Gender and Earnings Conference Calls Bill Francis Lally School of Management Rensselaer Polytechnic Institute [email protected]Thomas Shohfi Lally School of Management Rensselaer Polytechnic Institute [email protected]Daqi Xin Lally School of Management Rensselaer Polytechnic Institute [email protected]January 2019 . We thank the Donald Shohfi Financial Research Fund for computing support.
53
Embed
Gender and Earnings Conference Calls · Gender and Earnings Conference Calls January 2019 Abstract Using a sample of more than 65,000 earnings conference call transcripts from 2007
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
inconclusive. In this paper, we investigate various gender-related communication issues in the
setting of earnings conference calls. In particular, we examine four research questions: (1) whether
female analysts are discriminated against in earnings conference calls participation; (2) whether
female analysts and executives behave differently from their male counterparts; (3) how female
analysts react differently to information acquired from conference calls compared with male
analysts and; (4) how the markets interpret female and male participants’ behavior differently.
Women are not only solely outnumbered in the board room where participants have face-
to-face communication—their voice is drowned out in conference calls. Earnings conference calls
represent a unique and valuable setting to study gender issues but have yet been leveraged
specifically for this purpose in the literature.4 First, in conference calls, two parties—analysts and
executives—participate together, which makes conference calls different from other disclosure
venues in which only one party is involved. Prior research leverages this analyst-manager
interaction environment and shows that narratives in earnings conference calls convey “soft”
information. For example, Larcker and Zakolyukina (2012) classify CEO and CFO narratives from
conference call transcripts into “deceptive” and “trustful” parts based on psychological and
linguistic word lists and find that the deception measure can predict subsequent financial
restatements. By systematically inferring gender within a participant’s first name, we can directly
observe the interaction between analysts and management with various gender combinations.
Second, during the question-and-answer (Q&A) session, analysts and managers interact in
real time without rehearsal or scripting. Conference calls are a stressful environment for managers
because of potential interrogation by analysts. Matsumoto, Pronk and Roelofsen (2011) argue that
the spontaneous nature of the Q&A part of a conference call leads to more information disclosure
4 See Milian, Smith and Alfonso (2017) as an exception. The authors study positive tone in language during earnings
conference calls and find that female analysts exhibit significantly more favorable language.
4
by managers because they prefer to withhold bad news in prepared statements. Literature shows
that women have low preference for competition and perform poorly compared with men under
competition (Gneezy, Niederle and Rustichini, 2003; Niederle and Vesterlund, 2007). If women
generally experience stronger emotions and more nervousness and fear when faced with
unfavorable outcomes (Croson and Gneezy, 2009), we expect to observe gender differences in
behavioral patterns during earnings conference calls.
Third, compared with other information dissemination venues of analysts and executives,
conference calls make gender more visible to participants and are more likely to reflect any
possible gender gap. Investors are expected to pay less attention to analyst gender when analyst
reports, stock recommendation or earnings forecasts are issued. However, female voice is highly
distinguishable in conference calls, which may elicit different gender perceptions for all
participants (Sturm et al., 2014; Jannati et al. 2018).
We collect more than 65,000 conference call transcripts from Capital IQ from 2007 to
2016. Using multiple algorithms based on first names, we determine the gender of analysts who
participate in conference calls. First, we follow Mayew (2008) to use I/B/E/S to identify an analyst
population who are interested in asking questions in conference calls and show that female analysts
are less likely participate in conference calls. This relatively lower likelihood of participation
persists is robust to the number of covering analysts, analyst professional characteristics (e.g.
experience, all-star status), and across industries. Moreover, we find evidence that management
discriminates against female analysts in conference call participation by deferring their positions
in Q&A session and allowing fewer follow-up questions. Both female analyst and executives have
shorter discourses than their male executives. Regarding sentiment, female analysts’ questions
elicit more positive sentiment from executives. Female executives use less numeric content when
5
answering analysts’ question but their tone is more affirmative. Markets react similarly to male
and female executives’ conference call participation. Following calls, female analysts make
forecast revision with larger magnitude.
This paper contributes to the literature in three aspects. First, we extend earnings
conference call research by introducing gender effects in analyst-management communication.
While prior research on earnings conference calls focuses on incremental information and roles of
various participants, this paper focuses on gender differences in participant behavior. Given gender
differences in financial markets and the unique communication and disclosure form of earnings
conference calls, we enlarge the scope of earnings conference calls from an information
perspective.
