Executive Extraversion: Career and Firm Outcomes

T. Clifton Green, Russell Jame, and Brandon Lock*

November 2017

Abstract

Psychology research identifies extraversion as the personality trait most closely associated with leadership emergence. We examine executive extraversion, as measured by speech patterns during conference calls, and find extraverts experience significant career benefits. Controlling for executive and firm characteristics, including firm fixed effects, we find that extraverted CEOs and CFOs earn 6-9% higher salaries. Moreover, extraverts become top executives at a younger age, are less likely to experience job turnover, and extraverted CFOs are more likely to be promoted to CEO. Executive extraversion is also linked with firm outcomes. Analyzing a sample of manager transitions, we find that increases in CEO extraversion are associated with improvements in sales growth, revenue efficiency, and investor recognition. Further, extraverted CEOs are associated with higher acquisition announcement returns. Our findings highlight the role of personality traits in explaining executive career and firm outcomes. JEL: G14

Keywords: Executive Compensation, Personality, Labor Market

* Green is from Goizueta Business School, Emory University, clifton.green@emory.edu. Jame is from Gatton College of Business and Economics, University of Kentucky, russell.jame@uky.edu. Lock is from Kellogg School of Management, Northwestern University, b-lock@kellogg.northwestern.edu. We thank Jenny R. Guidry for excellent research assistance. We thank Emily Bianchi, Dan Bradley, Sinan Gokkaya, Elvira Alexandra Scarlat, Paul Smeets, Soojin Yim, Scott Yonker, and seminar participants at Clemson, Cornell (Dyson School), Florida, Georgia State, Luxembourg School of Finance, Miami University, Tulane, Wilfrid Laurier, the European Finance Association Meeting, the American Accounting Association, and the Jim and Jack conference at the University of Tennessee.

1

1. Introduction

Economic theory often assumes that managers vary in talent (e.g., Murphy and Zabojnik,

2004; Gabaix and Landier, 2008; Edmans, Gabaix, and Landier, 2009), and a growing empirical

literature suggests that managers have particular styles that can significantly impact corporate

performance.1 Corporate boards treat the selection of top executives as a critical element of firm

success, which has led to rapid growth in CEO salaries in recent decades (Khurana, 2004).

However, relatively little is known about which traits are viewed as important to boards in their

hiring of top executives. Indeed, Graham, Li, and Qiu (2012) conclude that unobservable

managerial traits explain a large fraction of the variation in executive pay.

In this article, we explore the role of personality in explaining variation in executive labor

market and firm outcomes. We place particular emphasis on extraversion, which is often

described as the single most important aspect of personality (Cain, 2012). Popularized by Jung

(1921), extraversion is a component of virtually all comprehensive models of personality

including the Big Five model and the Myers-Briggs Type Indicator. Extraversion tends to be

manifested in outgoing, talkative, energetic behavior, whereas introversion is manifested in more

reserved and solitary behavior.

Our emphasis on extraversion is motivated by a vast psychology literature that documents

a relation between extraversion and leadership. For example, Judge, Bono, Ilies, and Gerhardt

(2002) conclude in their survey that extraversion is the most consistent correlate of leadership

across study settings and leadership criteria, with extraverts more likely to be perceived as

effective by both supervisors and subordinates. Moreover, Bono and Judge (2004) find leaders

1 For example, Malmendier, Tate, and Yan (2011) and Benmelech and Frydman (2015) find a connection between early life experiences and management style, and Custódio and Metzger (2013, 2014) find evidence that CEOs’ employment histories influence corporate performance. Other work finds that financial management styles are influenced by measures of CEO optimism and overconfidence (Graham, Harvey, and Puri, 2013; Malmendier and Tate, 2005, 2008; and Hirshleifer, Low, and Teoh, 2012).

2

high in extraversion are more likely to express charisma, provide intellectual stimulation, and

offer individualized consideration to their employees.

In our analysis, we construct measures of extraversion for over 4,500 CFOs and CEOs at

S&P 1500 firms during 2006-2013. Our approach relies on linguistic algorithms fit to speech

patterns from the question and answer portion of quarterly conference calls. In particular,

linguistic research suggests that extraverts have a higher verbal output, use less formal language,

exhibit less word variety, and use more assertive language (Scherer, 1979; Furnham, 1990; Gill

and Oberlander, 2003). Extraverts also use more positive and negative emotion words than

introverts (Pennebaker and King, 1999). To help validate the textual approach, we compare our

linguistic measure of extraversion to listener-based assessments for a subset of the sample and

find listener-based and algorithm-based extraversion measures agree 68% of the time in a binary

setting. We also find that our linguistic measure is largely unrelated to firm performance near the

conference call, and it persists much more at the individual level than at the firm level, which is

consistent with the view of extraversion as an innate individual characteristic.2

We find strong evidence that extraverts experience greater labor market success. After

controlling for a number of firm factors known to influence compensation, we find that a one

standard deviation increase in CEO extraversion is associated with a compensation premium of

5.7% to 7.6% (or roughly $316,000 to $420,000). The relation between extraversion and

compensation is distinct from optimism or overconfidence. We control for the tone of the

executive responses to questions during conferences calls as well as option-based measures of

CEO overconfidence. The relation is also robust to controlling for other aspects of personality

and managerial characteristics known to affect compensation including education, the breadth of

2 Cain’s (2012) review of the psychology literature suggests extraversion is 40-50% heritable, and Borghans, et al. (2008) state that personality stability peaks between the ages of 50 and 70.

3

the CEO’s past work experience, and the size of the CEO’s network. We also document a very

similar relation among CFO extraversion and compensation.

We find extraverted executives also experience a number of additional career benefits. In

particular, extraverted executives tend to be first appointed to CEO at a younger age, are less

likely to experience job turnover, and have longer job tenures. We also find that extraverted

CFOs are more likely to be promoted to CEO. Specifically, a one-standard deviation increase in

CFO extraversion is associated with a 31% to 52% increase in the likelihood of being promoted

to CEO following the departure of the current CEO. Moreover, we find extraverted CEOs are

16% more likely to hold outside directorships, and the directorships they hold are at significantly

larger firms. Collectively, our results indicate that extraverted executives experience significant

career benefits.

We next explore the relation between executive extraversion and firm outcomes. If

extraverted managers’ employment success and salary premium reflect superior skill, we may

expect extraverts to deliver superior performance, consistent with managers not being able to

fully extract rents from their ability (e.g., Falato, Li, and Milbourn, 2015).3 On the other hand,

Custódio, Ferreira, and Matos (2013) reason that in an efficient labor market, CEOs are chosen

optimally by firms and therefore there should be no observed relation between CEO

characteristics and performance. Further complicating this issue, the matching process may

depend on firm performance. For example, firms may seek out extraverted managers to help

explain anticipated poor performance. Despite these potential challenges, we explore whether

variation in manager extraversion caused by CEO turnover are associated with changes in a

number of firm outcomes. 3 Gabaix and Lander (2008) argue that even if this prediction holds, it may be difficult to empirically detect due to the differences in scale between executive compensation and firm revenues (i.e., small percentage improvements in profits potentially justify large percentage increases in compensation).

4

We conjecture that extraverts’ tendency to attract social attention (e.g., Ashton, Lee, and

Paunonen, 2002) may lead to greater external firm visibility, and we begin by exploring

measures of investor recognition. We find that increases in CEO extraversion through manager

turnover are associated with increases in analyst coverage and more frequent firm presentations

at investor conferences. Increases in extraversion are also linked with improvements in liquidity,

as evidenced by significantly higher stock turnover and lower levels of Amihud’s (2002)

illiquidity measure.

Executive extraversion is also significantly related to some measures of firm

performance. Specifically, increases in CEO extraversion due to CEO transitions are associated

with increases in sales growth, market share, and firm efficiency (Demerjian, Lev, and Mcvay,

2012). If markets attribute some of the beneficial firm outcomes associated with executive

extraversion to managerial ability, we would expect to see negative announcement effects

associated with voluntary departures of extraverted CEOs. We explore this hypothesis using

hand collected data on more than 500 voluntary CEO departures and find point estimates ranging

from -0.19% to -0.57%.4 In additional analysis, we study announcement returns to mergers and

acquisitions activity and find complementary evidence, with an increase in CEO extraversion

being associated with M&A announcement returns that are 0.22% to 0.38% higher. Taken

together, the beneficial investor recognition, firm performance, and market reaction findings

provide support for a rational market-based explanation for the improved labor market outcomes

of extraverted CEOs.

Our study contributes to the growing literature that explores the determinants of

executive compensation, such as educational background, breadth of past work experience, or

4 We also find large negative associations between departure returns and CEO extraversion for a small sample of 15 unexpected CEO departures.

5

the size of an executive’s network (Falato, Li, and Milbourn, 2015; Custódio, Ferreira, and

Matos, 2013; Engelberg, Gao, and Parsons, 2013). While existing work largely focuses on

acquired attributes, we highlight an important underlying psychological factor that may influence

many of these variables (Hambrick and Mason, 1984). Our findings also add to the literature that

explores determinants of other labor market success such as the age at which an executive is first

appointed to CEO (Falato, Li, and Milbourn, 2015), determinants of internal promotion to CEO

(Parrino, 1997), and the number of outside directorships held by CEOs (Booth and Deli, 1996).

Further, our study relates to recent work highlighting the importance of non-cognitive

characteristics in the labor market for CEOs such as charisma or vocal pitch (Kaplan and

Sorensen, 2016; Mayew, Parson, and Venkatachalam, 2013).5 Finally, our work also contributes

to research that emphasizes the role of manager characteristics in explaining corporate financial

decision making (e.g., Bertrand and Schoar, 2003; Kaplan, Klebanov, and Sorensen, 2012), and

acquisitions returns (Malmendier and Tate, 2008; Custódio and Metzger, 2014).

Our paper extends the growing literature on textual analysis in accounting and finance.

Existing work analyzes the text of company disclosures to measure tone (Tetlock, Saar-

Tsechansky, and Macskassy, 2008; Loughran and McDonald, 2011; Jegadeesh and Wu, 2013),

uncertainty (Loughran and McDonald, 2013), readability (Li, 2008; Loughran and McDonald,

2014), and deception (Larcker and Zakolyukina, 2012). Our work is among the first to use

linguistic approaches to infer non-cognitive executive characteristics.6

5 Other examples include Persico, Postlewaite, and Silverman (2004), Graham, Harvey, and Puri (2014), Adams, Keloharju, and Knupfer (2014), and Otto (2014), who explore the relation between compensation and non-cognitive characteristics such as height, appearance, and optimism. 6 Dikolli, Keusch, Mayew, and Steffen (2014) proxy for CEO integrity using excessive explanations in annual shareholder letters and find an association between integrity and ethical behavior. Davis, Ge, Matsumoto, and Zhang (2014) find evidence of persistent conference call tone style that is influenced by involvement in charitable organizations. Gow et al. (2015) use textual algorithms to infer personality similar to our approach; they study firm policies rather than executive career outcomes. Other studies use voice (e.g., DeGroot et al., 2011; Mayew, Parsons,

6

2. Measuring extraversion

In this section we describe extraversion, the methodology we use to measure it, and the

sample datasets.

2.1 Measuring Extraversion from Speech

Individual personality traits are generally considered to be stable, persistent

characteristics (Costa and McCrae, 1988; Scherer, 2003; and Borghans, et al., 2008).

Psychologists commonly assess personality along five dimensions known as the Big Five

(Norman, 1963; John and Srivastava, 1999): (1) extraversion, (2) emotional stability, (3)

agreeableness, (4) conscientiousness, and (5) openness to experience. These personality traits

have been repeatedly obtained in factor analyses of personality description questionnaires

(Goldberg, 1990), and the Big Five Model has become standard in the psychology literature. We

focus on extraversion, which among the Big Five has produced the most findings related to

leadership and is easiest to infer from communication style (Dewaele and Furnham, 1999).

Extraverts are described as being outgoing and energetic whereas introverts tend to be

reserved and solitary.7 Jung (1921) describes extraverts as outward-turning and introverts as

inward-turning. Extraverts are more talkative and assertive; they draw energy from action

whereas introverts draw energy from reflection. Extraversion has been shown to correlate

strongly with speaker charisma (Bono and Judge, 2004). People that speak loud, quickly, and

and Venkatachalam, 2013), or facial features (e.g., Cook and Mobbs, 2017; Jia, van Lent, and Zeng, 2014; Kamiya, Kim, and Suh, 2016) to infer executive characteristics. 7 For example, in the factor analysis of John and Srivastava (1999), extraversion loads positively on talkative, assertive, active, energetic, outgoing, outspoken, dominant, forceful, enthusiastic, show-off, sociable, spunky, adventurous, noisy, and bossy, and the extraversion factor loads negatively on quite, reserved, shy, silent, withdrawn, and retiring.

7

often are rated as likable, smart, and more competent (Giles and Street, 1994; Paulhus and

Morgan, 1997; and Swann and Rentfrow, 2001).

Extraversion is relatively easy to detect due to its effect on communication patterns. In

spoken text, extraverts have a higher verbal output, speak more quickly and with fewer pauses,

use less word variety and more informal language, and are more assertive (Scherer, 1979;

Furnham, 1990; Gill and Oberlander, 2003). Extraverts also use more emotion words and show

more agreements and compliments than introverts (Pennebaker and King, 1999).

Researchers in psycholinguistics and artificial intelligence have developed personality

models based on linguistic outputs (e.g., Argamon et al., 2005; Oberlander and Nowson, 2006;

Mairesse et al., 2007).8 We rely on the trained personality algorithms of Mairesse et al. (2007),

who document a significant ability to predict observer-based personality scores based on spoken

language. They provide four linguistic models that generate continuous personality estimates

from a set of texts. They include two linear models: ordinary least squares and the support vector

regression (Vapnik, 1997), as well as two tree-based models that analyze conditional

relationships between linguistic features: the M5’ Model Tree and M5’ Regression Tree

(Quinlan, 1992). We refer the reader to Section 1 of the Internet Appendix for details on the

mechanics of each model.9

The four linguistic algorithms each implement regression models where the dependent

variable is the extraversion score of the individual and the explanatory variables are word

categories from the LIWC (Pennebaker et al., 2001) and MRC linguistic databases (Coltheart,

1981). Features include variables like the frequency of the use of different pronouns (i.e., “I”

8 An alternative approach for measuring personality traits is Profiler Plus (https://profilerplus.org), which has been used on textual data to infer the need for power, achievement, and affiliation (McClelland and Winter, 1969), and the seven leadership traits (Hermann, 1999). 9 We use an equal-weighted average over the output of all four algorithms to construct our measure of executive extraversion. However, our findings are qualitatively similar using any of the four algorithms individually.

