysts view conference calls as a low-cost method to collect new
information to update their research and that buy-side analysts’ participation is indicative of
institutions’ subsequent trading in the company’s stock.
In our final set of analyses, we test whether buy-side analyst participation is associated
with future changes in stock prices, trading volume, total institutional ownership, and short
5
interest of the company hosting the conference call. We predict that the buy-side analysts we
observe in conference calls, and the employing institutions they represent, are reflective of other
buy-side analysts and institutions that we do not observe. Our premise is that the investment
decisions of similar institutions should be driven by the same underlying economic factors. If
this is the case, then we should find that buy-side analyst participation is not only associated with
institution-level changes in ownership, but also company-level changes in stock prices, trading
volume, total institutional ownership, and short interest. Our results support this hypothesis, as
the number of buy-side analysts participating on the conference call is positively associated with
future absolute returns, absolute changes in share turnover, absolute changes in institutional
ownership, and absolute changes in short interest. Overall, this set of results provides support for
our prediction that the buy-side analysts we observe in conference calls and the subsequent
trading decisions by their employing institutions are reflective of other buy-side analysts and
institutions that we do not observe.
This study contributes to the literature by examining the role of buy-side analysts in the
capital markets. While a handful of prior papers have examined the research outputs of buy-side
analysts, little to no work has focused on the inputs of buy-side research.4 Using a broad sample
of transcripts of earnings conference calls, we highlight that: 1) participation by buy-side
analysts is not uncommon; 2) buy-side analysts are more likely to participate when the
company’s information environment is poor and when it misses consensus earnings forecasts; 3)
buy-side analyst participation is indicative of subsequent trading by their investment firms,
suggesting that the finding of Rebello and Wei (2014) and Frey and Herbst (2014) that buy-side
4 Brown et al. (2014) uses survey data to shed light on the incentives that shape buy-side research. In addition, two working papers (Shohfi 2015; Cen, Dasgupta and Ragunathan 2011) examine the implications of buy-side analysts’ participation in earnings conference calls using archival data. However, the research questions we examine and our sample, research design, and variables of interest are different from these studies.
6
analyst recommendations influence the investment firm’s trading decisions is generalizable to a
large sample of investment firms; and 4) the actions we observe for buy-side analysts on
conference calls are associated with the hosting company’s subsequent absolute stock return,
trading volume, total institutional ownership, and short interest.
This paper continues as follows. The next section reviews the literature and develops
testable hypotheses. Section 3 describes the sample and variable construction. Section 4 presents
the empirical findings. We discuss sensitivity and robustness checks in Section 5 and conclude in
Section 6.
2. Institutional Background, Literature Review and Hypothesis Development
2.1 Institutional Background
A buy-side analyst works for an institutional investment firm, which explains the “buy-
side” moniker. Some of the largest investment firms, according to Institutional Investor
magazine’s annual ranking, include Barclays Global Investors (now BlackRock), State Street
Global Advisors, Fidelity Investments, Capital Group Companies, and J.P. Morgan Asset
Management (Capon, 2005). These firms and other smaller investment firms typically employ a
team of buy-side analysts to analyze industry data and individual companies and make stock
recommendations to their firms’ portfolio managers. Each buy-side analyst covers approximately
40 companies broadly grouped within a single sector, with many more companies on their
“radar” (Retkwa 2009; Abramowitz 2006), which differs from sell-side analysts who typically
cover approximately 20 companies grouped within a narrow industry. Daily responsibilities
include analyzing company financial statements and disclosures, meeting with company
7
executives, and communicating with their sell-side counterparts to supplement their own
research.
Sell-side analysts are considered information intermediaries who gather information
about companies through many sources (such as conference calls), process and interpret that
information, and disseminate that information to institutional clients. Buy-side analysts, on the
other hand, are typically private in their information gathering activities. Thus, their observable
participation in conference calls makes for a unique research setting that has not been previously
explored. Perhaps the biggest difference between buy-side and sell-side analysts is that the
former are held accountable for their stock picks (Knox and Kenny, 2003; Brown et al. 2014).
This fact suggests that the research activities buy-side analysts conduct prior to recommending
that a portfolio manager buy or sell a particular stock are of utmost importance to the analysts. In
this study, we examine participation in earnings conference calls to better understand its
importance to buy-side analysts as a research activity and information source.
2.2 Literature Review
Despite the rise in institutional ownership in U.S. equities over the last several decades
(Gompers and Metrick, 2001), there is very little research on the buy-side analysts who influence
the investment decisions of portfolio and asset managers. This lack of research is due to a lack of
data on the activities of buy-side analysts and the inputs and outputs of their research. As a
result, the few studies that have been published are based on proprietary data usually obtained
from one investment firm.
Most of the prior studies have focused primarily on the outputs of buy-side research:
earnings forecasts and stock recommendations. Using proprietary data from one buy-side firm,
Groysberg, Healy and Chapman (2008) find that buy-side analysts make more optimistic and less
8
accurate forecasts than their sell-side counterparts and attribute this result to the different
benchmarks used by buy-side and sell-side firms to evaluate their analysts. Groysberg, Healy,
Serafeim, and Shanthikumar (2013) find that buy-side analysts issue less optimistic
recommendations, but the performance of the stocks recommended is not different from that of
their sell-side counterparts, after controlling for the different types of stocks that the two sets of
analysts cover. Other studies find that the stock recommendations of buy-side analysts are useful
and more valuable than those of sell-side analysts. Cheng, Liu, and Qian (2006) model a fund
manager’s optimal reliance on both buy-side and sell-side research and show that a fund manager
relies more on buy-side research when its quality increases relative to sell-side research and
when the bias or uncertainty in sell-side research increases. Survey data support the prediction of
their model. Empirical studies generally find that buy-side recommendations are associated with
fund manager trades and positive abnormal returns for the funds (Rebello and Wei, 2014; Frey
and Herbst, 2014). Once again, the generalizability of these studies is limited, as they are based
on data from a single buy-side firm or survey data.
In contrast to studies that focus on the outputs of buy-side research, Brown, Call,
Clement and Sharp (2014) survey 344 buy-side analysts from 181 investment firms and conduct
follow-up interviews with 16 analysts to gain insights about the inputs and incentives that shape
buy-side research. They highlight that buy-side analysts view 10-K and 10-Q filings to be more
useful than other forms of disclosure in producing their stock recommendations. Specifically
referring to earnings conference calls, statements from a few interviewees suggested that buy-
side analysts do not participate (i.e., ask questions) on conference calls to avoid revealing their
private information (e.g., ideas, thoughts, opinions, etc.) to the public. While we expect some
buy-side analysts take this position and never participate in earnings conference calls, our sample
9
of conference call transcripts includes nearly 18,000 instances in which a buy-side analyst asked
a question. As a result, we believe that examining the reasons why some buy-side analysts
participate in earnings conference calls and how that participation is associated with future
institutional investment decisions advances our understanding of the role that buy-side analysts
play in the capital markets.
2.3 Hypothesis Development
Earnings conference calls are typically considered venues for sell-side analysts and
company managers to interact. As a result, there is a large literature that examines conference
calls as a voluntary disclosure medium and a source of information for sell-side analysts (Tasker
1998; Frankel, Johnson, and Skinner 2001; Bowen, Davis, and Matsumoto 2002; Bushee,
Matsumoto, and Miller 2003, 2004). However, our data show that buy-side analysts also
participate in the question and answer portion of conference calls, which suggests that the call is
also a source of information for buy-side analysts. Given that Brown et al. (2014) show, using
survey and interview data, that some buy-side analysts are not inclined to participate because of
the public nature of the conference call, we posit that buy-side analysts are more inclined to ask
questions under certain conditions.
When a company’s information environment is poor, typically characterized by little to
no coverage by sell-side analysts, we expect more buy-side analysts must gather their own
information. Conversely, when there is a high level of coverage by sell-side analysts, buy-side
analysts can rely on their sell-side counterparts for industry knowledge, access to management,
and company-specific information. We also expect that even if a company has sell-side
coverage, greater uncertainty about firm fundamentals can be a reason for buy-side analysts to
ask questions in a conference call. These predictions are supported by the model in Cheng, Liu,
10
and Qian (2006), which shows that institutional fund managers rely more on buy-side research
when the quality of sell-side research decreases and the uncertainty in sell-side earnings forecasts
increases. For these reasons, our first prediction is as follows:
Prediction 1: Buy-side analysts are more likely to participate in a company’s earnings conference call when the company’s information environment is poor.