Second, we add to the gender research in the financial and accounting literatures. Extant
research documents gender differences in risk attitudes, competition preferences, performance etc.
(Faccio, Marchica and Mura, 2016; Post and Byron, 2015; Fang and Huang, 2017), but the results
are mixed. For example, Kumar (2010) argues that gender differences do not exist because some
females self-select into the competitive financial industry and they are less representative of the
general female population. We use earnings conference calls as a setting in which participants are
under certain pressure or constraints to examine their behavior. In addition, the interaction between
analysts and managers with the same or different genders provides a unique opportunity to directly
observe gender effects in financial markets.
The rest of the paper proceeds as follows. We review literature and develop hypotheses in
Section 2. Sections 3 describes data. In Section 4, we present analysis of gender effect in analyst-
manager interaction on conference call. Sections 5 concludes.
6
2. Literature Review and Hypothesis Development
2.1 Gender and Corporate Decisions
The theoretical foundation of gender’s effect on corporate decisions is the behavioral
difference from the perspective of risk attitude and moral standard (Ho et al., 2015). Given the
literature in general social sciences, women are more risk-averse than men (Byrnes, Miller and
Schafer, 1999; Powell and Ansic, 1997; Croson and Gneezy, 2009). Faccio, Marchica, and Mura
(2016) find that firms with female CEOs have lower leverage, less earnings volatility, and a higher
probability of survival. Huang and Kisgen (2013) document that males are overconfident about
significant corporate decisions compared to females. Specifically, female executives conduct
fewer acquisitions and issue debt less often. Accordingly, markets react more strongly to female
executives’ announcements. Moreover, female executives give a wider range of earnings estimates
than male executives. Gul, Srinidhi, and Ng (2011) find that the presence of a woman in a firm’s
board of directors leads to consistently higher stock price informativeness. They further show that
the channel of this relationship is through more public firm-specific disclosure in large firms and
by facilitating more private information collection in small firms. Regarding financial reporting,
female CFOs are more conservative (Francis, Hasan, Park and Wu, 2015), produce higher quality
earnings (Srinidhi, Gul, and Tsui, 2011;Krishnan and Parsons, 2015), and conduct less earnings
management (Barua, Davidson, Rama, and Thiruvadi, 2010; Peni and Vahamaa, 2010).
In addition, a gender gap also exists in ethical issues. Gender socialization theory argues
that personality differences between men and women are the result of divergent social expectations
and learning social rules differently. Gilligan (1982) finds that men and women are different in the
way they address moral dilemmas. Franke, Crown, and Spake (1997) conduct a meta-analysis and
find that women have higher standards for ethical business practices. Bernardi and Arnold (1997)
use a Defining Issues Test to measure moral development of managers in five of the Big Six
7
accounting firms and find that female managers are more morally developed. More directly, female
executives are associated with less securities fraud (Cumming, Leung, and Rui, 2015).
2.2 Conference Calls
Earnings conference calls are one of the most important venues to communicate with
institutional investors (Brown, Call, Clement, and Sharp, 2016). The majority of conference calls
are held immediately following a quarterly earnings release. A conference call usually starts with
a presentation session in which each participating executive discusses current operations and
forward-looking statements. After presentation, analysts and investors can ask managers questions
regarding the firm. Prior studies show that conference calls provide value-relevant information to
capital markets (Frankel et al. 1999; Bushee et al. 2004; Kimbrough, 2005). Matsumoto, Pronk,
and Roelofsen (2011) find that both presentation and discussion sessions have incremental
information over press releases and that discussions sessions are particularly more informative.
They further show that the informativeness of a Q&A session is associated with the number of
analysts following the firm. Their findings suggest analysts’ active role may contribute to the
informativeness of conference calls. Further, Bowen et al. (2002) show that conference calls
increase analysts’ forecast accuracy and decrease forecast dispersion. Mayew (2008) shows that
firms discriminate against unfavorable analysts by providing analysts who issue favorable stock
recommendation with more opportunities to ask question during conference calls. Further, Mayew,
Sharp, and Venkatachalam (2013) find that analysts who participate in conference calls by asking
questions issue more accurate and timelier earnings forecasts than non-participating analysts,
suggesting participating analysts may possess superior information.
One stream of literature examines soft information embedded in conference calls.
Linguistic cues contain soft information which is incremental to press releases (Matsumoto et al.