8

versus “We”), the use of articles, the use of assent words (e.g., agree, ok, yes), the total number

of words used, the ratio of unique words to total words (i.e., the type/token ratio), the total

number of words used longer than six letters, the tendency to use negations, the tendency to use

positive or negative emotion words, the use of imagery words, etc.

Mairesse et al. (2007) find that the following LIWC and MRC linguistic features are

significantly positively related to extraversion at the 1% level: affective or emotional processes,

anger, metaphysical issues, negative emotions, physical sates and functions, positive feelings,

religion, swear words, imageability, meaningfulness, word count, and language frequency, while

extraversion is negatively related to assent words and word uniqueness. Summarizing, extraverts

tend to use words that are more emotionally charged and are easier to visualize. They also

exhibit greater verbosity, use more common language, and exhibit less word uniqueness (i.e.,

repeat themselves).

Mairesse et al. (2007) confirm that the above linguistic features exhibit a significant

ability to predict observer-based extraversion scores from transcribed speech. Specifically, they

obtain binary classification accuracies as high as 73%, with a statistically significant

improvement over the baseline model (with 50% accuracy) at the 5% level. Although the

algorithms in Mairesse et al. (2007) are generally less successful in capturing other Big Five

personality traits from spoken language, we also consider measures of executive Emotional

Stability, Openness, Agreeableness, and Conscientiousness as controls in our analysis.

2.2 Estimation of Executive Extraversion

To measure executive extraversion, we apply the linguistic algorithms of Mairesse et al.

(2007) to the spoken language of executives’ from the questions and answers (Q&A) portion of

conference calls. We collect conference call transcripts from two sources: Thomson Reuters

9

StreetEvents and SeekingAlpha.com. From the Thomson Reuters dataset, we obtain a total of

88,792 transcripts (with matched I/B/E/S CUSIPs) from 2006-2011, and from the Seeking Alpha

dataset, we obtain 65,447 transcripts from 2006-2013. For each report, we retrieve an identifying

key, report title, date of transcript, and transcript text. When both datasets cover the same call,

we select the longer transcript.10

Conference call transcripts are generally well structured. Labels above each section allow

us to automatically identify speakers’ names and their associated speech. Transcripts consist of

three sections: the (1) call participants list, (2) management discussion section, and (3) Q&A

section. To extract speaker dialogue from the Thomson Reuters transcripts, we first convert the

PDF files to ASCII text, preserving as much of the document layout as possible. Transcripts with

multiple column layouts are omitted from the sample to ensure that word ordering is preserved.

Headers, footers, disclaimers, and branding information use consistent language and formatting

and are removed. For each quotation in the transcript we extract the speaker name in the line

directly above it and require that the name is included in the list of call participants. Transcripts

from the Seeking Alpha sample are retrieved as XHTML documents and are more easily parsed.

Page tags allow us to distinguish speakers and different transcript sections.

We next estimate the extraversion of each executive based on her dialogue from the Q&A

portion of each call. We focus on the Q&A portion because it is less scripted than the

management discussion section (Hollander, Pronk, and Roelofsen, 2010).11 The trained

personality models (Mairesse et al., 2007) take as input the dialogue for each executive from the

10 Roughly 41% (44%) of the combined conference call sample is from Seeking Alpha (Thomson Reuters). For the 15% overlapping sample, we choose the longer call to address potentially incomplete transcripts or quotations with “unknown” speaker labels. Also, to control for potential textual differences between the data sources, we include a Seeking Alpha (SA) dummy variable in Equation (1) which is one (zero) if the call transcript is from Seeking Alpha (Thomson Reuters). 11 Estimating extraversion from the full call (i.e., both the management discussion and the Q&A session) generates similar results.

10

call and generate personality ratings based on linguistic feature counts. Names extracted from the

transcripts are then matched with the Execucomp database by name and 6-digit CUSIP. We strip

all names of titles and suffixes (e.g., PhD, MD, Jr., etc.). A closest name match is chosen by

spelling distance as calculated with SAS fuzzy matching functions. For each name in the

transcript database, we are able to find a unique name with the minimum spelling distance. To

ensure match quality, we manually filter the non-exact name matches to obtain the final matched

pairs. We require that each firm have non-missing CRSP, Compustat, I/B/E/S, and Execucomp

data. We also limit the sample to Chief Executive Officers (CEOs) and Chief Financial Officers

(CFOs), and require at least 150 words of response for each call. The matched sample includes

39,783 CEO-call observations from 2,649 unique CEOs and 37,644 CFO-call observations from

2,953 unique CFOs. Unless otherwise stated, we further limit the sample to executives who

appear in at least three calls, resulting in a sample of 2,266 unique CEOs and 2,362 unique

CFOs.

For all CEOs and CFOs who appear on at least three conference calls, we construct an

extraversion score that is aggregated across all calls (Aggregate Extraversion). To compute

Aggregate Extraversion, we first winsorize the extraversion estimates from each of the four

different linguistic algorithms (discussed in the Internet Appendix) at the 1st and 99th percentile.

Next, we average across all four linguistic measures to compute a call-level measure of

extraversion (Call Extraversion).12 Aggregate Extraversion is then computed as the weighted

average Call Extraversion based on all calls, where each call is weighted by the number of words

spoken on the call. Thus, we treat extraversion as a time-invariant manager fixed effect.13

12We construct measures of the other Big Five personality traits (i.e., Emotional Stability, Openness, Agreeableness, and Conscientiousness) in a similar fashion. 13 Aside from helping reduce measurement error, using all available calls to measure extraversion allows us to include executives immediately, rather than waiting until they appear on three calls, which allows us to examine

11

2.3 Other Variable Construction and Summary Statistics

For all CEOs and CFOs who appear on at least three conference calls, we construct a

number of additional variables. Specifically, we collect information on total compensation (Total

Comp.), the number of years that they have held their current position (Tenure), whether they

were replaced during the year (Turnover), their gender (Male), their age (Exec Age), their

tendency to hold in-the-money stock options (Overconfidence), and the age at the time they were

first appointed to CEO or CFO (First Age) from Execucomp. We also proxy for whether the

executive was a founder based on whether the executive became CEO (as reported in

Execucomp) within one year of when the firm went public (Founder). We collect information on

the number of outside directorships held (Directorships) from RiskMetrics, and we compute a

measure of optimism based on the tone of the executive during the Q&A portion of the

conference call (Optimism). Following Li et al. (2014), we also compute the ratio of the number

of words spoken by the CEO (CFO) during the conference call to the number of words by all

company executives during the conference call (Percentage CEO (CFO) Text).

For the CEO sample, we also collect a number of managerial characteristics from

BoardEx including dummy variables for whether the executive graduated with honors

(GradHonors), received an MBA (MBA), a PhD (Doctorate), or an Ivy League education (Ivy

League). We also compute the sum of other external executives or directors related to the CEO

through past professional connections, social connections, and past universities attended

(Rolodex), as defined in Engelberg, Gao, and Parsons (2013); as well as a measure of general

managerial skills based on the breadth of the CEO’s past work experience (General Ability Index

or GAI) as defined in Custódio, Ferreira, and Matos (2013).

changes in relatively short-windows following CEO transitions. However, using forward looking calls to measure extraversion raises concerns of reverse causality, and we address this issue in Section 4.3.

12

We use CRSP data to compute the number of months since the firm first appeared in

CRSP (Firm Age), the standard deviation of daily returns (Vol), the firm’s market capitalization

(Size), share turnover (Share Turnover), the Amihud (2002) illiquidity ratio (Illiquidity), and the

annual return on the stock less the return on the CRSP value-weighted market portfolio (Return).

We collect information on Assets, Sales, Investment, Operating Cash Flows (Prof), Return on

assets (ROA), sales growth, profit margin and Tobin’s Q (Q) from Compustat. We also consider

Firm Efficiency as described in Demerjian, Lev, and McVay (2012).14 We collect from Factiva

the total number of media articles in the Wall Street Journal that mention the firm (Media

Articles) as well as the total number of words across all media articles (Media Words). We

measure the number of brokerage houses covering a firm (Analyst Coverage) from IBES; and we

collect information on the number of broker-hosted investor conferences attended by a firm

(Conference Presentation) from Bloomberg Corporate Events Database.15 More detailed variable

definitions are presented in Appendix A.

Panels A and B of Table 1 present descriptive statistics for the CEO and CFO sample,

respectively. For brevity, we only report summary statistics for the personality variables and the

dependent variables of interest (i.e., Compensation, Outside Directorships, and First Age). In

Table 2, we explore how other managerial characteristics relate to extraversion. The sample

includes 12,213 CEO-year observations and 11,326 CFO-year observations with non-missing

extraversion scores over the 2004-2013 sample period.16 The average aggregate extraversion

scores for CEOs is 4.15 compared to 3.61 for CFOs. This difference between the two estimates is

14 We thank Peter Demerjian for making the data available: http://faculty.washington.edu/pdemerj/data.html 15 See Bushee, Jung, and Miller (2011) and Green et al. (2014a, 2014b) for more details on broker-hosted investor conferences. 16 We begin gathering data in 2004, which permits us to have at least two years of data prior to executive transitions that occur in 2006 (the first year in which conference call data is available) while at the same time reducing potential backfill bias concerns. Beginning our data gathering from 1992-2003 yields similar conclusions.

13

significant at the 1% level, consistent with extraversion being a more prevalent trait for CEOs

relative to CFOs. However, the observed differences may also reflect differences in the expected

communication roles of CEOs and CFOs in conferences calls. In our analysis, we consider CEOs

and CFOs separately.

The aggregate extraversion scores of both CEOs and CFOs are based on a relatively large

sample of conference calls (20.2 and 18.8 average calls, respectively). Not surprisingly, CEOs,

on average, have a larger total compensation than CFOs ($5.6 million vs. $2.0 million) and sit on

more outside boards (0.45 vs. 0.02). CEOs and CFOs appear to be similar along the other four

dimensions of personality. Panel C reports the summary statistics for firm-level characteristics.

The sample includes 1,633 unique firms and 12,213 firm-years for which the CEO has a non-

missing extraversion score. The average firm in our sample has assets of $17.7 billion, a market

equity value of $8.5 billion, and sales of $6.9 billion.

2.4 Comparing Linguistic Algorithms to Observer-Assessed Ratings of Extraversion

An important advantage of using linguistic algorithms to infer extraversion is that they

can be applied to a large sample of executives in an objective manner. An alternative approach,

commonly used in the psychology literature, is to obtain observer assessments of personality. An

advantage of survey-based assessments is that observers may incorporate information not

contained in the call transcripts (e.g., pitch tone or variation, rate of speech, etc.), and thus may

more accurately capture extraversion. However, observer-based assessments are difficult to

objectively replicate, and more importantly, they are not practical for large samples of

executives.

Nevertheless, to help bolster the validity of our linguistic measure, we compare our

textual algorithm to observer assessments for a subset of the sample. In particular, we gather

14

audio excerpts for 100 conference calls from Earningscast.com. We choose calls for the 50 most

and least extraverted CEOs in our initial sample that have conference call audio files available

from Earningscast.17 For each CEO, we select the most recent call available during our sample

period, and from within each call we create an audio excerpt from the CEOs’ response to the first

question during the Q&A portion of the call. We require CEOs to speak at least 25 words during

the response, and we truncate longer audio excerpts at roughly one minute (e.g., near the end of a

sentence, etc.). The average audio response length is 42 seconds (minimum of 9 seconds and

maximum of 64 seconds) and contains 112 spoken words.

We ask BBA students to evaluate the level of extraversion for each executive using a 1 to

7 scale, based on the audio excerpt. Observers also have access to the written transcripts of the

responses to the questions. Each student evaluated 20 executives (10 extraverts and 10 introverts

in random order), and we obtained three separate evaluations for each executive. Additional

details are provided in Section IA.2 of the Internet Appendix, along with examples of executive

responses rated most and least extraverted by the observers with links to corresponding audio

files, as well as whether the linguistic algorithm classifies the executive as extraverted or

introverted.

Anecdotally, the speech patterns of listener-rated extraverts generally corroborate the

features emphasized by the linguistic algorithm. For example, the most extraverted excerpt in

Section IA.2 of the Internet Appendix contains more emotionally-charged words (e.g.,

wonderful, terrific, shame, etc.) relative to the least extraverted excerpt. The extraverted quote

also has a greater word count (120 words vs. 82), less word uniqueness (71.6% type/token ratio

17 We focus on executives with the highest and lowest extraversion scores to increase the likelihood that personality can be detected in one brief audio clip. However, this approach limits our ability to generalize the findings to the full sample of CEOs. In Section 3, we consider several additional validity tests of the extraversion measure for the full sample of executives.

15

vs. 74.4% for the least extraverted quote), and more words that score high on imageability. As a

result, the linguistic algorithm agrees with the listener assessments in this binary example. The

remainder of this section explores more systematic evidence across the 100 audio subsamples.

The first piece of supporting evidence from the audio subsample is that we find

extraverted CEOs speak faster than introverts. A higher rate of speech for extraverts is one of the

most frequently described aspects of personality on language (e.g., Furnham, 1990; and Feldstein

and Sloan, 1984). We observe that extraverted CEOs, as implied by the transcript-only linguistic

algorithm, speak 170.4 words per minute in the audio excerpts vs. 154.1 for introverts, and the

difference in means is statistically significant at the 1% level.