Using proprietary data from one institutional investment firm, Rebello and Wei (2014)
and Frey and Herbst (2014) find that buy-side analyst recommendations influence the investment
firm’s trading decisions. Anecdotal evidence suggests that buy-side analysts cover approximately
40 companies, but keep an additional 40 companies on their “radar” (Retkwa 2009; Abramowitz
2006). When they become interested in one of the companies on their radar, they focus their
attention to conduct more due diligence on that company. If participation in a company’s
earnings conference call is part of a buy-side analyst’s due diligence process in forming or
updating his or her stock recommendations, then it follows that conference call participation will
be a precursor to trading by the buy-side analyst’s investment firm. Specifically, for those
institutions that own the company’s stock as of the calendar quarter prior to the conference call
(i.e., owning institutions), we predict they are more likely to increase or decrease their
shareholdings by the next calendar quarter and by greater amounts compared to a control group
of owning institutions without buy-side participation on the conference call. For those
institutions that do not own the company’s stock prior to the conference call (i.e., non-owning
institutions), we predict that they are more likely to establish shareholdings by the next calendar
quarter and by greater amounts compared to a control group of non-owning institutions without
buy-side participation. Thus, our second prediction is as follows:
Prediction 2: Institutional investment firms trade more of a company’s stock after their buy-side analysts participate in the company’s conference call, relative to institutional investment firms without buy-side analyst participation.
11
Given prediction 2 that institutional investment firms trade more of a company’s stock
after their buy-side analysts participate in the company’s conference call, we next examine
whether these trades are reflective of the actions of other institutional investors. We conjecture
that the economic factors that motivate the institutions we observe to trade the stock also
motivate other institutions to make similar trades even though they do not have buy-side analysts
participating on the conference call. Therefore, the trading decisions of the institutional
investment firms with participating buy-side analysts on the conference call may serve as a proxy
for the trading decisions of other institutional investors not participating on the call.5 Under this
conjecture, we will observe changes in ownership not only at the institution level, but also at the
company level.
For example, if we observe one non-owning institution with a buy-side analyst on the
conference call establish an initial ownership position in the company’s stock, and this trading
decision is in fact reflective of the trading decisions of other non-owning institutions, then we
will observe an increase in total institutional ownership after the conference call. Similarly, if we
observe one owning institution (with a buy-side analyst on the call) sell the stock after a
conference call and other owning institutions do the same, then we will observe a decrease in
total institutional ownership. Finally, if we observe a non-owning institution not establish an
initial ownership position after the conference call, then that institution may be increasing or
decreasing a short position in the stock, which may proxy for the actions of other non-owning
5 There is empirical support for our conjecture that the buy-side analysts participating on the call, whether they work for institutions that own or do not own the stock prior to the call, are reflective of other buy-side analysts not participating on the call. Heinrichs, Park, and Soltes (2015) examine market participants who access earning conference calls (rather than ask questions on the calls) and find that more than half of the “consumers” of earnings conference calls are buy-side analysts, and roughly half of the buy-side analysts work for institutions that own the stock prior to the conference call. Therefore, given the large amount of interest in the calls from buy-side analysts from both owning and non-owning institutions, it is likely that the actions we observe for a few buy-side analysts who participate are reflective of the intentions for a larger number of unobserved buy-side analysts.
12
institutions. In any of these scenarios, we will also observe an increase in absolute stock return
and trading volume. Thus, our final prediction is as follows.
Prediction 3: Participation by buy-side analysts on a company’s earnings conference call is positively associated with future absolute stock return, trading volume, and changes in total institutional ownership and short interest.
3. Sample data
Our data is comprised of companies with available conference call transcripts from the
Thomson Reuters StreetEvents database from the second quarter of 2002 through the first quarter
of 2009. As shown in Table 1, Panel A, our full sample includes 56,285 earnings conference
calls from 3,414 companies.6
The transcripts contain the name and affiliation of anyone who asked a question during
the question and answer (Q&A) portion of the call. There are a total of 381,951 questioners in
our sample, roughly seven per conference call. To identify buy-side analysts, we first search all
affiliations for words that are common in the names of institutional investment firms including
“capital,” “asset,” “fund,” “investment,” “management,” “advisor,” “partner,” and “investor.”
Second, we exclude all affiliations that are known to be sell-side brokerage firms and investment
banks based on data from I/B/E/S and our own internet searches of sell-side firms. Third, with
the list of affiliations obtained from the first step and the remaining affiliations after the second
step, we manually match the affiliations to institutional investors in the Thomson Reuters
database of 13F filings.7 Through this procedure, we identify 17,691 questioners, or 4.6% of the
6 We require that the date of a company’s conference call (from Thomson Reuters) be the same or one day after the date of the earnings announcement provided by Compustat. We find that 78% of the conference calls occur on the same date as the earnings announcement and 22% occur on the next day. 7 Form 13F is the reporting form filed by institutional investment managers pursuant to Section 13(f) of the Securities Exchange Act of 1934. Institutional investment managers that use the United States mail (or other means or instrumentality of interstate commerce) in the course of their business and that exercise investment discretion over $100 million or more in Section 13(f) securities must file Form 13F.
13
total, as working for institutional investment firms.8 To estimate the number of unique buy-side
analysts, we compute the number of unique caller names from each investment firm, while
allowing for some variation and misspellings of names.9 We estimate there are 4,045 unique buy-
side analysts from 724 institutional investment firms in our sample.10 The top 45 institutional
investment firms ranked by total number of conference calls, along with their number of buy-
side analysts, is shown in Appendix 1.
We find that 24% of the earnings conference calls in our sample, or 13,373 calls, have at
least one buy-side analyst who asked a question. Table 1, Panel B, shows the percentage of
conference calls with buy-side analyst participation by year and calendar quarter. This
percentage was higher in the first half of our sample period, ranging from 25% to 30%, and has
settled to about 20% to 21% in the latter part of our sample period. Table 1, Panel C, shows that
there are 10,318 conference calls with one buy-side analyst, 2,404 conference calls with two
buy-side analysts, 521 conference calls with three buy-side analysts, and 130 conference calls
with four or more buy-side analysts. In our determinants tests discussed in the next section, we
combine the conference calls with two or more buy-side analysts participating into a single group
of 3,055 conference calls.
In addition to conference call data, we require other data sources for our empirical tests.
We use I/B/E/S data to compute sell-side analyst coverage and earnings forecast variables,
Thomson Reuters 13F filings data to compute companies’ quarterly institutional ownership,
8 Without the requirement of matching buy-side affiliations to the Thomson Reuters 13F filings database, we actually identity 22,800 buy-side analysts, or 5.9% of the total number of questioners. 9 Specifically, we compute the number of unique names per investment firm using only the first four letters of the analyst’s name. For example, “Stephan Smith” would be considered the same person as “Stephen Smith” because the first four letters of each name is “Step.” However, it is still possible for common names to vary in spelling within the first four letters, such as “John” and “Jon.” For such cases, our estimate of the number of unique buy-side analysts in our sample would be overstated. 10 In untabulated analyses, we estimate the average portfolio value for each of the 724 institutional investment firms during our sample period and find that most of them fall into the 3rd or 4th quartile (4th being the highest) in terms of total portfolio size among all institutional investment firms in the Thomson Reuters 13F database.
14
Compustat data to compute companies’ quarterly financial variables and monthly short interest,
and CRSP data to compute stock returns and trading volume. All variable definitions are shown
in Appendix 2.
4. Empirical Analysis
4.1 Determinants of Buy-Side Analyst Participation on Conference Calls
To test our first prediction, we run an ordered logistic regression with three ordinal levels
on the dependent variable NUMBUYSIDERS (Green, 1990; Allison 1999). The first level is
zero buy-side analysts on the conference call, the second level is one buy-side analyst, and the
third level is two or more buy-side analysts on the call (i.e., NUMBUYSIDERS=0, 1, or 2+).11
Our regression equation is as follows:
Prob[NUMBUYSIDERS=0, 1, or 2+]i,t = β0 + β1NUMSSANALYSTSi,t
+ β2DISPERSIONi,t + βkControl Variablesi,t +Year Fixed Effects + εi,t (1)
Our first prediction is that buy-side analysts are more likely to participate on a company’s
earnings conference call when the company’s information environment is poor. We use the level
of sell-side analyst coverage and forecast dispersion to proxy for a company’s information
environment. For each company i and quarter t, we measure the level of sell-side analyst
coverage with the variable NUMSSANALYST, defined as the number of unique sell-side
analysts in the I/B/E/S detailed earnings per share (EPS) database that provided any type of EPS
forecast for the company from the prior conference call date to one day before the current
conference call date. For companies with no EPS forecasts in a given quarter, we set
NUMSSANALYST to zero. We measure sell-side analyst forecast dispersion, DISPERSION, as
11 We also run OLS regressions in which the dependent variable ranges from zero to seven buy-side analysts in the conference call. The inferences from the OLS regressions are very similar to those from the ordered logistic regressions. For brevity, the results of the OLS regressions are not tabulated but are available upon request.