8
2011). For example, Allee and DeAngelis (2015) document that tone dispersion, the degree to
which tone is spread evenly in a narrative, is associated with firm performance, managers’ financial
reporting choices, and managers’ incentive to influence perception of the firm. Mayew and
Venkatachalam (2012) show that managers’ affective states in conference calls can predict future
firm performance and the effect is more prominent in the Q&A session when managers are under
greater scrutiny by analysts. Davis, Ge, Matsumoto, and Zhang (2015) show that there exists a
manager-specific component in the tone of earnings conference calls that cannot be explained by
current performance, future performance, or strategic incentives. They further add that this
manager-specific factor is related to demographic characteristics including career experience and
charitable organization involvement. The authors argue that tone of executives in earnings
conference calls is associated with their level of optimism. However, they only document weak
evidence that female executive use less favorable language.
2.3 Hypothesis Development
Gender discrimination is ubiquitous among male-dominated industries. For example,
Jacobi and Schweers (2017) examines the oral argument at the U.S. Supreme Court and show that
females Justices are disproportionately interrupted by both their male counterparts and male
advocates. Equity analysts are a male-dominated occupation. Given the extensive gender
discrimination and “old boys network”, establishing connections for female analysts is potentially
more difficult. Fang and Huang (2017) document that females account for only 12% of all analysts
in the 1993-2009 period. Although they find females analysts are equally likely to be selected as
Institutional Investor all-star analysts, performance improvements and recommendation impact for
female analysts are much lower compared to their male counterparts. They further show that these
connections mitigate the negative influence of forecast error on reputation for male analysts but
9
intensifies the negative effect for female analysts. Consequently, the disparity in the effectiveness
of connections can lower the intention of female analysts to make connections with managers.
Moreover, because managers have discretion over analysts’ conference call participation (Mayew,
2008), connections are a key determinant of conference call participation. Consistent with this
argument, Brown, Call, Clement and Sharp (2015) survey 365 analysts and find that analysts avoid
asking difficult questions in a conference call to maintain a good relationship with management
and leave harsh questions to private communication instead. Along the same line, Soltes (2014)
argues that public interaction between management and analysts is an approach to maintain a
relationship.
From the other perspective, female analysts are frequently documented as better performers.
Kumar (2010) proposes a self-selection hypothesis that female analysts are not representative of
common female characteristics such as higher risk aversion and lower preference for
competitiveness but are self-selected into the male-dominated profession due to their superior
ability. Consistent with the self-selection hypothesis, he finds that female analysts issue bolder and
more accurate forecasts. Female analysts are more likely to cover large stocks with higher
institutional ownership even in the early stage of career. He further shows that the market reacts,
both in the short and long term, more strongly to female analysts’ forecast revisions even when
they attract less media coverage. In addition, Kumar (2010) documents female analysts are more
likely to be promoted to a prestigious brokerage firm and less likely to receive a demotion to a less
prestigious one. Li, Sullivan, Xu, and Gao (2013) find that the markets render the same level of
importance to male and female analysts’ recommendations in terms of abnormal returns but less
idiosyncratic risk is generated by female analysts’ recommendation portfolios. They find no
evidence of gender discrimination measured by the likelihood of brokerage firm upward mobility
10
and find that female analysts are more likely to be selected as star analysts by both Institutional
Investor and the Wall Street Journal.
From the perspective of managers, coverage from prestigious analysts is valuable due to
increased credibility and stronger market impact (Stickel, 1992; Gleason and Lee, 2003; Park and
Stice, 2000). Mayew (2008) documents the role of analyst reputation as a deterrent of
discrimination given the high cost of discrimination against prestigious and unfavorable analysts.
In particular, even though managers have incentives to limit conference call access for unfavorable
analysts, recommendation downgrades are not associated with a decrease in conference call
participation for prestigious analysts. Given the superior performance of female analysts (Kumar,
2010) and their higher probability of being voted as all-star analysts (Green, Jegadeesh, and Tang,
2009), management may increase their conference call access. Since the relative importance of
gender discrimination in earnings conference calls is undetermined, we propose:
H1a: Female analysts are less likely to participate in earnings conference calls.
H1b: Female analysts are more likely to participate in earnings conference calls.