Turning to the listening-based evaluations of extraversion, we follow the

psycholinguistics literature and consider binary classification accuracy tests. We define the

executive as listener-based extraverted if the average listener-based rating (across the three

students) is in the top half of the distribution of listener-based scores. All other executives are

defined as listener-based introverted. Similarly, we classify an executive as algorithm-based

extraverted (introverted) if the executive is in the top (bottom) half of the distribution of

extraversion based on the algorithms score. We find that our listener-based and algorithm-based

measures agree with each other 68% of the time. If we limit the sample to more confident

listener assessments, in which the average listener score is greater than or equal to five

(extrovert) or less than or equal to three (introvert), the agreement between listeners and the

linguistic algorithm rises to 75.0%. Both ratios are statistically different from the null of 50% at

the 1% level. The economic magnitudes are generally in line with Mairesse et al. (2007), which

estimates binary classification accuracies between observer ratings and various statistical models

ranging from 59% to 73% (their Table 14). We find the correlation between the average listener

16

extraversion score and our linguistic extraversion score is 0.40, and the average correlation

between a single listener rating and other-observer average is 0.47.

Despite our rudimentary approach which relies on a single brief audio excerpt of at most

one minute for each executive, the general agreement between listener assessments and our

linguistic algorithm provides support for our measure of extraversion. In the next section, we

provide additional descriptive statistics to further evaluate the validity of our linguistic

extraversion measure.

3 Characterizing Executive Extraversion

In this section, we provide additional descriptive statistics to better understand our

measure of executive extraversion at the conference call level.

3.1 Determinants of Conference Call Extraversion

We first examine what drives variation in extraversion at the conference-call level. We

are particularly interested in understanding whether our measure of extraversion captures a stable

personality trait or merely reflect the circumstances of the call. For example, executives at firms

whose earnings exceed expectations may be happy to discuss the quarter and thus appear

extraverted, while executives reporting bad news may appear introverted.

An extraversion measure that is stable across calls is preferable to one that changes with

firm fundamentals for two broad reasons. First, using survey results, psychologists have found

that extraversion is a highly stable personality trait (Costa and McCrae, 1988). In our setting,

measured extraversion is meant to be representative of others’ view of the executive in a variety

of settings. Our call-based measure is likely to be a stronger proxy for perceived extraversion if it

is stable over time. Second, if extraversion varies considerably with changes in firm

17

fundamentals, there is a greater concern that extraversion may be capturing some omitted

fundamental variable.

To examine the determinants of conference call extraversion, we estimate the following

regression in Table 2:

Call Extraversion = D + β1Rett-63,t-2 + β2Rett-1,t+1 + β3Rett+2,t+63 + β4Earnings Call + (1) β5MBE+β6 Surprise + β7Loss + βCharacteristics+ Qtr +Manager + ε.

Call Extraversion is the call-level extraversion score based on averaging the winsorized

values of the extraversion estimates from the four linguistic models described in Section IA.1 of

the Internet Appendix. Rett-63,t-2 captures the return in the quarter prior to the call (i.e., the past 2

to 63 trading days). We also control for returns around the call (Rett-1,t+1) and returns over the

subsequent quarter (Rett+2,t+63). Roughly 83% of our sample of conference calls falls in the four-

day window [-1,2] around the earnings announcement (day 0), and we examine whether

extraversion differs systematically on earnings vs. other calls by including an Earnings Call

dummy variable. For the subset of earnings calls, we also include three variables related to the

earnings surprise. Meet-or-Beat is a dummy variable equal to one if the firm meets or beats the

mean consensus analyst forecast for the most recent quarter. Surprise is the difference between

quarterly EPS and the mean consensus analyst forecast scaled by the stock price at the beginning

of the quarter. Loss is a dummy variable equal to one for firms reporting negative earnings.

Characteristics is a vector that includes the following variables for both CEOs and CFOs

Tenure, Exec Age, Male, and Optimism. The CEO sample also includes Overconfidence,

Founder, General Ability Index, Rolodex, MBA, Doctorate, GradHonors, and Ivy League. All

variables are defined in Appendix A. All the regressions also include quarter fixed effects, and

Specifications 3 and 6 (of Table 2) include manager fixed effects. All continuous variables

18

(including extraversion) are standardized to have mean 0 and variance 1, and standard errors are

clustered by firm.

Specification 1 of Table 2 reports the results of equation (1) for the CEO sample before

including managerial characteristics or manager fixed effects. We find modest evidence that firm

fundamentals influence extraversion. For example, managers at firms with poor returns over the

past quarter or in the three days surrounding the conference call have higher extraversion ratings.

However, the size of the relationship between returns and extraversion is economically small.

For example, a one standard deviation change in either past quarterly returns or event-time

returns accounts for a change of 0.02 standard deviations in extraversion.

Specification 2 adds managerial characteristics. Extraversion tends to be higher for

males, younger executives, executives with greater tenure, and founders. Extraverts also appear

to have optimistic tones. We find no evidence that the education variables (MBA, Doctorate,

Grad Honors, and Ivy League) are correlated with extraversion. Collectively, including the

managerial characteristics improves the R-squared from 0.95% to 4.46%, suggesting that

managerial characteristics have a relatively modest ability to explain extraversion in the cross

section of CEOs.

We add manager fixed effects in Specification 3. We exclude all managerial

characteristics (except tenure) due to the fact that they are highly persistent over time. Most

interestingly, adding manager fixed effects to the CEO regression results in the R-squared

jumping to 49.6%. Further, the relatively high R-squared is driven almost entirely by the

inclusion of manager dummies. The within R-squared is only 1.02%, indicating that after

controlling for manager fixed effects, the additional variables have relatively little explanatory

power. The important role of manager fixed effects, and the weak effect of fundamental variables

19

in explaining extraversion suggests that the call-level estimates of extraversion largely capture a

stable personality trait of the manager rather than reflect the circumstances surrounding the call.

The patterns for CFOs in Specifications 4 through 6 are similar. However, one exception

is that CEOs appear more extraverted during earnings calls, whereas CFOs appear less

extraverted. This may capture the different responsibilities of CEOs and CFOs in earnings calls

compared with other calls, such as those discussing merger activity. In future tests, we control

for the relation between call extraversion and firm fundamentals by constructing an extraversion

measure based on the residuals from Specifications 1 and 4.18 In particular, for each executive,

we define Extraversion as the weighted average residual extraversion across all calls, where each

call is weighted by the number of words spoken in the Q&A portion of the call by the executive.

We also control for the relation between extraversion and other managerial characteristics by

directly including the set of managerial characteristics as controls.

3.2 Persistence in Call Extraversion

The significant increase in R-squared following the inclusion of manager fixed effects is

consistent with extraversion being highly persistent at the manager level. However, in many

cases we only observe the manager working for one firm, which points to the possibility that

manager fixed effects are simply capturing a firm fixed effect. To distinguish between a manager

and a firm fixed effect, we compare the persistence in extraversion for a sample of firms with

and without manager turnover. In particular, we split the conference call sample into two

periods, 2006-2009 and 2010-2013, and calculate separate executive extraversion scores (using a

minimum of three calls) for each period. We then compare correlations of extraversion scores

across the two sample periods for firms that do not experience a CEO change (i.e., a single CEO 18 The results are robust to using unadjusted extraversion ratings (i.e., Aggregate Extraversion as reported in Table 1).

20

across the two periods) with firms that do experience a CEO change (i.e., multiple CEOs across

the two periods). If extraversion is persistent over time at the manager level, we would expect to

observe a higher correlation for the former group (same CEO, same firm) than for the latter

group (different CEOs, same firm). We observe 1,145 firms with no CEO change and 609 firms

with a CEO change in our sample.

We observe that the correlation in the extraversion score for the same CEO over the two

sample periods is 0.75, whereas the correlation in extraversion for the same firm with two

different CEOs is only 0.25 (the results are tabulated in the internet appendix Table IA.2). While

both correlation estimates are significantly greater than zero, the correlation between

extraversion in the two periods is much lower for firms that experience a CEO change, and the

difference is statistically different from zero (t-statistic=12.37).19 Analogous tests for the sample

of CFOs produce similar results, with correlations in extraversion scores of 0.63 across periods

for the same firm and manager, and only 0.30 when there is a change in management at the firm,

and the difference between the two estimates is highly significant (t-statistic=8.59). These

findings indicate that managerial extraversion is distinct from firm-level extraversion.

3.3 Correlation Matrix

Although the persistence in executive extraversion diminishes following managerial

turnover, the coefficient estimate remains significantly greater than zero indicating that some

firms have a persistent preference for extraverts. To provide some descriptive evidence on the

types of firms that tend to hire extraverts, we include a correlation matrix between Extraversion

19 A more powerful test would be to examine the subset of managers that are CEOs for two different firms. Although our sample of CEOs who speak in conference calls for two different firms is small (19 observations), the correlation in measured extraversion for a given CEO across two firms is 0.76, consistent with persistence in extraversion over time for a given individual in different firm environments.

21

and the following firm characteristics (all in natural logs): Sales, Q, Vol, and Firm Age. For

reference, we also include a host of other CEO characteristics.

We expect extraverts to be overrepresented at larger firms, as well as growth firms which

tend to more visible, have greater investment opportunities, and offer higher executive

compensation (Smith and Watts, 1992; and Murphy, 1999). Extraversion is also associated with

greater risk taking (Zuckerman and Kuhlman, 2000), suggesting that extraverts may be attracted

to riskier firms, such as younger firms or those with more volatile stock returns. The correlation

matrix generally reveals support for these conjectures. In particular, we document a strong

positive correlation between Extraversion and Sales and Q, and a strong negative relation

between Extraversion and Age.20 We do not, however, find any evidence that extraverts work at

firms with greater return volatility.

Table 3 also reveals a strong correlation between Extraversion and Percent CEO Text.

The positive relation is not surprising given the strong correlation between extraversion and

verbosity (Mairesse et al., 2007). Importantly, all of our tests will control for Percent CEO Text,

which allows us to explore the incremental explanatory power of Extraversion after controlling

for the impact of Percent CEO Text on executive and firm outcomes documented in Li et al.

(2014).

4. Executive Extraversion and Career Outcomes

A long literature in psychology, originating with Mann (1959), documents a relation

between extraversion and leadership emergence. In this section, we explore whether the

perceived leadership advantage of extraverts translates into greater labor market success.

20 In Table IA.3 of the Internet Appendix, we examine the determinants of hiring extraverted CEOs in a multivariate regression that includes all four firm characteristics as well as industry fixed effects. We continue to find a strong relation between extraversion and both Sales and Q; the relation between Extraversion and Age remains negative but is no longer statistically significant.

22

4.1 Extraversion and Executive Compensation

We begin investigating the effects of extraversion on compensation using the following

panel regression:

𝐿𝑜𝑔(𝐶𝑜𝑚𝑝𝑒𝑛𝑠𝑎𝑡𝑖𝑜𝑛)𝑖𝑡 = 𝛽1𝐸𝑥𝑡𝑟𝑎𝑣𝑒𝑟𝑠𝑖𝑜𝑛𝑖 + 𝛾𝑭𝒊𝒓𝒎𝑪𝒉𝒂𝒓 + 𝛿𝑷𝒆𝒓𝒇𝒐𝒓𝒎𝒂𝒏𝒄𝒆 + ωCEOChar +𝑌𝑒𝑎𝑟𝑡 + 𝐹𝐸𝑖 + 𝜀𝑖𝑡 . (3)

Our primary measure of compensation is total compensation, which consists of salary, bonus,

value of restriction stock granted, value of options granted, long-term incentive payout, and other

compensation (TDC1 as reported in Execucomp). We winsorize compensation values at the 1st

percentile to address $1 salaries.

FirmChar is a vector that includes Log (Sales), Log (Assets), Log (Q), Log (Vol), and Log

(Firm Age); Performance is a vector that includes Log (Sales Growth) Fiscal Ret, Fiscal Ret (t-

1), Prof, Prof Growth, and Loss; and CEOChar is a vector that includes all the CEO

characteristics reported in Specification 2 of Table 2, plus Percent CEO Text and the other Big 4

personality traits. All continuous variables are standardized to have mean 0 and variance 1.

Appendix A provides more detailed variable definitions. All specifications include year fixed

effects (FE) and either Industry or Firm fixed effects.21

Specifications 1 through 4 of Table 4 report the results with industry fixed effects. Prior to

including firm characteristics, performance measures, or CEO characteristics, we observe that a

one-standard deviation increase in extraversion is associated with a 17.90% pay premium.

Specification 2 includes firm characteristics. Consistent with prior literature (e.g., Gabaix and

Landier, 2008), we find that CEO compensation is strongly related to proxies for firm size (Sales

21 Firm fixed effects help control for unobserved (time-invariant) firm characteristics, yet they also have shortcomings. Firm fixed effects require valid extraversion scores for at least two CEOs for the same firm, which precludes 2/3 of the sample. Aside from allowing a much larger sample, industry fixed effects are useful for conducting mediation analysis, since firm characteristics (e.g., size, growth opportunities, etc.) exhibit significantly more variation across firms than within firms.

23

and Assets) and growth opportunities (Q). Further, including firm characteristics significantly

reduces the extraversion pay premium from 17.90% to 6.10%. This suggests that extravert’s

tendency to be employed by larger firms and more growth oriented firms (Tables 3 and IA.3)

explains much, but not all, of the compensation premium.

Specification 3 includes the performance measures. As expected, compensation is greater

among firms with higher sales growth, higher returns, and higher profitability. Controlling for

the performance metrics further reduces the extraversion pay premium to 5.70%, but the estimate

remains highly significant. Finally, Specification 4 includes CEO characteristics. Consistent with

Custódio, Ferreira, and Matos (2013), CEOs with higher general ability or CEOs that serve as

chairmen earn significantly more.22 However, controlling for CEO characteristics has virtually

no effect on the extraversion pay premium. Specifications 5 through 8 include firm fixed effects.

Comparing Specification 4 to Specification 8, we find that including firm fixed effects

substantially increases the r-squared of the model (58.60% versus 81.26%) and also increases the

coefficient on Extraversion (5.70% versus 7.57%).

Overall, the evidence from Table 4 suggests that after controlling for basic firm

characteristics, the extraversion pay premium ranges from 5.70% to 7.57%. The average CEO

compensation is $5.5 million which suggests that a 5.70% (7.57%) premium would translate into

an additional $316,000 ($420,000) in annual compensation. The extraversion pay premium is in

line with existing literature on CEO characteristics and compensation. For example, Custódio,

Ferreira, and Matos (2013) find that a one-standard deviation increase in CEO’s general ability

22 Two CEO characteristic coefficients appear inconsistent with prior literature. We find an insignificant coefficient on the Percent CEO Text measure of Li et al. (2014), although this is driven by the correlation between Extraversion and Percent CEO Text (Table 3). Excluding Extraversion, the coefficient on Percent CEO Text increases to 2.95% (t=2.19). Also, we observe an insignificant relation between Rolodex and compensation (inconsistent with Engelberg, Gao, and Parson, 2013), which is partially driven by the correlation between Rolodex and GAI. Excluding GAI, the coefficient on Rolodex is 4.07% (t=2.84).