15
the standard deviation of analysts’ current quarter EPS forecast measured over the same period
as NUMSSANALYST. For companies with no EPS forecasts or one forecast from a single sell-
side analyst, we cannot compute a standard deviation. Therefore, to avoid losing observations of
companies with coverage from zero or one sell-side analyst, we set DISPERSION equal to the
mean value for the sample. 12 Our predictions are that the estimated coefficients for
NUMSSANALYST and DISPERSION are negative and positive, respectively (β1<0 and β2>0).
We include several variables to control for other factors that may be associated with
interest in the company in general and interest in the earnings conference calls in particular. We
control for total institutional ownership with the log of the number of institutional investors in
the company (NUMINSTINV), as the more institutional investors a company has, the more
likely it is that an institutional investor will participate on the conference call. We control for
company size using the log of market value of equity (COMPANYSIZE), company age using the
log number of months that the company has been listed in CRSP (COMPANYAGE), the
company’s book-to-market ratio (book value of equity divided by market value of equity), two
indicator variables for whether the company had a positive or negative earnings surprise
(POSEPSSURPRISE and NEGEPSSURPRISE), and the company’s stock return over the 90-day
period prior to the conference call (RETPRIOR90DAYS). To control for a time-of-day effect, in
which there is possibly more interest in the conference call if it occurs before or during trading
hours, we include an indicator variable (INTRADAY) for whether the start of the call occurs
between 6:00am and 3:45pm Eastern Time. Lastly, we include the absolute value of future
12 We set DISPERSION to the mean value for 9,902 observations out of 56,285 to avoid losing observations in which a company has no coverage from sell-side analysts or coverage from only one sell-side analyst. If we leave these observations as missing values, then the sample sizes for the ordered logistic regressions presented in Table 2, Panel C are reduced by an equivalent amount and the sample would be biased against companies with low or zero sell-side analyst coverage. In such regressions, the signs and statistical significance of the coefficients for the variables of interest remain similar to those presented in Table 2, Panel C.
16
three-month returns (ABSRET) to proxy for potential mispricing in the company’s stock at the
time of the conference call that would attract buy-side interest in the company. All variable
definitions are shown in Appendix 2.
We repeat our regressions using variants of our dependent variable, again coded for three
ordinal levels (0, 1, or 2+), to shed additional light on the determinants of buy-side participation
of various types. In addition to the number of buy-side analysts (NUMBUYSIDERS), we use the
number of buy-side analysts from owning institutions (NUMBUYSIDERS_OWN) and non-
owning institutions (NUMBUYSIDERS_NOTOWN) to examine whether prior ownership
matters in determining buy-side participation.
Descriptive statistics of the variables used in our determinants test are shown in Table 2,
Panel A. All continuous explanatory variables are winsorized at the 1st and 99th percentiles. The
mean values for the number of sell-side analysts (NUMSSANALYSTS) and forecast dispersion
are 7.6 and 0.04, respectively. Companies have on average 97 (exp(4.58)-1) institutional
investors. The mean company market capitalization (COMPANYSIZE) is roughly $1 billion
(exp(6.94) in $ millions) and company age is 139 months (exp(4.93)) or 12 years.
Panel B shows mean and median values, along with tests for differences, of the variables
for companies partitioned by whether buy-side analysts participated on their conference calls.13
The mean (median) number of sell-side analysts covering companies with buy-side analyst
participation is 6.3 (5.0), which is lower than 8.0 (6.0) for companies without buy-side
participation. The mean (median) forecast dispersion for companies with buy-side analysts is
0.044 (0.033), which is higher than 0.038 (0.024) for companies without buy-side participation.
Differences in values in the control variables indicate that companies with buy-side analyst
13 To ease presentation of univariate differences, we form two groups only (0 and 1 or more buy-side analysts on a call). In the subsequent ordered logistic regression, we form three levels of the dependent variable (0, 1, and 2 or more).
17
participation tend to be smaller in market capitalization, are older, have higher book-to-market
ratios, have higher prior returns, and have fewer positive and more negative earnings surprises.
All differences in the mean and median values for each group are significant at the 1% level,
except for absolute future three-month returns (ABSRET). These univariate results support
Prediction 1 and are consistent with buy-side analysts participating on earnings conference calls
of companies with lower sell-side analyst coverage and higher earnings forecast dispersion—
proxies for a poor information environment.
We next test our prediction in a multivariate setting. The results of estimating regression
equation (1) are presented in Table 2, Panel C, Columns (1) through (3). Year fixed effects are
included and robust standard errors are clustered by companies; z-statistics are presented in
parentheses below the coefficients. Across all columns, where the dependent variable is the
number of buy-side analysts in general, buy-side analysts from owning institutions, and buy-side
analysts from non-owning institutions, respectively, the estimated coefficient for the number of
sell-side analysts (NUMSSANALYSTS) is significantly negative. For example, the estimated
coefficient in Column (1) of −0.064 indicates that an interquartile decline in the number of sell-
side analysts from 11 to 3 (Table 2, Panel A), holding all other variables fixed, corresponds to a
50% [(exp(−0.064)−1)*(−8)] increase in the odds of the number of buy-side analysts on the call
being at a higher level (2+ vs. 1 or 1 vs 0). Taken together, these results indicate that buy-side
analysts (of each type) are more likely to participate on a company’s earnings conference call
when the company has less sell-side analyst coverage, consistent with the first part of Prediction
1.
The results for earnings forecast dispersion (DISPERSION) are consistent with the
second part of Prediction 1. Across all columns, the estimated coefficient for DISPERSION is
18
significantly positive. The coefficient in Column (1) of 1.697 indicates that an interquartile
increase in the dispersion of forecasts from 0.01 to 0.04, holding all other variables fixed,
corresponds to a 13% [(exp(1.697)-1)*0.03] increase in the odds of the number of buy-side
analysts on the call being at a higher level. The results reported under Columns (1) to (3) suggest
that more buy-side analysts (of each type) are likely to participate on a company’s earnings
conference call when forecast dispersion is higher. Overall, the results for our variables of
interest in Panel C are consistent with buy-side analysts taking a more active role in acquiring
information (i.e., asking questions) during earnings conference calls when the company’s
information environment is poor.
Regarding the control variables, there are several notable results. The positive
coefficients for COMPANYAGE, BOOK-TO-MARKET, NEGEPSSURPRISE, and
INTRADAY in most columns are consistent with the univariate results in Panel B and indicate
that buy-side analysts tend to participate on the calls of older, more value-oriented companies,
when earnings news is negative, and when the call occurs before or during market trading hours.
The positive coefficient for COMPANYSIZE differs from the univariate results and indicates
that after controlling for the other factors, more buy-side analysts tend to participate on the
conference calls of larger companies. Lastly, the positive coefficient for NUMINSTINV in
Column (2) indicates that more buy-side analysts from owning institutions tend to be on the call
when a company has more institutional investors, while the negative coefficient for the same
variable in Column (3) indicates that fewer analysts from non-owning institutions tend to be on
the call when a company has more institutional investors.14
14 We also estimate equation (1) with firm fixed effects included. Untabulated results indicate that all estimated coefficients have smaller t-statistics than those reported in Table 2, Panel C. The estimated coefficients on the two main variables of interest, NUMSSANALYSTS and DISPERSION, are still significant, while those on many of the control variables become insignificant.
19
4.2 Institutional Trading Subsequent to Buy-Side Analyst Participation on Conference Calls
In this section we examine whether the institutional investment firms with buy-side
analysts on companies’ conference calls subsequently trade (and by what amounts) the stock of
the company hosting the conference call. The purpose of this analysis is to better understand
whether buy-side analyst participation is a precursor to institutional trading in the company’s
stock and to investigate whether the results documented in Rebello and Wei (2014) and Frey and
Herbst (2014) can be generalized from one investment firm to a larger sample of investment
firms. Our second prediction is that participation of buy-side analysts on a company’s earnings
conference call is associated with greater changes in ownership by the institutional investment
firm employing the analysts than if there were no participation.