Firm are very sensitive about information disclosure in conference calls given that both
solid and soft information is disseminated to the public (Zhou, 2018; Suslava, 2017).5 To avoid
disclosing unfavorable information, management regularly chooses to not answer certain analysts’
questions (Hollander, Pronk, and Roelofsen, 2010) or disproportionately prioritize optimistic
analysts (Cohen, Lou, and Malloy, 2016). Given the time limit of conference calls, managers may
not be able give a thorough answer to questions asked by analysts appearing late in the queue
compared to questions asked earlier. According to firms’ Investor Relations Officers (IROs),
priority in the question queue is usually given to analysts who have a long coverage history with
5 For example, Elon Musk, the CEO of Tesla, Inc., said the questions from analysts are “boring, bonehead questions”
in its 2018 Q1 earnings conference call on May 2nd, 2018. The stock price plunged 5.6% on the following day.
11
the firm (Brown et al., 2017). Consequently, asking the first question in a conference call sends a
strong signal of firm’s special attention and connection with analysts (Cen, Chen, Dasgupta, and
Ragunathan, 2018; Call et al. 2018). Recent evidence shows that the Q&A session of earnings
conference calls is less spontaneous than it seems to be: sell-side analysts provide question to be
asked in conference calls to Investor Relations Officers (IROs) in advance (Brown et al., 2018).
The coordinated nature of Q&A session further entails a deep relationship between two parties.
Because of the lower benefits from connection to management for female analysts (Fang and
Huang, 2017) and potential in-group bias (Jannati et al., 2018), female analysts may be less capable
of building connections. Therefore, we propose:
H2: Females analysts are less likely to ask the first question and to have follow-up
interactions in the question-and-answer session on earnings conference calls.
Analysts benefit from connections with management both from the perspective of research
informativeness (Green, Jame, Markov, and Subasi, 2014) and compensation (Groysberg, Healy,
and Maber, 2011). Under Regulation FD, although firms must open conference calls to all
interested members of the general public (Bushee et al., 2004), the complementing role of public
information to private information (i.e. mosaic theory) on earnings conference calls remains
crucial for analysts (Mayew, 2008). Since management has discretion to decide who can ask
questions (Mayew, 2008), analysts’ connections with managers are crucial for analyst success.
Analysts value their reputation from recognition such as “all-star” status which is voted on by
influential institutional investors (Fang and Huang, 2017). Connections of analysts are also
associated with their quality of opinion and career advancement. Sell-side analysts have strong
incentives to curry favor of buy-side clients (Groysberg, Healy, and Maber, 2011). A considerable
amount of compensation paid by buy-side clients to sell-side firms is for corporate access (Brown
12
et al., 2018). Moreover, the limited time allocated to each analyst in a conference call and analysts’
concern of “tipping their hands” suggest conference call participation is more of a relationship
manifestation (Brown et al., 2015; Brown et al., 2016; Chen and Matsumoto, 2006). Management
often provides “call-backs” to well-connected analysts (Brown et al. 2018). To retain this
connection with management, analysts must not interrogate executives and/or cast them in a
negative light. “Assuming you want management to continue speaking with you, you have to avoid
making the C-suite lose face on the call…if you have difficult questions and you want management
to speak openly, you have to do that off-line.” (Soltes, 2014).
Moreover, men and women have different views on the purpose of conversation. Women
seek social connections and relationships in communication while men exhibit power (Leaper,
1991). Consequently, women are more expressive and politer in conversation while men are
aggressive (Basow and Rubenfeld, 2003). Research in linguistic documents conversation between
females as more fluent and affirmative compared to mixed-gender pairs and male-only pairs
(Hirschman, 1994).
Therefore, male analysts may be less concerned about the “conversation in harmony” and
may manifest their ability to the public by asking tough questions related to weaknesses of the
firm, which entails managers to explain with more words reflecting negative sentiment. On the
contrary, female analysts are expected to initiate a relatively relaxed conversation with
management in accordance with the “theater” nature of conference calls (Brown et al., 2018).
When both questioner and answerer are female, the cooperation attribute of the conversation will
be stronger given that females usually exhibit strong in-group favoritism (Rudman and Goodwin,
2004). Thus, we have:
13
H3: Female analysts’ interaction with management on earnings conference calls is
shorter than male analysts’ interaction with management. Interactions are shortest if the
manager is also female.
H4: The tone of female analysts’ interaction with management on earnings conference
calls is less negative than the tone of male analysts’ interaction with management. The tone is
least negative when both the analyst and the manager are female.