24

(GAI) is associated with an 11.7% increase in pay (Specification 3 of their Table 5), Engelberg,

Gao, and Parsons (2013) find that a one-standard deviation increase in professional connections

(Rolodex) is associated with a 10.0% increase in CEO pay (134.69 × 0.00076) and Li et al.

(2014) find that a one-standard deviation increase in industry knowledge (Percent CEO text) is

associated with a 5.5% premium (24.2 × 0.229). Our findings are also in line with the literature

that explores the premium associated with non-cognitive traits in the broader population. For

example, Case and Paxson (2008) find compensation effects of 9.2% associated with an

interquartile movement in height of roughly 4 inches. In our setting, an interquartile movement is

equal to 1.2 standard deviations, implying that an interquartile movement in extraversion is

associated with between a 6% to 9.6% premium.

4.2 Extraversion and Executive Compensation - Robustness

In this section, we consider several alternative specifications that for brevity are tabulated

in Table IA.4 in the Internet Appendix. In the discussion that follows, we focus on the

Extraversion coefficient from the full model that includes all controls and firm fixed effects (i.e.,

Specification 8 of Table 4). We first decompose CEO total compensation into cash versus equity

compensation. We find that a one-standard deviation increase in extraversion is associated with a

3.86% (t=2.85) premium in cash salary and a 13.84% (t=2.73) increase in equity compensation.

We also examine possible non-linearities by replacing Extraversion with extraversion quintile

rankings. We find a monotonic, and essentially linear compensation premium. Specifically,

relative to the executives in the bottom 20% of extraversion, the premium for being in the 2nd

through 5th quintile of extraversion are: 4.58%, 9.45%, 13.90%, and 16.29%, respectively.

We next explore whether extraverted CFOs also experience a pay premium. We find that a

one-standard deviation increase in CFO Extraversion is associated with a 9.07% (t=5.32)

25

premium in CFO compensation. Finally, we examine whether CEO relative extraversion

(Relative Extraversion), defined as CEO Extraversion – CFO Extraversion, predicts CEO

Relative Compensation, defined as CEO Log (Compensation) – CFO Log (Compensation). We

find that a one-standard deviation increase in Relative Extraversion is associated with a 9.60%

(t=5.56) increase in Relative Compensation.

4.3 Reverse Causality and Omitted Variables

Our measure of Extraversion is based on all available calls over the executive’s tenure

with the firm. While the treatment of extraversion as a manager fixed effect offers significant

methodological advantages, the use of forward-looking information raises concerns of reverse

causality (i.e., executives who experience increases in compensation become more extraverted).

We explore this possibility by estimating the following panel regression:

ΔExtraversioni,t+1 = α + β1ΔCompi,t + β2ΔCompi,t+1 + εit. (4)

The dependent variable is the extraversion score based on all calls in year t+1 less the

extraversion score for the same executive based on all calls in year t. ΔCompt+1 measures the

abnormal compensation in year t+ l less the abnormal compensation in year t. Abnormal

compensation is based on the residuals from Specification 4 of Table 4 after excluding

extraversion. We standardize all variables to have mean 0 and variance 1, and cluster standard

errors by firm. In untabulated findings, we estimate β1 to be 0.00 (t=0.20), and β2 to be -0.03 (t=-

1.85). Both coefficients are economically small, and the negative coefficient on β2 biases us

against finding a compensation premium. These findings suggest reverse causality is not a

substantial concern.

Another important concern is that extraversion may be correlated with other (perhaps

unobservable) managerial attributes that drive executive labor market success. In our analysis,

26

we control for a large set of existing managerial attributes known to influence executive

compensation including optimism, overconfidence, founder, education and education quality

(Falato, Li, and Milbourn), general ability (Custódio, Ferreira, and Matos, 2013), the extent of

the executive’s network (Engelberg, Gao, and Parsons, 2013), and CEO’s firm-specific

knowledge (Li et al., 2014). Our findings indicate that the career benefits associated with

extraversion are distinct from these existing managerial attributes.

Nevertheless, there may be additional characteristics that are correlated with extraversion

that may influence compensation, such as height or vocal pitch (Case and Paxson, 2008; Mayew,

Parsons, and Venkatachlam, 2013). We explore the potential severity of omitted variable bias

using the framework of Oster (2016), which demonstrates that omitted variable bias is less likely

to be a problem when 1) the coefficient of interest is stable after including a wide-range of

controls and 2) including controls significantly increases the R-squared of the model. In our

setting, we focus on Specifications 2 through 8 in Table 4, as it is unlikely that any omitted

individual characteristic will have the same type of explanatory power for compensation as firm

characteristics such as sales, assets, and Q.23 The fact that the coefficients are fairly stable

(ranging from 5.7% to 7.9%), despite the R-squared substantially increasing from 55% to 81% is

reassuring. While the diagnostics above are comforting, we acknowledge that we cannot

completely eliminate concerns regarding omitted variables. For example, extraversion may lead

to differences in early life experience such as increased participation or success in previous

leadership roles.24 A cautious interpretation of the relation between extraversion and

23 Even if one includes Specification 1 in the analysis, a more formal analysis of the bias suggests that omitted variables are unlikely to completely explain our results. Specifically, under “equal selection” in which observables and unobservable have the same proportional influence on extraversion, we estimate that the bias in Specification 8 of Table 4 is at most 2.68%. This suggests a lower bound for the extraversion pay premium of 4.89% (7.57% - 2.68%), which would be significant at a 5% level (t=2.17) if the standard errors remain unchanged. 24 Consistent with this view, evidence suggests that the career benefits of height and vocal pitch accrue relatively early in life (e.g., Perisco et al., 2004; Mayew and Venkatachalam, 2012).

27

compensation is that it reflects the direct effect of extraversion as well as any indirect effects

through earlier, unobserved personal or career successes.25 Our findings nevertheless highlight

the importance of a psychologically-motivated but previously unobserved managerial trait in

explaining career success.

4.4 Extraversion and Age of First Appointment to CEO or CFO

We next examine whether extraverted executives are first promoted to CEO at an earlier

age. We conjecture that executives with greater perceived ability will likely spend less time on

the corporate ladder and reach these prestigious positions more quickly (Falato, Li, and

Milbourn, 2015). Equivalently, the hurdle for appointing a young CEO is higher since such

executives have less experience, and thus younger CEOs must have other perceived advantages

to compensate for their lack of experience. We examine whether extraverted executives are

promoted to CEO at an earlier age by estimating the following regression.

𝐹𝑖𝑟𝑠𝑡_𝐴𝑔𝑒𝑖 = 𝛽1𝐸𝑥𝑡𝑟𝑎𝑣𝑒𝑟𝑠𝑖𝑜𝑛𝑖 + 𝛾𝑭𝒊𝒓𝒎𝑪𝒉𝒂𝒓 + 𝜔𝑪𝑬𝑶𝑪𝒉𝒂𝒓 + 𝑌𝑒𝑎𝑟𝑡 + 𝐹𝐸𝑖 + 𝜀𝑖𝑡. (4)

First_Age is the age at which the executive was first appointed to CEO. FirmChar and

CEOChar are defined as in equation 3. Yeart denotes year fixed effects, where t is the calendar

year in which the executive first becomes CEO (to control for systematic variation over time in

the tendency to hire young CEOs). The independent variables are defined in the Appendix A,

and FE reflects either industry or firm fixed effects.

Table 5 reports the results. We find that a one standard deviation increase in extraversion

is associated with being appointed to CEO 0.79 to 1.12 years earlier. We also repeat the analysis

separately for external versus internal hires. We conjecture that our estimates may be stronger

25 We note that this caveat is relevant for many of the documented determinants of compensation. For example, breadth of work experience, strength of personal networks, and even education may proxy for unobserved personal characteristics or early life experiences.

28

among external hires, since internal promotions to CEO are more likely to depend on firm-

specific factors (e.g., the age of the current CEO) that are largely outside of the control of the

executive. Consistent with this view, we find that extraversion is associated with being appointed

to CEO 0.98 years earlier for external hires, compared to 0.65 years earlier for internal hires.

4.5 Extraversion and Turnover

Executives with greater perceived ability may also be less likely to experience turnover,

particularly following poor performance. To examine the relation between CEO turnover and

extraversion, we estimate the following logistic regression:

𝐶𝐸𝑂 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟𝑖𝑡 = 𝛽1𝐸𝑥𝑡𝑟𝑎𝑣𝑒𝑟𝑠𝑖𝑜𝑛𝑖 + 𝛾𝑭𝒊𝒓𝒎𝑪𝒉𝒂𝒓 + 𝛿𝑷𝒆𝒓𝒇𝒐𝒓𝒎𝒂𝒏𝒄𝒆 + ωCEOChar +𝑌𝑒𝑎𝑟𝑡 + 𝐹𝐸𝑖 + 𝜀𝑖𝑡 ,

(5)

where the dependent variable, CEO Turnover, takes the value of 1 if a firm changes its CEO that

year, and 0 otherwise. All other variables are defined as in Equation 3.

Table 6 reports the odds ratios from the logistic regression. Specification 1 indicates that

prior to controlling for firm characteristics, performance, or CEO characteristics, a one-standard

deviation increase in extraversion is associated a statistically significant 19% reduction in the

likelihood of CEO turnover (relative to a 5% unconditional probability). Controlling for firm

characteristics and performance results in a slightly larger 22% decline. Further controlling for

the seventeen CEO characteristics reduces the estimate to still meaningful 12%, although the

decline is only marginal significant (p = 0.08). We conjecture that extraversion may be

particularly beneficial in the midst of poor performance, and we explore this idea by splitting the

sample into two groups based on the median breakpoint of prior fiscal year returns. We find that

extraversion is associated with an insignificant 4% decline in turnover among firms in the top

29

half of returns over the prior year, but a significant 20% decline in turnover among firms in the

bottom half of performance (results are tabulated in Table IA.5 in the internet appendix).

Overall, the above evidence suggests that extraversion is associated with a lower

likelihood of turnover and reduced turnover-performance sensitivity. Reduced turnover should

naturally translate into longer tenures. To test this prediction, we estimate:

𝐿𝑜𝑔 (𝑇𝑒𝑛𝑢𝑟𝑒)𝑖𝑡 = 𝛽1𝐸𝑥𝑡𝑟𝑎𝑣𝑒𝑟𝑠𝑖𝑜𝑛𝑖 + 𝛾𝑭𝒊𝒓𝒎𝑪𝒉𝒂𝒓 + 𝛿𝑷𝒆𝒓𝒇𝒐𝒓𝒎𝒂𝒏𝒄𝒆 + ωCEOChar + 𝑌𝑒𝑎𝑟𝑡 + 𝐹𝐸𝑖 + 𝜀𝑖𝑡 , (6)

where all variables are defined as in Equation 3 except we exclude Tenure from the CEOChar

vector.

We find that a one-standard deviation increase in extraversion is associated with an

increase of tenure of 5.15% to 7.68%. Longer tenures have economically meaningful

consequences for CEO pay. In particular, since the average tenure of a departing CEO is 8 years

(96 months), the 5.72% estimate (in Specification 6) implies that extraverted CEOs stay in office

for roughly 5.5 extra months.26 Assuming the annual incremental compensation premium from

being CEO is $3.5 million (the difference between the average CEO and CFO salary), then the

extra compensation associated with extraverted CEOs longer tenure is roughly $1.6 million ($3.5

× 5.5/12).

4.6 Extraversion and Outside Directorships

Extraverts’ tendency to attract social attention (e.g. Ashton, Lee, and Paunonen, 2002)

may lead to greater external visibility and more invitations to sit on outside boards. Extraverts’

outgoing nature may also make them more willing to accept board invitations. We therefore test

whether extraverted executives sit on more boards by estimating the following regression:

26 This estimate is line with Mayew, Parsons, and Venkatachalam (2013) who find that an interquartile increase in voice pitch is associated with an increase in tenure of roughly five months.

30

𝑌𝑖𝑡 = 𝛽1𝐸𝑥𝑡𝑟𝑎𝑣𝑒𝑟𝑠𝑖𝑜𝑛𝑖 + 𝛾𝑭𝒊𝒓𝒎𝑪𝒉𝒂𝒓 + 𝛿𝑷𝒆𝒓𝒇𝒐𝒓𝒎𝒂𝒏𝒄𝒆 + 𝜔𝑪𝑬𝑶𝑪𝒉𝒂𝒓 + 𝑌𝑒𝑎𝑟𝑡 + 𝐹𝐸𝑖 + 𝜀𝑖𝑡. (7)

We estimate two versions of Equation 7. In the first approach, we employ a logit

regression and Y is a dummy variable that is equal to one if the executive sits on any outside

boards and zero otherwise. In the second approach, Y is defined as the natural log of one plus the

number of outside directorships. All control variables are defined as in equation 3. The

independent variables are defined in Appendix A and all continuous variables are standardized to

have mean 0 and variance 1. Standard errors are clustered by executive.

The results are reported in Table 7. Of the 9,647 CEO-year observations, 3,340 involve

the CEO setting on an outside board, for an unconditional probability of 34.7%. The logit

approach in Specification 2 indicates that a one standard deviation increase in CEO extraversion

is associated with a 16% increase in the likelihood of being on any outside board. We also

observe a significant relation between extraversion and log of the number of directorships in

Specification 1.

We next examine whether extraverted CEOs receive higher quality outside directorships.

We consider firm size, as larger firms likely afford directors greater visibility and prestige

(Adams and Ferreira, 2008), higher compensation (Ryan and Wiggins, 2004), and an increased

likelihood of obtaining additional directorships (Yermack, 2004). Further, Masulis and Mobbs

(2014) find evidence that independent directors spend more time and effort in monitoring their

larger directorships, which is consistent with larger directorships being perceived as more

valuable.