We use a difference-in-differences approach with a control sample of owning and non-
owning institutions that did not have buy-side analysts on a conference call to examine if
conference call participation leads to a greater degree of trading. Using data from the Thomson
Reuters 13F filings database, we compute the percentage ownership that each institution has for
each company in our sample, measured at the calendar quarters ended before and after the
conference call, to compute changes in quarterly ownership. We note that ownership and
changes in ownership can be zero as well as non-zero for any company-institution pair in any
given quarter. With a sample of 56,285 conference calls (involving 3,414 companies) and 724
institutional investment firms, there is a total of 40,750,340 (56,285*724) possible company-
institution-quarters for our analyses.15 However, we eliminate about 30% of these observations in
15 Another approach is to expand the number of control observations to include all institutions in the Thomson Reuters 13F database (not just the ones we observe in transcripts) during our sample period. There are 4,003 such
20
which an institution did not appear to exist (i.e., not in the 13F database) at the time of a
company’s conference call, which leaves approximately 28 million observations in which any of
the institutions in our sample could have plausibly owned or not owned the company hosting the
conference call.
Table 3, Panel A presents summary statistics on the percentage of owning and non-
owning institutions with and without buy-side analysts on a conference call that subsequently
increased, decreased, or did not change their level of ownership in the company. In addition, the
mean and median of the changes are presented for both the institutions with and without buy-side
analyst participation and tests for differences. Of the 7,890 buy-side analysts from owning
institutions that participated on a conference call, 45% of the institutions increased their
ownership, 49% decreased their ownership, and 6% did not change their ownership after the call.
Of the owning institutions without buy-side analyst participation, 41% increased their ownership,
50% decreased their ownership, and 9% did not change their ownership. Thus, the comparison of
owning institutions suggests that there are slightly more changes in ownership (94% vs. 91%)
among institutions with buy-side analysts on the conference call. Of the 9,801 buy-side analysts
from non-owning institutions that participated on a conference call, 8% of the institutions
established initial ownership in the company. By comparison, only 1% of non-owning
institutions without buy-side analyst participation established initial ownership in the company.
Focusing on institutions that increased ownership, we find that owning institutions with
buy-side analysts on the call had a mean and median increase of 0.5%, compared to a mean
(median) increase of 0.2% (0.1%) for owning institutions without buy-side analysts on the call.
The difference in differences is significant at the 1% level using a two-tailed t-test for means and
institutions, which would result in 225,308,855 (56,285*4,003) possible company-institution-quarters for our analyses. Results are similar as those presented in Table 3 when using this expanded control set.
21
a Wilcoxon signed-rank test for medians. Non-owning institutions with buy-side analysts on the
call had a mean (median) ownership of 1.2% (0.4%) after the call, compared to a mean (median)
ownership of 0.3% (0.1%) for non-owning institutions without buy-side analysts on the call. The
difference in differences is significant at the 1% level for both the mean and median. Regarding
decreases in percentage ownership among owning institutions, we find that owning institutions
with buy-side analysts on the call had a mean and median decrease of 0.7%, compared to a mean
(median) decrease of 0.2% (0.1%) for owning institutions without buy-side analysts on the call.
The difference in differences is again significant at the 1% level for the mean and median.
The results in Panel A indicate that when a buy-side analyst participates on a company’s
earnings conference call, the analyst’s investment firm will alter its ownership of the company’s
stock by a greater degree than when its buy-side analyst does not participate on the call. Overall,
this evidence is consistent with the buy-side analysts using conference calls as an avenue to
acquire information to update their recommendations and to identify new investment
opportunities, which often leads to subsequent trading in the company’s stock.
We next test Prediction 2 in a multivariate setting. We regress the absolute change in
percentage ownership that institution j has in company i in quarter t (ABSCHGOWNERSHIPi,j,t)
on an indicator variable (PARTICIPATEINCALLi,j,t) for whether a buy-side analyst employed
by institution j participated on company i's conference call in quarter t.
ABSCHGOWNERSHIPi,j,t = β0 + β1PARTICIPATEINCALLi,j,t
+ βkControl Variablesi,t + Year Fixed Effects + εi,t (2)
If conference call participation leads to greater changes in ownership, then the coefficient for
PARTICIPATEINCALL will be significantly positive (β1>0).
22
We also control for several institution- and company-specific factors that may be
associated with institutions’ quarterly changes in ownership in companies. In particular, changes
in ownership may be related to the size of an institution’s shareholding in a given company and
the size of the institution’s total portfolio. Accordingly, we include the variable
VALUEOFOWNERSHIP, defined as the log of the dollar value of ownership that institution j
has in company i, and INVFIRMSIZE, defined as the log of total dollar value of all the
investment firm’s shareholdings. We also include NUMFIRMSINPORT as the log of the total
number of companies in an institution’s portfolio to capture the amount of attention (or lack of)
that a company may receive from the portfolio manager. Finally, we include a company’s
absolute stock return from the prior 90 calendar days (ABSRETPRIOR90DAYS).16 All control
variables are measured as of the calendar quarter before the conference call and are shown in
Appendix 2.
The results of estimating regression equation (2) are presented in Table 3, Panel B.
Standard errors are clustered by institutional investment firm and year fixed effects are included.
The estimated coefficient for PARTICIPATEINCALL is significantly positive, indicating that
when buy-side analysts participate on a company’s conference call, the analyst’s institution tends
to increase or decrease its ownership in the company by a greater amount. This result is
consistent with Prediction 2 and with the univariate results in Panel A. Among the control
variables, we find that there is also a greater absolute change in ownership when the value of
shareholdings is greater, when the institutional investment firm is larger, and when absolute prior
returns are greater. When there is a greater number of companies in an institution’s portfolio,
there tend to be smaller changes in ownership.
16 Prior returns are not available for all company-institution-quarters used in the univariate analyses of Table 3 Panel A, which results in a relatively smaller sample size in for the regression.
23
4.3 Conference Call Participation and Future Absolute Returns, Share Turnover, Institutional
Ownership, and Short Interest
Thus far we have examined the participation of individual buy-side analysts on
companies’ earnings conference calls and the subsequent trading decisions of individual
institutional investment firms that employ the analysts. However, it is unclear whether such
decisions at the institution level are noticeable at the company level, where a company may have
hundreds or thousands of institutional investors. Even if we observe one institution increase or
decrease its ownership in a company, a similar decision by other institutional investors is needed
to move the company’s stock price, trading volume, total institutional ownership, and short
interest. We conjecture that the investment decisions of similar institutions are driven by the
same underlying economic factors and that buy-side analyst participation is reflective of broader
interest by many institutional investors. Hence, buy-side analyst participation will be associated
with company-level changes in absolute stock price, trading volume, total institutional
ownership, and short interest.
To test this conjecture, we regress a company’s future absolute stock returns, absolute
changes in share turnover, absolute changes in total institutional ownership, and absolute
changes in short interest on the number of buy-side analysts on the company’s conference call
and control variables. The basic regression equation is as follows.
FUTURECHANGEi,t = β0 + β1NUMBUYSIDERSi,t + βkControl Variablesi,t
+ Year Fixed Effects + εi,t (3)
FUTURECHANGE represents ABSRET, ABSCTURNOVER, ABSCPIO, and
ABSCSHORTINT. For each company i and quarter t, ABSRETi,t is the company’s absolute
24
stock return over the 90 calendar days after its conference call, ABSCTURNOVERi,t is the
absolute daily average share turnover percentage over the 90 calendar days after a conference
call less the daily average share turnover percentage over the 90 calendar days before the
conference call, ABSCPIOi,t is the absolute change in total percentage institutional ownership
from the calendar quarter ended before the conference call to the calendar quarter ended after the
conference call, and ABSCSHORTINTi,t is the absolute change in short interest as a percentage
of total shares outstanding from the month of the prior conference call to the month of the
current conference call. Regarding the third dependent variable (ABSCPIO), while we predict
that changes in individual institution’s ownership in a company are associated with changes in
the company’s total institutional ownership, we do not expect there to be a mechanical relation.
But to alleviate that potential concern, we compute ABSCPIOi,t excluding the institutions with
buy-side analysts on company i’s conference call in quarter t to capture changes in ownership by
institutions without buy-side analysts on the call.