Women are less resistant to pressure (Gneezy, Niederle, and Rustichini, 2003; Niederle
and Vesterlund, 2007) and evoke more negative feelings when anticipating negative outcomes
(Croson, and Gneezy, 2009). Given lower resilience to pressure and high ethical standards, when
faced with interrogation from analysts, especially male analysts, female managers will reveal more
information truthfully. Therefore, we propose:
H5a: Female managers exhibit less uncertainty in their narratives.
H5b: Female managers use more numeric information in their narratives.
Women are more conservative than men (Byrnes, Miller and Schafer, 1999; Powell and
Ansic, 1997; Croson and Gneezy, 2009; Niederle and Vesterlund, 2007) and conservative
individuals are more likely to exhibit status quo bias (Samuelson and Zeckhauser, 1988;
Kahneman et al. 1991). However, female analysts are found to rely more on independent research
relative to earnings news and are less likely to issue forecast revisions than men after earnings
announcements (Green, Jegadeesh, and Tang, 2009). If female analysts have less access to
earnings conference calls, they may be more sensitive to new information obtained therefrom. In
addition, female analysts issue bolder forecast revision because of their superior ability and low
employment risk (Kumar, 2010). We expect female analysts will issue bolder forecast revision
than male analysts.
14
H6: Female analysts’ forecast revision magnitude is larger than that of male analysts.
3. Data
3.1 Sample selection
Earnings conference call transcripts of Standard and Poor’s 500 (S&P 500) constituent
firms are collected via Capital IQ from 2007 to 2016. Additionally, we also collect transcripts of
another 2,700 firms which are not included in S&P 500 index but appear in the Center for Research
in Security Prices (CRSP) database. Our sample construction starts with 81,677 earnings
conference call transcripts for 3,346 unique publicly traded companies. Firms without data in
I/B/E/S or CRSP are removed. For each transcript, we record the call date, time stamp, names of
firm executives, names of analysts participating in the question-and-answer (Q&A) session, and
analyst affiliation. We follow Mayew (2008) to use I/B/E/S as the universe of analysts who are
potentially interested in attending conference calls and construct a corresponding I/B/E/S sample.
For the initial I/B/E/S sample, we require each firm-quarter-analyst observation to have both an
outstanding earnings forecast and an outstanding stock recommendation. We refer to this as the
“full I/B/E/S sample.” Earnings forecasts must be issued within one year of a given fiscal quarter
end for an analyst to be considered as actively following the firm.
To determine analyst gender, we need to obtain analyst full analyst first names. I/B/E/S
only provides each analyst’s last name and first initial (item “ANALYST” in I/B/E/S).
Observations with missing brokerage ID (ESTIMID in I/B/E/S) or analyst name are removed. In
addition, forecasts made by research teams are eliminated.6 To ensure the accuracy of analyst
gender determined from first name, we remove analysts for which two or more analysts (indicated
6 Analyst names for forecast issued by team are record in I/B/E/S as two or more last names (e.g.,
“GERRY/ADKINS”, “RESEARCH DEPT”).
15
by analyst code in I/B/E/S) share the same analyst last name in the same brokerage (Bradley,
Gokkaya, and Liu, 2017). Next, to determine the first name of analysts in I/B/E/S, we match the
analysts on the transcripts with the analysts in the I/B/E/S at the brokerage level. We also remove
observations for which a firm is covered by only one analyst for a fiscal quarter end. Only the most
recent forecasts prior to an earnings conference calls are used. We apply the R package, gender
and a Python package, gender-guesser, to determine the gender of analyst based on the first name.7
For androgynous names, we further use the dataset on gender-api.co.8 All of these tools use
publicly available government databases and social network data to construct name-gender
databases. For executives who appear in conference calls, we match names with Execucomp
records which have gender and other information. Among 224,452 call-executive observations,
1,592 (0.7%) are unidentified. 9 Finally, we complement missing analyst and executive gender by
manually searching a variety of sources including Capital IQ, LinkedIn, Bloomberg, Seeking
Alpha, etc. We successfully identify the full name and gender for 5,687 analysts (99.8% of 5,722
unique analysts corresponding to all conference calls) in I/B/E/S. The final I/B/E/S sample
includes 708,592 analyst-firm-quarter observations related to 2,876 firms and 65,850 conference
calls. For the call-analyst sample, we identify analyst gender for 97% of observations.
In order to investigate the dynamics of analyst-management conversation in each
conference call, we parse all conference call transcripts into question-answer blocks. Each
conference call transcript is scanned from the beginning to the end to identify these blocks.