We examine the relation between the quality of directorships and extraversion by setting

Y in Equation 7 equal to Ln(Sales), Ln(Assets), or Ln(Market Equity) of the executive’s largest

outside directorship, and we require CEOs to hold at least one outside directorship position. The

31

results, reported in Specifications 3 through 5 of Table 7, are consistent with extraverted CEOs

sitting on larger (i.e., higher quality) outside boards. In particular, a one-standard deviation

increase in extraversion is associated with sitting on outside boards that are 11% to 22% larger.

4.7 Extraversion and CFO Promotion to CEO

In competitive sorting theories of firm hierarchies, individuals with the greatest perceived

ability are assigned to the most important job (i.e., the CEO), where their perceived differential

ability is likely to have the largest impact. Thus, if extraversion is linked to perceived managerial

ability, then extraverted CFOs should be more likely to be promoted to the top job following the

departure of the CEO.27 To test this conjecture, we estimate the following regression:

𝑃𝑟𝑜𝑚𝑜𝑡𝑖𝑜𝑛𝑖𝑡 = 𝛽1𝐸𝑥𝑡𝑟𝑎𝑣𝑒𝑟𝑠𝑖𝑜𝑛𝑖 + 𝛾𝑭𝒊𝒓𝒎𝑪𝒉𝒂𝒓 + 𝛿𝑷𝒆𝒓𝒇𝒐𝒓𝒎𝒂𝒏𝒄𝒆 + ωCFOChar+ 𝜁𝑪𝒖𝒎𝑷𝒆𝒓𝒇𝒐𝒓𝒎𝒂𝒏𝒄𝒆 +𝑌𝑒𝑎𝑟𝑡 + 𝐹𝐸𝑖 + 𝜀𝑖𝑡 , (8)

Promotion is a dummy variable equal to one if the internal CFO is subsequently promoted to

CEO following a CEO departure, FirmChar are defined as in equation 3, and CFOChar is a

vector that includes all the CFO characteristics reported in Specification 5 of Table 2, plus the

other Big 4 personality traits. We also expect that the promotion to CEO will depend not only on

performance metrics over the prior year, but also over the CFO’s entire tenure with the firm prior

to the CEOs departure. 28 Accordingly, CumPerformance is a vector of cumulative performance

measures that include: the average returns over the CFO’s tenure with the firm (Cumulative

Return), the average Percent CFO Text, and the average Relative Forecast Error (as defined in

Hutton and Stocken, 2009). Li et al. (2014) argue that Percent Text proxies for firm-specific

knowledge, and Goodman et al. (2014) find higher quality earnings forecasts are associated with

27 The broader prediction is that among all inside executives, those with higher extraversion are more likely to take on the role of CEO. By focusing on CFOs, we implicitly assume that CFOs with high extraversion scores are likely to be more extraverted than other inside executives. 28 We measure the CFO’s tenure with the firm based on the first year the CFO appears in Execucomp for the firm. Limiting the CFO’s tenure to only years where the executive held the title of CFO leads to very similar results.

32

better investment decisions. Finally, because compensation may be an effective way of

summarizing unobservable performance measures, we also include the average CFO Pay Slice,

defined as CFO’s compensation scaled by the compensation of the top three executives over the

executive’s tenure.

Specification 1 of Table 8 reports the odds ratios from a logistic regression prior to

including the cumulative performance measures. The results indicate that a one standard

deviation increase in CFO extraversion is associated with a 52% percent increase in the

likelihood of being promoted to CEO (relative to an 8% unconditional probability). Specification

2 adds Cumulative Returns, Relative Forecast Errors, and Percent CEO Text. None of the

cumulative performance measures are statistically significant, although their inclusion does

reduce the coefficient on Extraversion to 41%. Specification 3 includes CFO Pay Slice, a strong

predictor of promotion to CEO, which further reduces the coefficient on extraversion to 31%.

The evidence that the relation between CFO extraversion and promotion to CEO is attenuated

when including cumulative performance measures is consistent with the view that previous

career successes mediate the relation between executive extraversion and superior labor market

outcomes.

We conjecture that boards may place greater emphasis on extraversion for CFOs with

little or no experience with the firm, since other signals of skill such as performance measures

are less reliable or unavailable. We test this hypothesis by grouping the samples into CFOs with

tenures below and above the sample median of four years. We find that a one-standard deviation

increase in extraversion is associated with a highly significant 161% increase in the likelihood of

being promoted to CEO for CFOs with below median tenure, compared to a statistically

insignificant 23% increase for CFOs with above median tenure. We also find that boards place

33

significantly more weight on cumulative performance metrics as tenure increase. Both

Cumulative Return and CFO Pay Slice are statistically insignificant for the low tenure sample,

but very strong predictors of promotion for the high tenure sample.

The evidence in Section 3.2 and Table IA.2 indicates that firms tend to have persistent

preferences for extraverted CEOs. Accordingly, we may expect that extraversion may be

strongly related to CEO promotion when the departing CEO is more extraverted. In

Specifications 6 and 7 of Table 8, we split the sample of CEOs into those with below and above

the median extraversion.29 Consistent with our conjecture, we find that a one standard deviation

increase in extraversion is associated with a statistically significant 90% increase in the

likelihood of being promoted to CEO following the departure of a relatively extraverted CEO,

compared to an insignificant 3% increase following the departure of a relatively introverted

CEO.

5. Executive Extraversion and Firm Outcomes

The results from the previous section indicate that extraverted executives experience a

number of career benefits including higher compensation, longer tenures, more outside

directorships, and a greater likelihood of promotion of CEO. These findings are consistent with a

long literature in psychology, originating with Mann (1959), documenting that people perceive

extraverts as superior leaders. While the psychology literature has robustly documented that

people perceive extraversion as an important leadership quality, the literature is mixed on

whether extraversion is related to performance. For example, Stogdill (1974) and Bentz (1985)

show that extraverted leaders receive high ratings of effectiveness from both peers and superiors,

29 We exclude 339 CEO departures with missing CEO extraversion scores.

34

whereas recent work by Bendersky and Shah (2013) finds evidence that extraverts underperform

expectations.

In an efficient labor market, we would expect optimal matching between managers and

firms, and as a result there may be no observed relation between CEO characteristics and

performance (e.g., Custódio, Ferreira, and Matos, 2013). Relatedly, firms may optimize over a

wide range of CEO traits, in which case firms that may benefit from an extraverted CEO could

nevertheless hire an introvert who has other valuable traits. However, such firms could expend

resources in other ways to synthetically obtain the benefits associated with CEO extraversion

(such as increased investments in investor relations, etc.).

On the other hand, if extraverts’ incremental pay reflects skill that they are not able to

fully extract through salary, then we could observe a positive relation between extraversion and

firm performance. However, even in this setting, detecting a relation between extraversion and

performance is challenging. For example, Gabaix and Landier (2008) calibrate a CEO talent

model and argue that small, hard to empirically detect effects on firm value can justify

economically large differences in compensation. The selection of CEOs is also endogenous. For

example, if firms that anticipate poor performance seek out extraverted executives, this will bias

downward the effects of extraversion on performance. With these caveats in mind, in this section

we explore the relation between extraversion and several firm outcomes.

5.1 Changes in Investor Recognition and Performance around CEO Transitions

We begin by examining changes in firm outcomes around CEO transitions. We require

that the incoming CEO remains in office for at least two years, and we eliminate transitions that

coincide with major corporate restructurings (e.g., spinoffs and mergers) and interim CEOs.

Using the resulting sample of 619 CEO transitions, we estimate the following regression:

35

𝛥𝑌𝑖𝑡+3,𝑡−1 = 𝛽1𝛥𝐸𝑥𝑡𝑟𝑎𝑣𝑒𝑟𝑠𝑖𝑜𝑛𝑖𝑡+3,𝑡−1 + 𝜷2𝛥𝑪𝑬𝑶𝑪𝒉𝒂𝒓𝑖𝑡+3,𝑡−1 + 𝜀𝑖𝑡 . (9)

Y denotes one of several firm outcome measures, and the dependent variable examines the

difference between the average level of Y in the three years after CEO transition (years t+1 to

t+3) relative to the value of Y in the year prior to the transition (year t-1, we exclude the year of

the transition). Following Perez-Gonzalez (2006), we adjust each firm outcome measure by

subtracting the median Y from a control group of firms that are matched by industry and Y. The

control group of firms consists of firms in the same Fama-French 12 industry group and the same

industry-adjusted quintile ranking of Y in the year prior to the executive transition.

ΔExtraversion is the difference in the extraversion between the incoming and departing

CEOs, and ΔCEOChar is a vector that includes the changes in all CEO characteristics in

Equation 3.30 If data is unavailable for either the incoming (departing) CEO, we set the value of

their extraversion or other CEO characteristics equal to sample mean, and we include a missing

incoming (departing) CEO dummy for each of the specific missing variables.31 All variables are

defined in Appendix A. Continuous independent variables are standardized to have mean zero

and variance equal to one, and outcome measures are winsorized at the 1st and 99th percentile.

Standard errors are clustered by firm.

Before turning to the firm outcome results, we first confirm that extraverts command a

compensation premium when focusing on CEO transitions. Specifically, we estimate Equation

(9) with Y set equal to the log of total compensation. The Extraversion coefficients are reported

in Panel A of Table 9. In a univariate setting in Specification 1, we find that incoming CEOs that

30 By focusing on CEO transitions, we examine within firm variation in firm outcomes (relative to similar firms in its industry) around a short window. We therefore exclude firm characteristics such as growth or size from the set of controls since these may endogenously reflect manager ability. 31 This approach allows us to significantly expand our sample by including observations where we have extraversion scores for only one of the two transitioning CEOs. Excluding these observations results in very similar point estimates and slightly reduced statistical significance.

36

have extraversion scores that are one standard deviation higher than the departing CEO receive

compensation that is 6.78% higher (t-stat=2.90). The result is robust to including controls for

(changes in) CEO characteristics in Specification 2, with a coefficient estimate of 7.18% (t-

stat=2.75).

Extraverts tend to attract social attention (e.g., Ashton, Lee, and Paunonen, 2002), which

could lead to greater external visibility for the firm. We begin the firm outcome analysis by

exploring measures of investor recognition. We consider three broad measures: brokerage

analyst attention, media attention, and stock liquidity. Analyst attention is measured using the

number of analysts covering the firm (Analyst Coverage) and the number of times the firm

presents at a broker-hosted investor conference (Conference Presentations) during the calendar

year. Media attention is measured using the number of articles (Media Articles) or the number of

words (Media Words) across all articles that mention the firm in the Wall Street Journal during

the calendar year. Finally, we proxy for firm liquidity using share turnover (Turnover) and the

Amihud (2002) illiquidity ratio (Illiquidity). All investor recognition variables are measures in

natural logs.

Panel B of Table 9 presents the results. After controlling for CEO characteristics in

Specification 2, we observe that the appointment of a CEO that has an extraversion score that is

one standard deviation greater than the departing CEO is associated with a 4.4% increase in

Analyst Coverage, a 5.2% increase in Conference Presentations, a 5.1% increase in Media

Articles, a 24.9% increase in Media Words, a 5.1% increase in Turnover, and a 14.4% decline in

Illiquidity. With the exception of the Media variables, all of the estimates are statistically

significant at a 5% level. The findings are consistent with the view that the appointment of an

extraverted executive is associated with improvement in investor recognition. As discussed

37

earlier, however, the positive association may reflect the endogenous matching of firms and

CEOs. For example, firms that are interested in improving investor recognition or anticipate

greater attention from analysts or the media may place a greater emphasis on hiring an

extraverted CEO. However, this interpretation is also consistent with the view that extraverts

may be valuable for firms with heightened visibility.32

We next examine the relation between extraversion and measures of firm performance

around CEO transitions. In particular, we set Y in Equation 9 equal to one of eight different

performance measures: sales growth, market share, firm efficiency (as defined in Demerjian,

Lev, and Mcvay, 2012)33, profit margin, profitability (i.e. scaled OCF), return on assets (ROA),

Tobin’s Q, and returns. All variables are defined in Appendix A.

Panel C of Table 9 reports the performance results. The evidence indicates that firms that

replace the departing CEO with a more extraverted incoming CEO experience a significant

increase in sales growth (2.5%), market share (4.2%), and firm efficiency (1.4%), and all

coefficients are statistically significant at the 5% level. Extraversion is positively related to

profitability, yet the relation is only marginally significant at the 10% level. Extraversion shows

a positive relation with the other performance measures, although none of the remaining

estimates are statistically significant.

5.2 CEO Extraversion and Departure Returns

The positive association between CEO extraversion and firm outcomes is consistent with

extraverted CEOs adding value to their firms. We examine whether the market shares this view

32 The evidence suggests that controlling for measures of investor recognition may also mediate some of the extraversion pay premium. Supporting this view, augmenting Specification 4 of Table 4 to include the investor recognition measures above reduces the extraversion pay premium from 5.70% to 4.50%, although the effect remains significant (t=3.01). 33 They use a Data Envelope Analysis approach to measure the efficiency by which managers convert assets, investments, and expenses, into revenues.

38

by studying the stock price response to CEO departure announcements. We focus on departures

since the market is likely better at assessing the contributions of the departing CEO relative to

incoming CEOs, about whom relatively less information may be available.

We hand collect data on CEO departure announcements for all firms in Execucomp

during our sample period (2004-2013). Specifically, we focus on the sample of departures in

which we can obtain a valid extraversion score for the departing CEO, and we search company

press releases or other news sources (through Factiva) to identify the earliest reported departure

dates. We also collect additional information about the circumstances surrounding each

succession. We exclude departures that are directly related to major restructurings (e.g., mergers,

spinoffs, going private, etc.) The final sample includes 757 CEO departures.

We examine the relation between extraversion and departure returns by estimating the

following regression

𝐶𝐴𝑅𝑖𝑡 = 𝛽1𝐸𝑥𝑡𝑟𝑎𝑣𝑒𝑟𝑠𝑖𝑜𝑛𝑖𝑡 + 𝜷𝟐𝑭𝒊𝒓𝒎𝑪𝒉𝒂𝒓𝑖𝑡 + 𝜷𝟑𝑪𝑬𝑶𝑪𝒉𝒂𝒓𝑖𝑡 + 𝐼𝑛𝑑𝑖 + 𝑌𝑒𝑎𝑟𝑡 + 𝜀𝑖𝑡 . (8)

CAR is the three-day market-adjusted returns of the acquiring firm centered around the departure

date. Extraversion is the extraversion score of the departing CEO. FirmChar and CEOChar are

vectors of the firm and CEO characteristics included as controls in Equation 3.