We include control variables that may be associated with future absolute stock returns,
absolute changes in share turnover, absolute changes in total institutional ownership, and
absolute changes in short interest. ABSEPSSURPRISE is defined as the absolute value of actual
reported EPS minus the most recent sell-side analyst mean consensus forecast prior to the
earnings announcement. We also include company size and the book-to-market ratio as
previously defined. To control for stock return momentum, we use RETPRIOR11MO as the
stock return for the 11-month period starting one year prior to the conference call data and
ending one month prior to the conference call date. We control for increases and decreases in
sell-side analyst interest in a conference call using NC_ANALYSTS and
COV_ANALYSTS_ABSENT, as defined in Jung, Wong, and Zhang (2015). Specifically,
25
NC_ANALYSTS is the number of non-covering sell-side analysts on the conference call and
COV_ANALYSTS_ABSENT is the number of covering sell-side analysts who are absent from
the current conference call (but participated on the prior conference call), both scaled by the
number of sell-side analysts on the conference call.17 To control for other possible omitted
factors that may be associated with each of our dependent variables, we include a lagged version
(measured for the prior quarter) of the dependent variable (ABSRET_LAG,
ABSCTURNOVER_LAG, ABSCPIO_LAG and ABSCSHORTINT_LAG). Descriptive
statistics of the variables used in these regressions are shown in Table 4, Panel A.
For the final control variable, we include the inverse mills ratio (IMR) to control for
potential sample selection bias related to buy-side analysts’ decision to participate on certain
companies’ earnings conference calls but not others. That is, since we do not observe when buy-
side analysts do not participate in a call (but are perhaps listening), our dependent variable for
whether there are any buy-side analysts on a conference call is zero for a significant fraction of
the observations. The results of our univariate and multivariate determinants tests presented in
Table 2 indicate that buy-side analysts tend to participate on the conference calls of companies
with poorer information environments and companies that are older, have higher book-to-market
ratios, and have negative earnings news. As a result, regressions of future absolute stock returns,
absolute changes in share turnover, absolute changes in institutional ownership, and absolute
changes in short interest without correcting for the potential selection bias may lead to biased
coefficient estimates (Heckman 1979). To compute the inverse mills ratio, we first model the
decision for a buy-side analyst to participate on a company’s conference call using a probit
17 As in Jung et al. (2015), the computation of COV_ANALYSTS_ABSENT requires one lag quarter of data to determine if the absent sell-side analysts were on the prior conference call. Thus, for the regressions presented in Table 4 in which COV_ANALYSTS_ABSENT is included, the sample period is from the third quarter of 2002 to the first quarter of 2009.
26
regression similar to equation (1) except that the dependent variable is binary (1 for any buy-side
analysts on the call and 0 for none).18 We then include the IMR from the probit regression into
regression equation (3) as a control variable.
The first set of results from estimating regression equation (3) are presented in Table 4,
Panel B, which shows Fama-MacBeth regressions in which the dependent variable is ABSRET
and the coefficient estimates are the average of quarterly estimates over 27 quarters from the
third quarter of 2002 to the first quarter of 2009. The estimated coefficient for
NUMBUYSIDERS is significantly positive at the 10% level. This result suggests that a
company’s post-earnings announcement return volatility is positively associated with the number
of buy-side analysts who participate on the conference call. In Panel C, Columns (1) through (3)
present the results of pooled regressions in which the dependent variable is the absolute change
in daily average share turnover (ABSCTURNOVER), absolute change in percentage institutional
ownership (ABSCPIO), and absolute change in short interest (ABSCSHORTINT), respectively.
Robust standard errors are clustered by companies. Across all columns, the estimated coefficient
for NUMBUYSIDERS is significantly positive at the 1% level. These results, along with those
for future return volatility, are consistent with Prediction 3 that buy-side analyst participation is
reflective of broader interest by many institutional investors.19
5. Sensitivity and Robustness Tests
5.1 Alternative specifications of variables of interest
18 For brevity, the results of this first-stage probit regression are not tabulated, but they are available upon request. 19 The estimated coefficients on IMR are statistically significant in all three sets of regressions reported in Panels B and C of Table 4. While the IMR variable is included to correct for self-selection in the estimation of equation (3), it can also be interpreted as an estimate of the private information underlying the decision of a buy-side analyst to participate on a conference call or not (see Li and Prabhala (2007) for a detailed explanation).
27
One of the explanatory variables of interest in Table 2, Panel C, is DISPERSION, defined
as the standard deviation of analyst EPS estimates, and it proxies for earnings uncertainty within
a firm’s information environment. Cheong and Thomas (2011) show that analyst EPS forecast
dispersion does not vary with scale, so we do not scale DISPERSION in our main tests. As a
robustness check, we define dispersion alternatively as the standard deviation of analysts’ EPS
estimates, scaled by the company’s stock price on the date of the conference call
(DISPERSIONSC). To avoid a small denominator problem, we require that the stock price be
equal to or greater than $1.00. Results using this alternative definition (not tabulated) are similar
to those presented in Table 2, Panel C, with the exception that the estimated coefficient for
DISPERSIONSC is insignificant in Column (2).
In Table 4, Panels B and C, our explanatory variable of interest is NUMBUYSIDERS,
which measures the level of buy-side analyst participation on companies’ conference calls.
Alternatively, we can define the absolute change in buy-side analyst participation
(ABSCNUMBUYSIDERS) to explain absolute changes in company-level stock prices, trading
volume, total institutional ownership, and short interest. Results using this alternative
specification (not tabulated) are similar to those presented in Table 4, Panels B and C.
5.2 Limiting sample of companies to those with sell-side analyst coverage of ten or fewer
The results in Table 2, Panels B and C, indicate that buy-side analysts are more likely to
participate in a company’s earnings conference call when there are fewer sell-side analysts that
cover the company. However, for companies that are covered by a large number of sell-side
analysts, there may be no opportunity for buy-side analysts to ask questions if sell-side analysts
get priority to ask a question and there is a time constraint on the conference call. Therefore, to
check the robustness of our results, we repeat our analysis using a subsample of firms that are
28
covered by ten (the third quartile value) or fewer sell-side analysts. This restriction eliminates
14,412 conference calls from 1,144 firms (out of 3,414) from our analysis. The results (not
tabulated) of estimating regression equation (1) using this subsample remain similar to those
presented in Table 2, Panel C despite the loss of statistical power. For example, the coefficient
for NUMSSANALYSTS in Column (1) is −0.056 with a Z-stat of −6.46, compared to the
coefficient of −0.064 with a Z-stat of −12.25 shown in Table 2, Panel C. Therefore, we do not
believe that time constraints on conference calls drive our main results.
5.3 Limiting conference calls to those near the beginning of a calendar quarter
The results in Table 3, Panel A, indicate that institutional investment firms are more
likely to change their level of shareholdings, and to a greater degree, after their buy-side analysts
participate on a company’s conference call. One caveat to this analysis is that since the Thomson
Reuters database of 13F filings shows quarterly shareholdings, we cannot identify whether
changes in shareholdings occur before or after (or both) the company’s conference call. As a
robustness check, we repeat this analysis (not tabulated) using a reduced sample in which the
dates of companies’ conference calls are close to the beginning of the calendar quarter, which
increases the likelihood that quarterly changes in ownership occur after the conference call. For
conference calls that are within three weeks of the beginning of the calendar quarter, the sample
size is reduced by 83% (from 17,691 to 2,961 buy-side analysts in a conference call), yet we still
find results very similar to those presented in Table 3 Panel A. For example, among owning
institutions with buy-side analyst participation, 46% increased their ownership, 51% decreased
their ownership, and 4% did not change their ownership. Among non-owning institutions without
buy-side analyst participation, 7% established new ownership in the company and 93% did not.
29
Results for institutions without buy-side analyst participation are also similar to those presented
in Panel A.
6. Conclusion
In this study, we examine a large sample of buy-side analysts who participated on
companies’ earnings conference calls to better understand the process in which buy-side analysts
gather information. Contrary to the common perception that earnings conference calls are venues
for sell-side analysts and company managers to interact, we find evidence consistent with the
notion that buy-side analysts participate in such calls when a company’s information
environment is poor. We also document that institutional investors tend to increase or decrease
their ownership in the stock to a greater degree after the conference call when their buy-side
analysts participate on the call. This result suggests that buy-side analysts use information
gathered from conference calls to revise or update their stock recommendations, which
ultimately leads to their institution trading the stock after the conference call. Finally, we find
associations between buy-side analyst participation on conference calls and companies’ future
absolute stock returns, share turnover, changes in total institutional ownership and short interest,
which we interpret as evidence that the buy-side analysts we observe on conference calls and the
subsequent trading decisions by their employing institutions are reflective of other buy-side
analysts and institutions that we do not observe. Overall, our study contributes to the literature on
buy-side analysts by highlighting earnings conference calls as one of their important research
activities, and by also highlighting the implications of their participation for the institutional
investment firms that employ them and for the companies that host the conference calls.