Conversation is defined as continuous back-and-forth comments between the analyst and
executives in which call no conference call operator speaks out in between. One difficulty is to
7 https://cran.r-project.org/web/packages/gender/gender.pdf and https://pypi.python.org/pypi/gender-guesser/ 8 https://gender-api.com/en/ 9 Unidentified company participants are recorded as “Unidentified Company Representative”, “Unknown
where 𝑓𝑜𝑟𝑒_𝑒𝑟𝑟𝑜𝑟𝑖,𝑗,𝑡 is the absolute forecast error (the absolute difference between the last
earnings per share (EPS) forecast and actual EPS) for analyst i of firm j in quarter t and
𝑓𝑜𝑟𝑒_𝑒𝑟𝑟𝑜𝑟𝐽,𝑡 is the mean absolute forecast error (average 𝑓𝑜𝑟𝑒_𝑒𝑟𝑟𝑜𝑟𝑖,𝑗,𝑡 across all analysts
covering firm j in quarter t. Positive (negative) fore_accuracy indicates an analyst’s forecast is
more (less) accurate than the forecasts of the same firm in the same quarter. This measure of
forecast accuracy is relative to other analysts and eliminates heteroskedasticity across firm-quarter
(Ke and Yu, 2006).
10 The method could lead to misidentification of analyst’s target executive for several reasons. First, an analyst may
not have a target executive to whom the question is asked. Second, an analyst may not indicate an executive to
answer the question and the executive who answers the question is not the one who is expected by the analyst.
Third, the executive who speaks out first may ask another executive to take the question.
17
If analyst ability varies systematically with gender, the relationship between analyst gender
and conference call or market outcomes will be biased. To account for analyst ability, we follow
Clement (1999) to include variables which are related to analyst ability. AllStar is an indicator
variable for Institutional Investor All-American analysts in a given year. GenExp is the number of
years between the conference call data and the date on which the analyst issued a forecast on
I/B/E/S for the first time. FirmExp is the number of years between the conference call data and the
date on which the analyst issued a forecast for the firm on I/B/E/S for the first time.
4. Empirical findings
4.1 Analyst gender distribution
Table 1 reports the call-analyst level analyst gender distribution by year (Panel A), Global
Industry Classification Standard (GICS) sector (Panel B) and brokerage affiliation (Panel C).11 In
Panel A, the number of quarterly earnings conference calls exhibits a steady increasing trend
except for 2016.12 Total number of unique analysts is lower in the later years compared with
earlier. We find a slightly decreasing trend for female analyst participation. The percentage of
female analyst participation and the percentage of unique female analysts indicate that for those
participating analysts in our sample, the likelihood of participation of female and male analysts are
similar. Panel B shows the gender distribution across 11 GICS sectors. Female analysts are more
concentrated in Consumer Staples and Consumer Discretionary, followed by Health Care, Real
Estate, and Utilities. The evidence is consistent with the self-selection hypothesis that female
analysts choose sectors in which they have more expertise (Kumar, 2010). In Panel C, we follow
Green et al. (2009) to rank brokerage firms in I/B/E/S database based on the number of affiliated
11 The number of observations is this sample is less than the conversation sample for two reasons. First, each
analyst-call has only one observation. Second, analysts without determinable gender are not included. 12 The relatively small number of conference calls in 2007 is due to data availability in Capital IQ.
18
analysts in each year and separate top 10 and other brokerages. Analysts without an affiliated
brokerage found in I/B/E/S are either with buy-side institutions, media outlets, and other
institutions (Call et al., 2018). We find that the proportion of female analysts who are in I/B/E/S
is higher than that of non-IBES analysts. In I/B/E/S analysts, women account for 16.59% analysts
in top 10 brokerages compared with 11.52% in other brokerages. Green et al. (2009) argue that the
relative high representation of female analysts in large brokerages is because of the emphasis on
the employee diversity and better working conditions which are attractive to women.