The results are reported in Table 10. For the set of all departures, a one standard deviation

increase in extraversion is associated with a -0.21% lower announcement return, although the

estimate is not statistically significant. We conjecture that extraversion may play a larger role for

voluntary departures, and we identify and remove forced departures using the methodology of

Parrino (1997).34 For the sample of 531 voluntary departures, the effect of extraversion on

34 Departures are classified as forced if media reports indicate that the CEO is fired, forced from the position, or departs due to unspecified policy differences. Departures are also deemed forced if the departing CEO is under the

39

announcement returns increases to more than 50 basis points and becomes statistically significant

in Specifications 4 and 5, although Extraversion loses significance after including the full set of

eighteen CEO characteristics as controls in Specification 6.

CEO departures may be anticipated by the market, which should significantly attenuate

the market reaction to the announcement. We explore whether the effect of extraversion is

stronger when the departure announcement is more likely to covey new information to the

market by identifying “unexpected” CEO departures. The search of departures includes three

sudden deaths (as defined in Nguyen and Nielsen, 2014) and 12 other departures that are

explicitly described in the press release as “unexpected,” “unanticipated,” or “surprising.” In a

univariate setting, we find that the departure of more extraverted CEOs are associated with

significantly lower announcement returns.35

5.3 Acquisition Announcement Returns

In our final tests, we examine whether the market reaction to firm investment decisions

varies with CEO extraversion. One area in which extraverts may be particularly valuable is

mergers and acquisitions (M&A). For example, Sitkin and Pablo (2005) note that M&A can act

as a “revealing litmus test that highlights the quality of leadership,” and Waldman and Javidan

(2009) argue that charismatic leadership qualities are particularly important during post-merger

integrations.

We collect M&A data from the Thomson Financial SDC Platinum database. Following

Custódio and Metzger (2013), we include only transactions in which control is transferred (i.e.,

age of 60 and media reports do not mention death, poor health, or the acceptance of another position, or reports that the CEO is retiring, but does not announce the retirement at least six months prior to the succession. 35 Due to the small sample size we are not able to determine whether this effect is robust to the inclusion of all other controls. However, in unreported analyses, we find that the coefficient on extraversion remains significant after including any single firm or CEO characteristic.

40

the acquirer has less than a 50% stake in the target before the transaction, and greater than 50%

after the transaction), and we require that the transaction value of the merger is at least $50M.

We also limit the sample to observations for which the acquiring CEO has a non-missing

extraversion score. The final sample includes 1,503 acquisitions. We examine the relation

between extraversion and announcement returns by estimating the following regression:

𝐶𝐴𝑅𝑖𝑡 = 𝛽1𝐸𝑥𝑡𝑟𝑎𝑣𝑒𝑟𝑠𝑖𝑜𝑛𝑖𝑡+3,𝑡−1 + 𝜷𝟐𝑫𝒆𝒂𝒍𝑪𝒉𝒂𝒓𝑖𝑡 + 𝜷𝟑𝑭𝒊𝒓𝒎𝑪𝒉𝒂𝒓𝑖𝑡 + 𝜷𝟒𝑪𝑬𝑶𝑪𝒉𝒂𝒓𝑖𝑡 + 𝐼𝑛𝑑𝑖 + 𝑌𝑒𝑎𝑟𝑡 + 𝜀𝑖𝑡 . (10)

CAR is the three-day market-adjusted returns of the acquiring firm centered around the

announcement date of the acquisition. Extraversion is the extraversion score of the CEO of the

acquiring firm. DealChar is a vector of deal characteristics that are known to influence

announcement returns, FirmChar and CEOChar are vectors of the firm and CEO characteristics

included as controls in Equation 3 (details are in Appendix A).

Table 11 reports the results. In a univariate setting (with industry and year fixed effects),

a one-standard deviation increase in CEO extraversion is associated with a 0.22% increase in 3-

day CARs around the acquisitions, although the estimate is not statistically significant.

Controlling for deal, firm, and CEO characteristics increases the estimate to a statistically

significant 0.38%. The evidence supports the view that extraverted CEOs add value in the M&A

process.36

Overall, while the analysis of CEO extraversion on firm outcomes is subject to important

caveats, two main findings emerge. First, we find no evidence to suggest that extraverts are

associated with worse firm outcomes, which helps rule out that extraverts charismatically attract

board attention but subsequently deliver disappointing performance (e.g., Khurana, 2002 and

36 This finding points to the possibility that extraverted executives may also be more likely to make an acquisition. In Table IA.6 in the Internet Appendix, we find evidence that extraverted executives are significantly more likely to engage in M&A activity.

41

Malmendier and Tate, 2009). In contrast, we do find evidence that CEOs are associated with

superior firm outcomes in some facets of the job, including improvements in investor recognition

and acquisitions that are better received by the market. The negative announcement response to

voluntary departures of extraverted CEOs corroborate this view. While the positive associations

do not necessarily imply a causal relation, they do point towards possible channels that may help

explain the extraversion pay premium.

6. Conclusion

Relatively little is known about which executive traits are viewed as important to boards

in their hiring of top executives. We explore the role of an important individual characteristic,

personality extraversion, on career outcomes. We use linguistic algorithms to measure executive

extraversion based on speech patterns during conference calls. Our call-based measure is highly

persistent at the executive level and considerably less so at the firm level, which is consistent

with psychology literature that indicates extraversion is a persistent personality trait.

We find compelling evidence that executive extraversion is associated with improved

career outcomes. After controlling for firm characteristics as well as including controls for

manager education and experience, we find that extraverted CEOs receive 5.70%-7.57% higher

compensation. Extraversion also affects career trajectory. Extraverts become top executives at a

younger age, are less likely to experience job turnover, serve on more outside boards, hold

directorships at larger firms, and extraverted CFOs are more likely to be promoted to CEO.

Collectively, our findings highlight convincing salary and other labor market benefits to

extraversion. Although we control for a host of managerial education and experience variables in

our analysis, it is not possible to control for all potentially relevant intermediate effects. Thus, a

cautious interpretation of the observed relation between extraversion and executive labor market

42

outcomes is that it reflects the direct effect of extraversion as well as any indirect effects of

earlier, unobserved experiences or successes. Nevertheless, our analysis makes an important step

forward in understanding which managerial traits are associated with career success.

We also examine the implications of CEO extraversion on firm outcomes. Focusing on

manager transitions, we find that hiring an extraverted CEO is associated with improvements in

investor recognition, sales growth, and firm efficiency. Extraverted CEOs are also associated

with higher acquisition returns. While the positive associations between extraversion and firm

outcomes do not necessarily imply a causal relation, the findings point towards possible channels

that may help explain the extraversion pay premium, thereby providing support for a rational

market-based explanation for the improved labor market outcomes of extraverted CEOs.

43

Appendix A: Variable Definitions:

A.1 Measures of Executive Extraversion

x Call Extraversion – the extraversion score of an executive based on his/her speech during the questions-and-answers portion of a conference call. For each call, the extraversion score is computed using the average of four linguistic algorithms described in Section IA.1 of the Interenet Appendix. The extraversion score for each linguistic algorithm is winsorized at the 1st and 99th percentile. (Source: Thomson Reuters and Seeking Alpha).

x Aggregate Extraversion – A weighted average measure of Call Extraversion, where each call is weighted by the number of words spoken in the Q&A portion of the call by the executive.

x Extraversion – the weighted average residuals from the following panel regression:

Call Extraversion = β1Retit-63,t-2 + β2Retit-1,t+1 + β3Retit+2,t+63 + β4Earnings Call + β5MBE + Β6 Surprise + β7Loss + Qtr + ε. The regression is estimated separately for CEOs and CFOs (i.e., Specifications 1 and 4 of Table 2). The residuals from each conference call are weighted by the number of words spoken by the executive during the questions-and-answers portion of the conference calls. Executives who speak on fewer than 3 conference calls are dropped from the sample.

A.2 Dependent Variables

x Total Comp. – total compensation over the fiscal year, comprised of the following components: salary, bonus, total value of restricted stock granted, total value of stock options granted (estimated using Black-Scholes), long-term incentive payouts, and other compensation (i.e., TDC1). (Source: Execucomp).

x Cash Comp. – Salary + Bonus (Source: Execucomp). x Equity Comp. – Total Comp. – Cash Comp. (Source: Execucomp). x First Age – the age at which the executive was first appointed to CEO of an Execucomp firm,

estimated using the Became CEO variable (Source; Execucomp). x Tenure – the number of years the executive has held the same position at the firm (Source:

Execucomp) x Turnover – a dummy variable equal to one if the executive is replaced in a year, and 0

otherwise (Source: Execucomp). x Promotion – A dummy variable equal to one if the internal CFO was promoted to CEO

following the departure of the CEO. (Source: Execucomp). x Directorships – the number of outside boards held by the executive during the calendar year.

(Source: ISS/RiskMetrics). x Directorship Size – the size of the executive’s largest outside directorship. Size is measured

using either sales, total assets, or market equity. (Source: ISS/RiskMetrics). x Profit Margin – Net income divided by sales, winsorized at the 1st and 99th percentile.

(Compustat)

44

o Ind_Adj_Profit Margin – Profit Margin less the median Profit Margin of a control group of firms that are in the same Fama and French (1997) 12 industry classification and are in the same Profit Margin quintile in the year prior to the executive transition.

x Operating Cash Flow (Prof) – Annual cash flows from operations scaled by assets as of the end of the prior fiscal year, winsorized at the 1st and 99th percentile. (Source: Compustat) o Ind_Adj_Prof – Prof less the median Prof of a control group of firms that are in the

same Fama and French (1997) 12 industry classification and are in the same Prof quintile in the year prior to the executive transition.

x ROA - EBITDA scaled by assets as of the end of the prior fiscal year, winsorized at the 1st and 99th percentile. (Source: COMPUSTAT). o Ind_Adj_ROA –ROA less the median ROA of a control group of firms that are in the

same Fama and French (1997) 12 industry classification and are in the same ROA quintile in the year prior to the executive transition.

x Q – (total assets + market value of equity – book value of equity)/total assets. We drop negative values and winsorize at the 99th percentile. (Source: COMPUSTAT). o Ind _Adj Q – Q less the median Q of a control group of firms that are in the same Fama

and French (1997) 12 industry classification and are in the same Q quintile in the year prior to the executive transition.

x Market Share – the percentage of revenues earned by the firm within its Fama and French (1997) 49 industry classification. (Source: COMPUSTAT). o Ind_Adj Market Share – Market Share less the median Market Share of a control group

of firms that are in the same Fama and French (1997) 12 industry classification and are in the same Market Share quintile in the year prior to the executive transition.

x Firm Efficiency – a measure of how efficient a firm is in generating revenue for a given set of inputs, as described in greater detail in Demerjian, Lev, and McVay (2012). (Source: http://faculty.washington.edu/pdemerj/data.html). o Ind_Adj_Firm Efficiency – Firm Efficiency less the median Firm Efficiency of a control

group of firms that are in the same Fama and French (1997) 12 industry classification and are in the same Firm Efficiency quintile in the year prior to the executive transition.

A.3 Partitioning Variables

x External Hire – a dummy variable equal to one if the incoming CEO was hired from another firm.

x Internal Hire – a dummy variable equal to one if the incoming CEO was hired from within the same firm.

x Voluntary Departure – a dummy variable equal to 1 if the CEO departure was voluntary, as defined in Parrino (1997). Specifically: o If a news article reports that the CEO is fired, forced from the position, or departs due to

unspecified policy difference then Voluntary Departure = 0. o If the departing CEO is under 60 and 1) the departure is not explicitly attributed to

death, poor health, or accepting another position then Voluntary Departure = 0.

45

o If the departing CEO is under 60 and the article reports the CEO is retiring, but does not announcement the retirement at least 6 months before the succession, then Voluntary Departure = 0.

o All other cases are classified as voluntary (i.e., Voluntary Departure =1). x Unexpected Departure – this includes departures due to sudden deaths defined as either heart

attacks, strokes, accidents, or any death of natural causes described as “sudden” or “unexpected”. The sample also includes departures that are described in new articles as “sudden”, “unexpected”, or “surprising”.