30
References
Abramowitz, P. 2006. “The Best of the Buy Side.” Institutional Investor, 40 (11): 42-52. Allison, P.D. 1999. “Logistic Regression Using SAS: Theory and Application.” Cary, NC: SAS
Institute Inc. Bowen, R.M., A.K. Davis, and D.A. Matsumoto. 2002. “Do Conference Calls Affect Analysts’
Forecasts?” The Accounting Review 77 (2): 285-316. Brown, L.D., A.C. Call, M.B. Clement, and N.Y. Sharp. 2014. “Skin in the Game: The Inputs
and Incentives that Shape Buy-Side Analysts’ Stock Recommendations.” Working paper, University of Texas at Austin.
Bushee, B., D. Matsumoto, and G. Miller. 2003. “Open versus Closed Conference Calls: The
Determinants and Effects of Broadening Access to Disclosure.” Journal of Accounting and Economics 34: 149-180.
Bushee, B., D. Matsumoto, and G. Miller. 2004. “Managerial and Investor Responses to
Disclosure Regulation: The Case of Reg FD and Conference Calls.” The Accounting Review 79 (3): 617-643.
Capon, A. 2005. “Hollowing out the core: America’s Top 300 Money Managers.” Institutional
Investor 39 (7): 45-63. Cen, L., S. Dasgupta, and V. Ragunathan. 2011. “Buy-Side Participation and Information
Production in Earnings Conference Calls.” Working paper, University of Toronto, HKUST, and University of Queensland.
Cheng, Y., M.H. Liu, and J. Qian. 2006. “Buy-Side Analysts, Sell-Side Analysts, and Investment
Decisions of Money Managers.” Journal of Financial and Quantitative Analysis 41 (1): 51-83.
Cheong, F.S., and J. Thomas. 2011. “Why Do EPS Forecast Error and Dispersion Not Vary with
Scale? Implications for Analyst and Managerial Behavior.” Journal of Accounting Research 49 (2): 359-401.
Crawford, S., W. Gray, B. Johnson, and R. Price. 2012. “The Investment Value of Contrarian
Buy-Side Recommendations.” Working paper, Rice University, Drexel University and Creighton University.
Fama, E. F., and J. D. MacBeth. 1973. “Risk, return, and equilibrium: Empirical tests.” Journal
of Political Economy 81 (3): 607–636. Frankel, R., M. Johnson, and D. Skinner. 1999. “An Empirical Examination of Conference Calls
as a Voluntary Disclosure Medium.” Journal of Accounting Research 37 (1): 133-150.
31
Frey, S., and P. Herbst. 2014. “The Influence of Buy-Side Analysts on Mutual Fund Trading.”
Journal of Banking and Finance 49: 442–458. Gompers, P.A., and A. Metrick. 2001. “Institutional Investors and Equity Prices.” The Quarterly
Journal of Economics 116 (1): 229-259. Green, W.H. 1990. “Econometric Analysis.” New York, NY. Macmillan Publishing Company. Groysberg, B., P. Healy, and C. Chapman. 2008. “Buy-Side vs. Sell-Side Analysts' Earnings
Forecasts.” Financial Analysts Journal 64 (4): 25-39. Groysberg, B., P. Healy, G. Serafeim, and D. Shanthikumar. 2013. “The Stock Selection and
Performance of Buy-Side Analysts.” Management Science 59 (5): 1062-1075. Heckman, J.J. 1979. "Sample Selection as a Specification Error." Econometrica 47 (1): 153–161. Heinrichs, A., J. Park, E.F. Soltes. 2015. “Who Consumes Firm Disclosures? Evidence from
Earnings Conference Calls.” Working paper, Columbia Business School. Jung, M.J., M.H.F. Wong, and X.F. Zhang. 2015. “Analyst Interest as an Early Indicator of Firm
Fundamental Changes and Stock Returns.” The Accounting Review 90 (3): 1049-1078. Knox, L., and J.B. Kenney. 2003. “The Best of the Buy Side.” Institutional Investor, 37 (11): 34-
42. Li, K. and N.R. Prabhala. 2008. “Self-Selection Model in Corporate Finance.” Handbook of
Empirical Corporate Finance, Vol. 1. Edited B. Espen Eckbo. Amsterdam: Elsevier B.V., 2008. 37- 86.
Martin, S. 2005. “The Best of the Buy Side.” Institutional Investor, 39 (11): 76-84. Rebello, M., and K.D. Wei. 2014. “A Glimpse Behind a Closed Door: The Long-Term
Investment Value of Buy-Side Research and Its Effect on Fund Trades and Performance.” Journal of Accounting Research 52 (3): 775-815.
Retkwa, R. 2009. “Blue-Ribbon Researchers.” Institutional Investor, 42 (11): 82-83. Shohfi, T. 2015 “Implications of Buy Side Analyst Earnings Conference Call Participation.”
Working paper, Rensselaer Polytechnic Institute. Tasker, S.C. 1998. “Bridging the Information Gap: Quarterly Conference Calls as a Medium for
Voluntary Disclosure.” Review of Accounting Studies 3: 137-167.
32
Appendix 1: Top 45 Investment Firms with Buy-Side Analysts on Conference Calls Institutional Investment Firm Number of Analysts Number of Conference Calls Lord Abbett 36 701 Kennedy Capital Management 25 290 Zimmer Lucas Partners 52 282 Millennium Partners 46 274 Neuberger Berman 61 217 Cobalt Capital Management 17 204 SAC Capital 56 189 Gates Capital Management 21 185 Omega Advisors 17 180 Capital Returns Management 2 178 Columbia Management Investment Advisors 43 175 BlackRock 34 173 Gagnon Securities 14 159 State of Wisconsin Investment Board 15 156 Bricoleur Capital Management 3 151 Wellington Management 46 150 Tieton Capital Management 3 143 Sage Asset Management 11 142 Heartland Advisors 13 139 T. Rowe Price 26 136 Duquesne Capital Management 12 133 Adage Capital Management 11 132 Arcadia Investment Management 3 129 Alliance Capital Management 30 121 Kalmar Investments 2 121 Sentinel Trust Company 7 121 Fenimore Asset Management 4 111 Davidson Investment Advisors 5 106 Talon Capital 8 105 Knott Capital Management 12 104 Kern Capital Management 4 98 Barclays 42 92 Bank of New York 18 92 Priority Capital Management 8 92 Entrust Capital 5 87 KBW Asset Management 45 84 NWQ Investment Management 4 81 Sigma Capital Management 32 80 Wells Capital Management 15 79 Pilot Advisors 12 79 Frontier Capital Management 7 77 Insight Capital Research and Management 5 77 JLF Asset Management 7 77 Westcliff Capital Management 3 77 Fidelity Investments 30 75
33
Appendix 2: Definition of Variables Variable Definition Data Source NUMBUYSIDERS Number of buy-side analysts who asked a question on the conference call. Thomson Reuters Streetevents
NUMBUYSIDERS_OWN Number of buy-side analysts from owning institutions who asked a question on the conference call.
Thomson Reuters Streetevents and 13F database
NUMBUYSIDERS_NOTOWN Number of buy-side analysts from non-owning institutions who asked a question on the conference call.
Thomson Reuters Streetevents and 13F database
EPSSURPRISE Actual reported quarterly EPS minus sell-side analyst consensus mean EPS forecast. I/B/E/S
ABSEPSSURPRISE Absolute value of actual reported quarterly EPS minus sell-side analyst consensus mean EPS forecast.
I/B/E/S
POSEPSSURPRISE Indicator variable set to 1 (0 otherwise) if EPSSURPRISE>0 I/B/E/S
NEGEPSSURPRISE Indicator variable set to 1 (0 otherwise) if EPSSURPRISE<0 I/B/E/S
NUMINSTINV Log of the number of institutional investors in the company, measured as of the calendar quarter ended prior to the conference call.
Thomson Reuters 13F database
COMPANYSIZE Log of market value of equity (in $ millions), measured as of the fiscal quarter ended prior to the conference call.
Compustat
COMPANYAGE Log of the number of months that a company has been listed in CRSP, measured as of the calendar quarter ended prior to the conference call.
CRSP
BOOK-TO-MARKET Book value of equity divided by market value of equity, measured for the fiscal quarter ended prior to the conference call.
Compustat
ABSCHGOWNERSHIP The absolute change in percentage ownership that an institution has in company in a given quarter.
Thomson Reuters 13F database
INTRADAY Indicator variable set to 1 (0 otherwise) is a conference calls starts between 6:00am and 3:45pm Eastern Time.
Thomson Reuters StreetEvents
INVFIRMSIZE Log of the total dollar value of all of an investment firm's portfolio shareholdings, measured as of the calendar quarter ended prior to the conference call.
Thomson Reuters 13F database
VALUEOFOWNERSHIP Log of the dollar value ownership that an institutional investment firm has in a company, measured as of the calendar quarter ended prior to the conference call.