[Insert Table 1 here]
4.2 Descriptive Statistics
Table 2 presents descriptive statistics for conference call level variables. On average, a firm
has market capitalization is $7 billion (MktCap), market-to-book ratio of 2.89 (MB), a leverage
ratio of 2.59 (Leverage) and return on asset of 0.01 (ROA). 22% (58%) of firms are S&P 500 (S&P
1500) constituents. Institutional ownership accounts for 67% of total shares (InstOwn). Each firm
is covered by 7.24 analysts (NumAna). The average standardized unexpected earning (SUE) is
0.033. Mean stock recommendation consensus is 0.721 (RecCon).
Regarding conference call characteristics, the total number of words spoken in the Q&A
session is 3,808 at the mean (TotalWords). On average, 7.6 conversations (ConverCall) are
conducted by 7.2 analysts (AnaCount) and 3.4 executives (ExeCount). Among analysts, the median
number of analysts in I/B/E/S is 5 (IBESCount). 95.6% of conference calls have at least one I/B/E/S
analyst (IBESPart). The number (percentage) of female analysts is 0.768 (9.7) at the mean. 3
executives attend a conference call at the median. Average number of female executives is 0.44
(FemaleExeCount) and the mean percentage of female executives is 12.6% (FemaleExePct).
Since one duty of Investor Relations Officers (IROs) are organizing conference calls and they are
19
therefore frequent participants but do not answer analysts’ questions, we limit executives to only
CEOs and CFOs who are regarded as the most important corporate participants. 58.1% (56.6%) of
conference calls have CEO (CFO) attended (CEOPart and CFOPart) and 50.8% have both CEO
and CFO attended (CEOCFOPart). We find that the percentage of female CEO or CFO is 6%
(FemaleCEOCFOPct) compared with 12.6% of female firm participants. This is consistent with
previously reported evidence that about percentage of female IROs are are much higher than other
executives (Brown et al., 2018).
[Insert Table 2 here]
4.3 Univariate analysis
4.3.1 Gender difference in narratives
We first compare the mean of a series of analyst-call level variables between male and
female analysts. Table 3 Panel A shows the results. Variable definition is summarized in Appendix
B. Female analysts are much more likely to be all-star analysts (AllStar), are hired by large broker
firms (BrokerSize), have less general experience (GenExp) but similar firm-specific experience
(FirmExp), cover fewer industries (IndCover) and companies (CompCover), are more accurate in
earnings forecasts (ForeAcc), and have less favorable stock recommendation (Rec) and shorter
recommendation horizons (RecHorizon). Results are consistent with prior studies (Mayew, 2008;
Kumar, 2010). Regarding analyst participation characteristics, we report four variables: the order
of analyst question in Q&A session (Order), first questioner dummy (First), the number of
conversations between analyst and managers (ConverAna), and abnormal conversation length
(AbnLength). Following Call et al. (2018), AbnLength is defined as:
Specifically, female analysts appear later in the Q&A queue, are less likely to ask the first question,
are less likely to have a follow-up interaction with executives, and have shorter conversations,
consistent with H2 and H3.
Narrative sentiment is measured for analysts and managers separately with the Loughran
and McDonald (2011) (LM) dictionaries in each conversation block. 13 Prior research has
established that the LM dictionary is an effective measure of financial statement sentiment. Given
that the LM dictionary is specially designed for financial statements and conference call transcripts
are derived from verbal communication, we also use Harvard General Inquirer (Harvard GI)
dictionary to measure sentiment. 14 To capture the general sentiment in analyst-management
interaction, we construct a net tone measure, which is the difference between positive and negative
tone (net and netGI). Positive net tone indicates that an interaction exhibits more positive sentiment
than negative sentiment. Each tone variable is the number of words in each tone dictionary divided
by the total number of words spoken in percentage. In addition, we follow Zhou (2018) to examine
the percentage of number of in the narratives (number). Numbers are expected to contain more
value-relevant information than lexical content (Zhou, 2018). In addition, we include three
variables related to conversation characteristics: the percentage of interruption (Interrupt),
percentage of hesitation (Hesit), and the number of back-and-forth comments for the call-analyst
(Rally) in each interaction block. Prior work finds that females are more likely to be interrupted
and men are likely to be the interrupter on the Supreme Court (Jacobi and Schweers, 2017). In a
conference call, being interrupted indicates managers (analysts) strongly disagree with the
analyst’s (manager’s) comments. In the example given in Appendix C Panel A, the CEO of Tesla
Inc., Elon Musk, interrupted analyst Galileo Russell’s comment. A more oppressive comment will
13 The Loughran and McDonald (2011) dictionaries can be found at http://sraf.nd.edu/ 14 The Harvard General Inquirer dictionaries can be accessed at http://www.wjh.harvard.edu/~inquirer/