A.3 Control Variables

x Return – the return on the stock less the value-weighted market return. (Source: CRSP). o FRett – the Return over fiscal year t. o Rett – the Return over calendar year t.

x Earnings Call – a dummy variable equal to one if the conference call occurred around the 4-day window [-1, 2] around the earnings announcement (day 0).

x Meet-or-Beat – a dummy variable equal to one if earnings meet or beat the consensus analyst forecast for the most recent quarter. This variable is set to zero for all conference calls that are not earnings calls. (Source: IBES).

x Surprise – the most recent earnings surprise, measured as the difference between quarterly EPS and the mean consensus analyst forecast scaled by the stock price at the beginning of the quarter. This variable is winsorized at the 1st and 99th percentile and set to zero for all conference calls that are not earnings calls. (Source: IBES).

x Loss – a dummy variable equal to one for firms reporting negative earnings in the most recent quarter. This variable is set to zero for all conference calls that are not earnings calls. (Source: IBES).

x Exec Age – the age of the executive. (Source: Execucomp). x Male – a dummy variable equal to one if the executive is a male. (Source: Execucomp). x Optimism – the total number of positive words spoken by an executive during the a Q&A

section of the conference call scaled by the sum of the total number of both positive and negative words [i.e., Positive Words/(Positive + Negative Words)]. The list of the positive and negative words is taken from the Loughran and McDonald (2011) dictionary.

x Overconfidence – a measure of an executive’s tendency to hold in-the-money stock options as defined in Campbell et al. (2011). (Source: Execucomp).

x Founder – a dummy variable equal to one if the year the current executive first became CEO (as reported in Execucomp) is within one year of when the firm went public (as reported in CRSP).

x General Ability Index (GAI) – a measure of general managerial ability as defined in Custódio, Ferreira, and Matos (2013). Specifically, GAI = 0.268X1 +0.312X2 +0.309X3

+0.218X4 +0.153X5, where: o X1 = number of different positions that a CEO performed during his career o X2 = number of firms where a CEO worked

46

o X3 = number of industries at the four-digit SIC level where a CEO worked. o X4 = a dummy variable equal to one if a CEO held a CEO position at another firm o X5 = a dummy variable equal to one if the CEO worked for a multi-division firm

(Source: Boardex).

x Rolodex – the sum of other external executives or directors related to the CEO through past professional connections, social connections, and past universities attended, as defined in Engelberg, Parsons, and Gao (2013). (Source: Boardex).

x MBA – a dummy variable equal to one if the executive has an MBA. (Source: Boardex). x Doctorate – a dummy variable equal to one if the executive has a PhD. (Source: Boardex). x GradHonors – a dummy variable equal to one if the an executive graduated with distinction,

honors, summa cum laude, magna cum laude, or cum laude for any degree. (Source: Boardex).

x Ivy League – a dummy variable equal to one if the executive graduated from an ivy league university for any degree. (Source: Boardex).

x CEO Percent Text – The ratio of the number of words spoken by the CEO during the conference call to the number of words spoken by all company executives during the conference call

x Sales – total sales. (Source: COMPUSTAT). x Vol – the standard deviation of daily returns over the past 60 months. (Source: CRSP). x Firm Age – the total number of months since the firm first appeared in CRSP. x Emotional Stability (Emo_Stab) – the weighted average call-level measure of emotional

stability. The calls are weighted by the number of words spoken by the executive during the Q&A portion of the conference call. The call-level measure is computed using the average of four linguistic algorithms described in Appendix B. The extraversion score for each linguistic algorithm is winsorized at the 1st and 99th percentile. (Source: Thomson Reuters and Seeking Alpha).

x Openness (Open) – the weighted average call-level measure of openness. The calls are weighted by the number of words spoken by the executive during the Q&A the conference call. The call-level measure is computed using the average of four linguistic algorithms described in Appendix B. The extraversion score for each linguistic algorithm is winsorized at the 1st and 99th percentile. (Source: Thomson Reuters and Seeking Alpha).

x Agreeableness (Agree) – the weighted average call-level measure of agreeableness. The calls are weighted by the number of words spoken by the executive during the Q&A portion of the conference call. The call-level measure is computed using the average of four linguistic algorithms described in Appendix B. The extraversion score for each linguistic algorithm is winsorized at the 1st and 99th percentile. (Source: Thomson Reuters and Seeking Alpha).

x Conscientiousness (Consc) – the weighted average call-level measure of conscientiousness. The calls are weighted by the number of words spoken by the executive during the QA portion of the conference call. The call-level measure is computed using the average of four linguistic models described in Appendix B. The extraversion score for each linguistic algorithm is winsorized at the 1st and 99th percentile. (Source: Thomson Reuters and Seeking Alpha).

47

x CFO Pay Slice – the total compensation of the CFO scaled by the total compensation of the three highest paid executives at the firm. (Source: Execucomp).

x Assets – total assets. (Source: COMPUSTAT). x R&D/Assets – research and development expenses scaled by assets at the end of the prior

year, winsorized at the 99th percentile. We set missing values of R&D to zero and include an indicator variable that equals one where there is a missing value and zero otherwise. (Source: COMPUSTAT).

x Cumulative Returns – the average annual return over the executive’s entire tenure with the firm.

x Relative Forecast Error – the absolute forecast error of the management forecast relative to the absolute forecast error of the prevailing consensus analyst forecast. Absolute forecast errors are computed as the absolute difference between realized earnings and forecasted earnings, scaled by price at the end of the prior quarter.

x Guidance Dummy – a dummy variable equal to one if the executive ever issued earnings guidance.

x Tender – a dummy variable equal to 1 for tender offer acquisitions. x Equity Finance – a dummy equal to 1 if the merger is 100% paid with equity. x Mixed Finance – a dummy equal to one if the merger is financed with a mix of cash and

equity. x Cash Finance – a dummy equal to 1 if the merger is 100% paid with cash. x Public Target – a dummy equal to 1 if the target is a public company x Private Target – a dummy equal to 1 if the target is a private company x Subsidiary – a dummy equal to 1 if the target is a subsidiary of the company.

48

References Adams, R., Ferreira, D., 2008. Do directors perform for pay? Journal of Accounting and

Economics 46, 154–171. Adams, R., Keloharju, M., Knupfer S., 2014. Match made at birth? What traits of a millions

Swedes tell us about CEOs. Working paper, University of New South Wales. Amihud, Y., 2002. Illiquidity and stock returns: Cross-section and time-series effects. Journal of

Financial Markets 5, 31-56. Argamon, S., Dhawle, S., Koppel, M., Pennebaker, J., 2005. Lexical predictors of personality

type. In Proceedings of the Joint Annual Meeting of the Interface and the Classification Society of North America.

Ashton, M., Lee, K., Paunonen, S., 2002. What is the central feature of extraversion? Social

attention versus reward sensitivity. Journal of personality and social psychology 83, 245-252.

Bendersky, C., Shah, N.P., 2013. The downfall of extraverts and rise of neurotics: The dynamic

process of status allocation in task groups. Academy of Management Journal 56, 387-406. Benmelech, E., Frydman, C., 2015. Military CEOs. Journal of Financial Economics 117, 43-59. Bentz, V.J. (1985). Research findings from personality assessment of executives. In J.H.

Bernardin & D.A. Bownas (Eds.), Personality assessment in organizations (pp. 82–144). New York: Praeger.

Bertrand, M., Schoar, A., 2003. Managing with style: the effect of managers on corporate policy.

Quarterly Journal of Economics 118, 1169–1208. Bono, J., and Judge, T., 2004, Personality and transformational and transactional leadership: A

meta-analysis. Journal of Applied Psychology 89, 901-910. Booth, J., Deli, D., 1996. Factors affecting the number of outsider directorships held by CEOs.

Journal of Financial Economics 40, 81-104. Borghans, L., Duckworth, A., Heckman, J. and ter Weel, B., 2008. The economics and

psychology of personality traits. The Journal of Human Resources, 43: 972-1059 Breiman, L., Friedman, J., Stone, C. J., Olshen, R. A., 1984. Classification and regression trees.

CRC press. Bushee, B., Jung, M., and Miller, G., 2011. Conference presentations and the disclosure milieu.

Journal of Accounting Research 49, 1163-1192

49

Cain, S., 2012. Quiet: The power of introverts in a world that can’t stop talking. Crown Publishers, New York.

Campbell, T. C., Gallmeyer, M., Johnson, S. A., Rutherford, J., Stanley, B W., 2011. CEO

optimism and forced turnover. Journal of Financial Economics 101, 695-712. Case, A., Paxson, C., 2008. Stature and status: Height, ability and labor market outcomes.

Journal of Political Economy 116, 499-532. Coltheart, M., 1981. The MRC psycholinguistic database. Quarterly Journal of Experimental

Psychology 33, 497–505. Cook, D.O., Mobbs, S., 2017. CEO selection and executive appearance. Working Paper,

University of Alabama. Costa, P., McCrae, R., 1988. Personality in adulthood: A six-year longitudinal study of self-

reports and spouse rating on the NEO personality inventory. Journal of Personality and Social Psychology 54, 853-863.

Custódio, C., Ferreira, M., Matos, P., 2013. Generalists versus specialists: Lifetime work

experience and chief executive officer pay. Journal of Financial Economics 108, 471-492. Custódio, C., Metzger, D., 2013. How do CEOs matter? The effect of industry expertise on

acquisition returns. Review of Financial Studies 26, 2008-2047. Custódio, C., Metzger, D., 2014. Financial expert CEOs: CEO’s work experience and firm’s

financial policies. Journal of Financial Economics 114, 125-154. Davis, A., Ge, W., Matsumoto, D., and Zhang, J., 2014. The effect of manager-specific optimism

on tone of earnings conference calls. Review of Accounting Studies 20: 639-673. DeGroot, T., Aime, F., Johnson, S.G., Kluemper, D., 2011. Does talking the talk help walking

the walk? An examination of the effect of vocal attractiveness in leader effectiveness. The Leadership Quarterly 22, 680-689.

Demerjian, P., Lev, B., McVay, S., 2012. Quantifying managerial ability: A new measure and

validity tests. Management Science 58, 1229-1248. Dewaele, J. M., Furnham, A., 1999. Extraversion: The unloved variable in applied linguistic

research. Language Learning 49, 509-544. Dikolli, S., Keusch, T., Mayew, W., and Steffen, T., 2014. Using shareholder letters to measure

CEO integrity. Working Paper, Duke University. Edmans, A., Gabaix, X., Landier, A., 2009. A multiplicative model of optimal CEO incentives in

market equilibrium. Review of Financial Studies 22, 4881–4917.

50

Engelberg, J., Gao, P., Parsons, C., 2013. The Price of a CEO’s Rolodex. Review of Financial

Studies 26, 79-114. Falato, A., Li, D., Milbourn, T., 2015. Which skills matter in the market for CEOs? Evidence

from pay for CEO credentials. Management Science, 61: 2845-2869. Fama, E.F., French, K.R., 1997. Industry cost of equity. Journal of Financial Economics 43,

153-193. Feldstein, S., Sloan, B., 1984. Actual and stereotyped speech tempos of extraverts and introverts.

Journal of Personality 52: 188-204. Furnham, A., 1990. Language and personality. In Giles, H., and Robins, W. (Eds), Handbook of

Language and Social Psychology. Winley. Gabaix, X., Landier,A., 2008. Why had CEO pay increased so much? Quarterly Journal of

Economics 123, 49-100. Giles, H., Street, R. L., 1994. Communicator characteristics and behavior. In M. L. Knapp and

G. R. Miller, eds., Handbook of Interpersonal Communication, 2nd ed. Thousand Oaks, California.

Gill, A., Oberlander, J., 2003. Perception of e-mail personality at zero-acquaintance:

Extraversion takes care of itself, neuroticism is a worry. In Proceedings of the 25th Annual Conference of the Cognitive Science Society, 456-461.

Goldberg, L.R., 1990. An alternative “description of personality”: The big-five factor structure.

Journal of Personality and Social Psychology 59, 1216-1229. Goodman, T., Neamtiu, M., Shroff, N., and White, H., 2014. Management forecast quality and

capital investment decisions. The Accounting Review 89, 331-365. Gow, I., Kaplan, S., Larcker, D., and Zakolyukina, A., 2015. CEO personality and firm policies.

Working Paper, Harvard University. Graham, J. R., Harvey, C. R., Puri. M., 2013. Managerial attitudes and corporate actions. Journal

of Financial Economics 109, 103-121. Graham, J. R., Harvey, C. R., Puri. M., 2014. A corporate beauty contest. Working Paper, Duke

University. Graham, J. R., Li., S., Qiu, J., 2012. Managerial attributes and executive compensation. Review

of Financial Studies 25, 144-186. Green, T.C., Jame, R., Markov, S., and Subasi, M., 2014a. Broker-hosted investor conferences.

51

Journal of Accounting and Economics 58, 142-166. Green, T.C., Jame, R., Markov, S., and Subasi, M., 2014b. Access to management and the

informativeness of analyst research. Journal of Financial Economics 114, 239-255. Hambrick, C., Mason, P., 1985. Upper Echelons: The organization as a reflection of its top

managers. The Academy of Management Review 9, 193-206. Hermann, M.G., 1999. Assessing leadership style: A trait analysis. Social Science Automation,

https://socialscience.net/docs/LTA.pdf/ Hirshleifer, D., Low, A., Teoh, S.H., 2012. Are confident CEOs better innovators? Journal of

Finance 67, 1457-1498. Hollander, S., Pronk, M., Roelofsen, E., 2010. Does silence speak? An empirical analysis of

disclosure choices during conference calls. Journal of Accounting Research 48, 531-563. Hutton, A., and Stocken, P., 2009. Prior forecasting accuracy and investor reaction to

management earnings forecasts. Working Paper, Boston College. Jegadeesh, N., Wu. D., 2013. Word power: A new approach for content analysis. Journal of

Financial Economics 110, 712-729. Jia, Y., Van Lent L., Zeng, Y., 2014. Masculinity, testosterone, and financial misreporting.

Journal of Accounting Research 52, 1195-1246. John, O. P., Srivastava, S., 1999. The Big Five trait taxonomy: History, measurement, and

theoretical perspectives. Handbook of personality: Theory and research, 2, 102-138. Judge, T., Bono, J., Ilies, R., Gerhardt, M., 2002. Personality and leadership: A qualitative and

quantitative review. Journal of Applied Psychology 87, 765-780. Jung, C. J., 1921. Psychologischen Typen. Rascher Verlag, Zurich – translation H.G. Baynes,

1923. Kamiya, S., Kim, Y.H., Suh, J., 2016. The face of risk: CEO testosterone and risk taking?

Working paper, Nanyang Business School. Kaplan, S., Klebanov, M., Sorensen, M., 2012. Which CEO characteristics matter? Journal of

Finance 67, 973-1007. Kaplan, S., Sorensen, M., 2016. Are CEOs different? Characteristics of top managers. Working

Paper, University of Chicago. Khurana, R., 2004. Searching for a corporate savior: The irrational quest for charismatic CEOs.

Princeton, N.J.: Princeton University Press.

52

Khurana, R., 2002. The curse of the superstar CEO. Harvard Business Review: September, 3-8 Larcker, D. F., Zakolyukina, A., 2012. Detecting deceptive discussions in conference calls.

Journal of Accounting Research 50, 495-540. Li, F. 2008. Annual report readability, current earnings, and earnings persistence. Journal of

Accounting and Economics 45, 221-247. Li, F., Minnis, M., Nagar, V., and Rajan, M., 2014. Knowledge, compensation, and firm value:

An empirical analysis of firm communication. Journal of Accounting and Economics 58, 96-116.

Loughran, T., McDonald, B., 2011. When is a liability not a liability? Textual analysis,

dictionaries, and 10-Ks. Journal of Finance 66, 35-65. Loughran, T., McDonald, B., 2013. IPO first-day returns, offer price revisions, volatility, and

form S-1 language. Journal of Financial Economics 109, 307-326. Loughran, T., McDonald, B., 2014. Measuring readability in financial disclosures. Journal of

Finance 69, 1643-1671. Mairesse, F., Walker, M. A., Mehl, M. R., Moore, R.K., 2007. Using linguistic cues for the

automatic recognition of personality in conversation and text. Journal of Artificial Intelligence Research 30, 457-500.