Thomson Reuters 13F database
34
Appendix 2: Definition of Variables (Continued) Variable
Definition Data Source
PARTICIPANTINCALL Indicator variable for whether an institution's buy-side analysts asked a question on the conference call.
Thomson Reuters Streetevents
NUMFIRMSINPORT Log of the number of companies in an institutional investment firm's portfolio, measured as of the calendar quarter ended prior to the conference call.
Thomson Reuters 13F database
NUMSSANALYSTS Number of sell-side analysts that issued any type of EPS forecast for a company during the period from the prior conference call to one day before the current conference call.
I/B/E/S
DISPERSION Standard deviation of sell-side analysts' current quarter EPS forecasts published in the period from the prior conference call to the current conference call.
I/B/E/S
NC_ANALYSTS Number of non-covering analysts who participated on the conference call, scaled by the total number of sell-side analysts on the conference call.
Thomson Reuters Streetevents and I/B/E/S
COV_ANALYSTS_ABSENT Number of covering sell-side analysts that participated on the prior conference call but did not participate on the current conference call, scaled by the total number of sell-side analysts on the conference call.
Thomson Reuters Streetevents and I/B/E/S
ABSRET Absolute value of stock return for the 90 calendar days after the conference call.
CRSP
ABSCTURNOVER Absolute value of the change in the average percentage share turnover from 90 days before the conference call to 90 days after the conference call
CRSP
ABSCPIO Absolute value of the change in the total percentage institutional ownership (in %), measured from the calendar quarter ended prior to the conference call to the calendar quarter ended after the conference call. ABSCPIO excludes the institutions with buy-side analysts on the call.
Thomson Reuters 13F and Streetevents databases
ABSCSHORTINT Absolute value of the change in the short interest of the company as a percentage of total shares outstanding, measured from the prior conference call month to the current conference call month.
Compustat
RETPRIOR90DAYS Stock return for the 90 calendar days before the conference call.
CRSP
ABSRETPRIOR90DAYS Absolute value of stock return for the 90 calendar days before the conference call.
CRSP
RETPRIOR11MO Stock return for the 11 months from one year before the conference call to one month before the conference call.
CRSP
IMR Inverse Mills Ratio from a probit regression of an indicator for buy-side analyst conference call participation on variables included in eq. (1)
Thomson Reuters 13F & Streetevents databases
35
Table 1: Sample and Summary Statistics Panel A: Earnings Conference Calls by Year and Calendar Quarter Year 1st Qtr. 2nd Qtr. 3rd Qtr. 4th Qtr. Total 2002 - 520 975 1,189 2,684 2003 1,329 1,334 1,631 1,762 6,056 2004 1,815 1,767 1,859 1,907 7,348 2005 1,964 2,039 2,098 2,133 8,234 2006 2,154 2,160 2,296 2,305 8,915 2007 2,341 2,373 2,460 2,554 9,728 2008 2,637 2,679 2,684 2,678 10,678 2009 2,642 - - - 2,642 Total 14,882 12,872 14,003 14,528 56,285
Panel B: Percentage of Conference Calls with at Least One Participating Buy-Side Analyst
Year 1st Qtr. 2nd Qtr. 3rd Qtr. 4th Qtr. Total 2002 - 23% 17% 25% 21% 2003 31% 30% 29% 29% 30% 2004 28% 26% 28% 26% 27% 2005 26% 26% 24% 27% 26% 2006 25% 24% 25% 24% 24% 2007 22% 20% 19% 23% 21% 2008 20% 19% 20% 19% 20% 2009 21% - - - 21%
24% 23% 23% 24% 24% Panel C: Number of Buy-Side Analysts Participating per Conference Call
Number of Conference Buy-Side Analysts Calls Percent
0 42,912 76.2% 1 10,318 18.3% 2 2,404 4.3% 3 521 0.9% 4 98 0.2% 5 25 0.0% 6 6 0.0% 7 1 0.0%
Table 1 presents descriptive statistics of the sample of earnings conference calls used in this study. Panel A shows the number of conference calls by year and calendar quarter. Panel B shows the percentage of conference calls each year and quarter that have buy-side analysts that participated on the call. Panel C shows the number of conference calls with zero, one, or multiple buy-side analysts participating on the call.
36
Table 2: Determinants of Buy-Side Analysts on Conference Calls Panel A: Descriptive statistics of variables used in the determinants regressions Variable N Mean Min 1st Quartile Median 3rd Quartile Max NUMBUYSIDERS 56,285 0.31 0 0 0 0 7 NUMBUYSIDERS_OWN 56,285 0.14 0 0 0 0 5 NUMBUYSIDERS_NOTOWN 56,285 0.17 0 0 0 0 5 NUMSSANALYSTS 56,285 7.61 0 3 6 11 29 DISPERSION 56,285 0.04 0 0.01 0.03 0.04 0.43 NUMINSTINV 56,285 4.58 0 4.17 4.77 5.30 7.38 COMPANYSIZE 53,882 6.94 1.03 5.83 6.84 7.96 13.13 COMPANYAGE 55,987 4.93 1.37 4.44 5.00 5.53 6.92 BOOK-TO-MARKET 52,495 0.54 0.02 0.28 0.45 0.67 5.48 POSEPSSURPRISE 56,285 0.51 0.00 0.00 1.00 1.00 1.00 NEGEPSSURPRISE 56,285 0.28 0.00 0.00 0.00 1.00 1.00 RETPRIOR90DAYS 56,274 0.00 −0.59 0.12 0.01 0.12 0.72 INTRADAY 56,285 0.71 0.00 0.00 1.00 1.00 1.00 ABSRET 56,285 0.18 0.00 0.06 0.12 0.24 0.96
Panel B: Differences in means and medians of variables partitioned by buy-side conference call participation NUMBUYSIDERS>0 NUMBUYSIDERS=0 Differences N Mean Median N Mean Median Mean Median NUMSSANALYSTS 13,373 6.323 5.000 42,912 8.013 6.000 −1.690 *** −1.000 *** DISPERSION 13,373 0.044 0.033 42,912 0.038 0.024 0.006 *** 0.009 *** NUMINSTINV 13,373 4.501 4.736 42,912 4.603 4.787 −0.103 *** −0.051 *** COMPANYSIZE 12,768 6.813 6.735 41,114 6.983 6.868 −0.170 *** −0.133 *** COMPANYAGE 13,286 4.987 5.057 42,701 4.917 4.977 0.070 *** 0.080 *** BOOK-TO-MARKET 12,411 0.575 0.483 40,084 0.530 0.438 0.046 *** 0.045 *** POSEPSSURPRISE 13,373 0.482 0.000 42,912 0.518 1.000 −0.036 *** −1.000 *** NEGEPSSURPRISE 13,373 0.300 0.000 42,912 0.270 0.000 0.030 *** 0.000 *** RETPRIOR90DAYS 13,371 0.011 0.014 42,903 0.000 0.004 0.011 *** 0.010 *** INTRADAY 13,373 0.761 1.000 42,912 0.696 1.000 0.065 *** 0.000 *** ABSRET 13,373 0.176 0.124 42,912 0.177 0.124 −0.001 0.000
*, **, *** 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.
37
Table 2: Determinants of Buy-Side Analysts on Conference Calls (Continued) Panel C: Ordered logistic regressions with three ordinal levels on the dependent
variable (0, 1, or 2+)
Dependent variable: NUMBUYSIDERS NUMBUYSIDERS NUMBUYSIDERS Pred. _OWN _NOTOWN
Sign (1) (2) (3) NUMSSANALYSTS − −0.064 *** −0.064 *** −0.059 *** (−12.25) (−11.15) (−9.40) DISPERSION + 1.697 *** 1.007 *** 2.051 *** (6.55) (3.05) (7.84) NUMINSTINV 0.021 0.559 *** −0.142 *** (1.05) (14.94) (−6.95) COMPANYSIZE 0.066 *** −0.126 *** 0.088 *** (3.10) (−4.66) (3.58) COMPANYAGE 0.099 *** 0.028 0.123 *** (4.06) (0.97) (4.37) BOOK-TO-MARKET 0.155 *** 0.062 0.157 *** (4.16) (1.39) (3.81) POSEPSSURPRISE 0.056 0.006 0.098 ** (1.57) (0.14) (2.28) NEGEPSSURPRISE 0.175 *** 0.130 *** 0.216 *** (4.50) (2.67) (4.59) RETPRIOR90DAYS −0.027 −0.019 −0.026 (−0.52) (−0.28) (−0.39) INTRADAY 0.185 *** 0.190 *** 0.181 *** (4.02) (3.51) (3.24) ABSRET 0.113 −0.054 0.211 ** (1.48) (−0.53) (2.35) Intercept1 −3.714 *** −5.794 *** −4.309 *** (−23.66) (−31.75) (−23.70) Intercept2 −2.001 *** −3.580 *** −2.271 *** (−12.45) (−19.49) (−12.12) Year Fixed Effects Incl. Incl. Incl. N 52,318 52,318 52,318 Pseudo R2 0.024 0.030 0.026
Z-statistics are shown in parentheses below the coefficient estimates. Standard errors are clustered by companies. *, **, *** Significantly different from zero at the 0.10, 0.05, and 0.01 level, respectively, using a two-tailed test. Table 2 presents results related to testing the determinants of buy-side analyst participation on earnings conference calls. Panel A shows descriptive statistics of the variables used in an ordered logistic regression. Panel B shows tests for differences in mean and median values of the variables for companies with buy-side analysts on their conference calls and companies without buy-side analysts on their conference calls. Panel C shows the results of ordered logistics regressions with three ordinal levels (0, 1, and 2 or more) on the dependent variable. All variable definitions are included in Appendix 2.
38
Table 3: Institutional Trading After Buy-Side Analyst Participation on Conference Calls Panel A: Changes in institutional ownership conditional on buy-side analyst participation % Change in Change in % Change in Change in % Increased Ownership Ownership Decreased Ownership Ownership No Change in Total Ownership Mean Median Ownership Mean Median Ownership Owning institutions: with buy-side analyst on the call 7,890 45% 0.5% 0.5% 49% −0.7% −0.7% 6% without buy-side analyst on the call 2,467,139 41% 0.2% 0.1% 50% −0.2% −0.1% 9% Difference in differences 0.3%*** 0.4%*** −0.4%*** −0.6%*** Non-owning institutions: with buy-side analyst on the call 9,801 8% 1.2% 0.4% - - - 92% without buy-side analyst on the call 25,868,052 1% 0.3% 0.1% - - - 99% Difference in differences 0.9%*** 0.3%*** - -
39
Table 3: Institutional Trading After Buy-Side Analyst Participation on Conference Calls (Continued) Panel B: Regression of absolute changes in ownership by institutional investment firms employing buy-side analysts
on conference calls
Dependent Variable: Pred. ABSCHGOWNERSHIP Sign. PARTICIPATEINCALL + 0.067 *** (11.70) VALUEOFOWNERSHIP 0.007 *** (20.37) INVFIRMSIZE 0.003 *** (6.51) NUMFIRMSINPORT −0.002 *** (−4.58) ABSRETPRIOR90DAYS 0.009 *** (14.05) Intercept −0.054 *** (−6.45) Year Fixed Effects Included N 27,274,097 R2 0.306
*, **, *** Significantly different from zero at the 0.10, 0.05, and 0.01 level, respectively, using a two-tailed test. Standard errors are clustered by institutional investment firm. Table 3 presents results related to examining institutional trading after buy-side analyst participation on companies’ earnings conference calls. Panel A presents changes in institutional ownership conditional on buy-side analyst participation. Panel B shows the results of a regression in which the dependent variable is the absolute change in percentage ownership that institution j has in company i in quarter t on an indicator variable (PARTICIPATEINCALLi,j,t) for whether a buy-side analyst employed by institution j participated on company i's conference call in quarter t. All variable definitions are included in Appendix 2.
40
Table 4: Conference Call Participation and Future Absolute Returns, Change in Share Turnover, Change in Institutional Ownership, and Change in Short Interest Panel A: Descriptive statistics of additional variables used in the regressions of future returns, change in share turnover, and change in percentage institutional
ownership Variable N Mean Min 1st Quartile Median 3rd Quartile Max ABSRET 56,285 0.02 0.00 0.06 0.12 0.24 0.97 ABSCTURNOVER 56,274 0.26 0.00 0.06 0.14 0.33 1.65 ABSCPIO (%) 56,285 3.50 0.00 0.68 2.18 4.83 19.16 ABSCSHORTINT (%) 50,572 1.72 0.00 0.20 0.74 2.04 39.91 ABSEPSSURPRISE 56,285 0.01 0.00 0.00 0.001 0.004 0.20 COMPANYSIZE 53,882 6.94 1.03 5.83 6.84 7.96 13.13 BOOK-TO-MARKET 52,495 0.54 0.02 0.28 0.45 0.67 5.48 RETPRIOR11MO 56,206 0.10 −0.79 −0.20 0.05 0.30 2.24 NC_ANALYSTS 54,237 0.24 0.00 0.00 0.17 0.36 1.00 COV_ANALYSTS_ABSENT 51,116 0.29 0.00 0.00 0.18 0.40 2.00
41
Table 4: Conference Call Participation and Future Absolute Returns, Change in Share Turnover, Change in Institutional Ownership, and Change in Short Interest (Continued) Panel B: Fama-MacBeth regressions of 90-day absolute returns after conference calls
Pred. Sign. ABSRET NUMBUYSIDERS + 0.003 * (1.91) ABSEPSSURPRISE 0.639 *** (6.80) COMPANYSIZE −0.021 *** (−12.97) BOOK-TO-MARKET −0.001 (−0.23) PRIORRET11MO −0.007 (−0.63) NC_ANALYSTS 0.007 * (2.09) COV_ANALYSTS_ABSENT −0.001 (−0.69) ABSRET_LAG 0.136 *** (13.41) IMR 0.065 *** (6.12) Intercept 0.195 *** (13.74) N 49,311 R2 0.137
Fama-MacBeth t-statistics are shown in parentheses below the coefficient estimates. *, **, *** Significantly different from zero at the 0.10, 0.05, and 0.01 level, respectively, using a two-tailed test.
42
Table 4: Conference Call Participation and Future Absolute Returns, Change in Share Turnover, Change in Institutional Ownership, and Change in Short Interest (Continued) Panel C: Regressions of changes in daily average share turnover, total institutional ownership,
and short interest after conference calls
t-statistics are shown in parentheses below the coefficient estimates. Standard errors are clustered by companies. *, **, *** Significantly different from zero at the 0.10, 0.05, and 0.01 level, respectively, using a two-tailed test.
Pred. ABSCTURNOVER ABSCPIO ABSCSHORTINT Sign. (1) (2) (3) NUMBUYSIDERS + 0.007 *** 0.149 *** 0.123 *** (2.80) (4.27) (4.66) ABSEPSSURPRISE 0.966 *** 10.939 *** −2.085 *** (10.15) (9.46) (−3.01) COMPANYSIZE −0.008 *** −0.254 *** −0.093 *** (−6.41) (−18.27) (−8.62) BOOK-TO-MARKET 0.001 0.117 * −0.044 (0.22) (1.88) (−1.03) PRIORRET11MO 0.044 *** 0.657 *** 0.399 *** (10.32) (13.87) (10.53) NC_ANALYSTS −0.007 −0.253 *** −0.290 *** (−1.31) (−3.17) (−5.68) COV_ANALYSTS_ABSENT −0.005 0.019 0.080 ** (−1.45) (0.40) (2.25) ABSCTURNOVER_LAG 0.417 *** (56.46) ABSCPIO_LAG 0.224 *** (35.34) ABSCSHORTINT_LAG 0.294 *** (29.58) IMR 0.130 *** 0.398 *** 0.346 *** (10.33) (3.02) (3.27) Intercept 0.006 3.527 *** 1.126 *** (0.32) (16.07) (4.96) Year Fixed Effects Incl. Incl. Incl. N 49,306 49,311 46,935 R2 0.211 0.086 0.119
43
Table 4: Conference Call Participation and Future Absolute Returns, Change in Share Turnover, Change in Institutional Ownership, and Change in Short Interest (Continued) Table 4 presents results related to examining the association between the number of buy-side analysts participating on a company’s earnings conference call and that company’s future absolute stock return, absolute change in share turnover, absolute change in total percentage institutional ownership, and absolute change in short interest. Panel A shows descriptive statistics of the variables used in the regressions. Panel B shows results of Fama-Macbeth regressions in which the dependent variable is absolute value of stock returns over the 90 calendar days after a conference call. The coefficient estimates are the average of quarterly estimates over 27 quarters from the third quarter of 2002 to the first quarter of 2009. Fama and MacBeth (1973) t-statistic is presented in parentheses under the estimated coefficient. Panel C show results of panel regressions in which the dependent variable is absolute change in share turnover, absolute change in total percentage institutional ownership, and absolute change in short interest. All variable definitions are included in Appendix 2.