Malmendier, U., Tate, G., 2005. CEO overconfidence and corporate investment. Journal of

Finance 60, 2661–2700. Malmendier, U., Tate, G., 2008. Who makes acquisitions? CEO overconfidence and the market’s

reaction. Journal of Financial Economics 89, 20–43. Malmendier, U., Tate, G., 2009. Superstar CEOs. Quarterly Journal of Economics 124, 1593–

1638. Malmendier, U., Tate, G., Yan, J., 2011. Overconfidence and early-life experiences: the effect of

managerial traits on corporate financial policies. Journal of Finance 66, 1685–1731. Mann, R.D., 1959. A review of the relationship between personality and performance in small

groups. Psychological Bulletin 56, 241-270. Masulis, R., Mobbs, S., 2014. Independent director incentives: Where do talented directors spend

their limited time and energy? Journal of Financial Economics 111, 406-429. Mayew, W., Venkatachalam. M., 2012. Speech analysis in financial markets. Foundations and

Trends in Accounting 7, 73-130.

53

Mayew, W., Parsons, C., Venkatachalam. M., 2013. CEO voice pitch predicts labor market

success. Evolution and Human Behavior 34, 243-248. McClelland, D.C., and Windter, D.G., 1969. Motivating economic achievement. New York: Free

Press, 1969. Murphy, K.J., 1999. Executive compensation. North Holland, Amsterdam. Murphy, K. J., Zabojnik, J., 2004. CEO pay and appointments: A market-based explanation for

recent trends. American Economic Review Papers and Proceedings 94, 192-196. Nguyen, B.D., and Nielsen, K.M., 2014. What death can tell: Are executives paid for their

contributions to firm value? Management Science 60, 2994-3010. Norman, W. T., 1963. Toward an adequate taxonomy of personality attributes: Replicated factor

structure in peer nomination personality rating. Journal of Abnormal and Social Psychology, 66, 574–583.

Oberlander, J., Nowson, S., 2006. Whose thumb is it anyway? classifying author personality

from weblog text. In Proceedings of the 44th Annual Meeting of the Association for Computational Linguistics (ACL).

Oster, E., 2016. Unobservable selection and coefficient stability: Theory and evidence. Journal

of Business and Economic Statistics, DOI:10.1080/07350015.2016.1227711. Otto, C., 2014. CEO optimism and incentive compensation. Journal of Financial Economics

114, 366-404. Parrino, R., 1997. CEO turnover and outside succession a cross-sectional analysis. Journal of

Financial Economics 46, 165-197. Paulhus, D., Morgan, K.L., 1997. Perceptions of Intelligence in leaderless groups: The dynamic

effects of shyness and acquaintance. Journal of Personality and Social Psychology 72, 581-591.

Pennebaker, J. W., King, L.A., 1999. Linguistic styles: Language use as an individual difference.

Journal of Personality and Social Psychology 77, 1296–1312. Perez-Gonzalez, F., 2006. Inherited control and firm performance. American Economic Review

96, 1559-1588. Persico, N., Postlewaite, A., Silverman, D., 2004. The effect of adolescent experience on labor

market outcomes: The case of height. Journal of Political Economy 112, 1019-1053.

54

Quinlan, J. R., 1992. Learning with continuous classes. In Proceedings of the 5th Australian joint Conference on Artificial Intelligence 92, 343-348.

Ryan, H., Wiggins, R., 2004. Who is in whose pockets? Director compensation, board

independence, and barriers to effective monitoring. Journal of Financial Economics 73, 497–524.

Scherer, K. R. 1979. Personality markers in speech. In Scherer, K. R., & Giles, H. (Eds.), Social

markers in speech, pp. 147–209. Cambridge University Press. Scherer, K. R. 2003. Vocal communication of emotion: A review of research paradigms. Speech

Communication 40, 227–256. Sitkin, S. B., Pablo, A., 2005. The neglected importance of leadership in mergers and

acquisitions. In G. K. Stahl & M. E. Mendenhall (Eds.), Mergers and acquisitions: Managing culture and human resources (pp. 412-422). Stanford, CA: Stanford University Press.

Smith, C.W., Watts, R., 1992. The investment opportunity set and corporate finance, dividend,

and compensation policies. Journal of Financial Economics 32, 263-292. Stogdill, R., 1974. Handbook of leadership: A survey of theory and research. New York: Free

Press. Swann, W. B., Rentfrow, P.J., 2001. Blirtatiousness: Cognitive, behavioral, and physiological

consequences of rapid responding. Journal of Personality and Social Psychology 81, 1160-1175.

Tetlock, P. Saar-Tsechansky, M., Macskassy, S., 2008. More than words: Quantifying language

to measure firms’ fundamentals. Journal of Finance 63, 1437-1467. Vapnik, V., Golowich, S., Smola, A., 1997. “Support Vector Method for Function

Approximation, Regression Estimation, and Signal Processing”, in M. Mozer, M. Jordan, and T. Petsche (eds.), Neural Information Processing Systems, Vol. 9. MIT Press, Cambridge, MA.

Waldman, D., Javidan, M., 2009. Alternative forms of charismatic leadership in the integration

of mergers and acquisitions. The Leadership Quarterly 20, 130-142. Yermack, D., 2004. Remuneration, retention, and reputation incentives for outside directors.

Journal of Finance 59, 2281–2308. Zuckerman, M., Kulhman, D.M., 2000. Personality and risk-taking: Common biosocial factors,

Journal of Personality 68, 999-1029.

55

Table 1 Summary Statistics

This table presents descriptive statistics for the sample of CEO, CFO, and firm-level variables over the 2004-2013 sample period. Panel A reports summary statistics for the 2,266 unique CEOs (12,213 CEO-years) for whom we are able to estimate a valid extraversion score. Panel A also presents analogous results for 2,362 unique CFOs (11,326 CFO-years). Panel B reports summary statistics for the 1,633 unique firms (12,213 firm-years) for which we observe CEO extraversion. The transcripts of the conference calls are obtained from Seeking Alpha and Thomson Reuters during the 2006-2013 sample period. Aggregate Extraversion scores are estimated based on the executive’s responses during the questions-and-answers portion of conference calls. The extraversion scores are computed using the average of four linguistic algorithms described in Section IA.1 of the Internet Appendix. We limit the sample to executives that appear on at least three conference calls. We match the executives on the conference calls with executives in the Execucomp database by CUSIP and name. Definitions of the other variables are presented in Appendix A. Panel A: Executive Variables

CEOs

CFOs

Mean Median Std. Dev

Mean Median Std. Dev

Extraversion 4.15 4.14 0.37

3.61 3.60 0.37 Total Calls 20.16 20.00 10.19

18.75 18.00 10.09

Tenure 5.15 5.00 2.83

3.84 3.00 2.70 Compensation ($ Mil) 5.55 3.72 6.30

1.97 1.34 2.41

Outside Directorships 0.45 0.00 0.69

0.02 0.00 0.23 Age at First Appointment 50.68 51.00 6.65

46.29 46.00 6.11

Executive Turnover 0.05 0.00 0.22

0.06 0.00 0.25 Agreeableness 3.66 3.66 0.13

3.70 3.70 0.12

Conscientiousness 3.68 3.68 0.21

3.67 3.67 0.19 Openness 3.73 3.73 0.12

3.75 3.75 0.12

Emotional Stability 3.23 3.22 0.18

3.23 3.24 0.16 Panel B: Firm-Level Variables

Mean Median Std. Dev

Assets ($ Bil) 17.73 2.15 107.22 Sales ($ Bil) 6.89 1.59 21.03 Market Equity ($ Bil) 8.54 1.87 26.19 Volatility 2.65 2.33 1.43 Fiscal Return 7.18 1.25 54.11 Firm Age (months) 268.19 235.00 150.81 Q 1.85 1.49 1.17 Sales Growth 0.08 0.07 0.20 Market Share (%) 0.73 0.15 2.20 Firm Efficiency 0.35 0.29 0.17 Profitability (OCF) 0.12 0.11 0.10 Profit Margin 0.05 0.06 0.15 ROA 0.14 0.13 0.12 Analyst Coverage 14.40 12.00 9.52 Conference Presentations 4.80 4.00 5.22 Media Articles 7.55 0.00 40.50 Media Words (Thousands) 5.01 0.00 27.86 Share Turnover 11.31 8.97 8.69 Amihud Illiquidity (x 100) 27.49 0.97 370.85

56

Table 2 Determinants of Conference Call Extraversion

This table reports estimates from regressing executive extraversion, measured from conference call Q&A responses, on manager, firm, and call characteristics (see Equation 1 in Section 3.1). We report the results separately for CEOs and CFOs. All specifications include quarter fixed effects. Specifications 2 and 5 add managerial characteristics, and Specifications 3 and 6 add manager fixed effects. The sample includes conference calls obtained from Seeking Alpha and Thomson Reuters over the 2006-2013 sample period that can be matched with Execucomp. Definitions for all variables are provided in Appendix A. All continuous variables are standardized to have mean 0 and variance 1. Standard errors are clustered by firm, and t-statistics are reported below each estimate. CEOs CFOs

[1] [2] [3] [4] [5] [6]

Returnt-63,t-2 -0.02 -0.02 -0.02 -0.01 -0.02 -0.01 (-3.32) (-4.67) (-4.12) (-2.79) (-3.19) (-1.52)

Returnt-1,t+1 -0.02 -0.02 -0.02 -0.01 -0.01 0.00 (-3.65) (-4.19) (-3.50) (-2.56) (-2.52) (0.00)

Returnt+2,t+63 0.00 -0.01 -0.01 0.00 0.00 0.01 (-0.69) (-1.91) (-2.13) (0.65) (0.46) (1.86)

Earnings Call 0.09 0.11 0.14 -0.26 -0.26 -0.16 (3.00) (3.72) (6.42) (-9.75) (-9.80) (-7.15)

MBE 0.06 0.02 0.00 0.07 0.06 0.00 (2.96) (1.20) (0.29) (3.82) (3.32) (-0.27)

Surprise -0.02 -0.01 0.00 -0.03 -0.03 0.00 (-2.23) (-1.59) (-0.82) (-3.16) (-2.96) (0.45)

Loss -0.17 -0.11 -0.06 -0.04 -0.01 0.02 (-3.78) (-2.76) (-2.84) (-1.24) (-0.43) (0.98)

Tenure 0.08 -0.03 0.09 0.04 (4.67) (-2.94) (5.99) (4.11)

Exec Age -0.13 -0.01 (-7.48) (-0.93)

Male

0.30

0.25

(4.40)

(5.54)

Optimism (Tone)

0.04

0.05

(2.53)

(3.65)

Overconfidence (HD67)

0.01

(0.33)

Founder

0.11

(2.65)

General Ability Index

0.03

(1.75)

Rolodex

0.02

(1.24)

MBA

0.00

(-0.05)

Doctorate

0.06

(0.71)

Grad Honors

0.07

(1.38)

Ivy League

-0.04

(-1.40)

Quarter FE Yes Yes Yes Yes Yes Yes Manager FE No No Yes No No Yes Obs. 36,934 36,934 36,934

37,644 37,644 37,644

Total R-squared 0.95% 4.46% 49.54%

1.30% 2.62% 41.06% Within R-squared

1.02%

1.23%

57

Table 3 Correlations Between CEO Extraversion and Firm and Manager Characteristics

The table reports Pearson correlations between CEO extraversion and firm and manager characteristics. Extraversion of CEOs is estimated using the weighted average of residuals from Specification 1 of Table 2. Firm and manager characteristics are defined in Appendix A. Statistical significance at the 5% level is indicated with a bold estimate. Statistical significance is computed from standard errors clustered by firm.

Log

(Com

p)

Log

(Sal

es)

Log

(Q)

Log

(Vol

)

Log

(Firm

A

ge)

Tenu

re

Exec

Age

Mal

e

Opt

imis

m

Ove

r C

onfid

ent

Foun

der

Cha

ir

GA

I

Rol

odex

MB

A

Gra

dHon

ors

Ivyl

eagu

e

Perc

ent C

EO

Text

Extraversion 0.20 0.14 0.13 0.02 -0.07 0.07 -0.15 0.06 0.07 0.02 0.07 0.01 0.02 0.04 0.00 0.02 -0.01 0.52 Log (Comp) 0.67 0.09 -0.27 0.18 -0.06 0.07 0.01 0.09 -0.10 -0.12 0.19 0.33 0.37 0.05 0.09 0.10 0.12 Log (Sales)

-0.12 -0.40 0.33 -0.12 0.12 0.00 0.04 -0.06 -0.21 0.22 0.32 0.41 0.03 0.06 0.07 0.06 Log (Q)

-0.09 -0.17 0.02 -0.13 0.01 0.16 0.00 0.09 -0.04 -0.08 0.01 0.01 0.00 0.02 0.09 Log (Vol)

-0.29 0.05 -0.07 -0.02 0.02 -0.01 0.15 -0.16 -0.13 -0.17 -0.02 -0.03 -0.06 0.04 Log (Firm Age)

-0.01 0.20 0.02 -0.08 0.09 -0.56 0.12 0.11 0.17 0.06 -0.02 0.04 0.00 Tenure

0.37 0.07 -0.05 0.05 0.44 0.33 -0.13 -0.06 -0.07 -0.03 0.05 0.09 Exec Age

0.04 -0.15 0.04 0.10 0.28 0.15 0.05 -0.03 -0.04 0.02 -0.06 Male

-0.02 0.03 0.01 0.03 -0.08 -0.05 0.03 0.00 0.00 0.02 Optimism

-0.03 0.01 -0.04 -0.03 0.07 0.03 -0.03 -0.02 0.04 Overconfidence

0.00 0.02 -0.08 -0.03 0.03 -0.04 0.02 -0.05 Founder

0.12 -0.17 -0.16 -0.10 -0.02 0.00 0.01 Chair

0.13 0.15 -0.02 0.03 0.06 -0.02 GAI

0.40 0.08 0.12 0.12 0.05 Rolodex

0.06 0.13 0.13 0.01 MBA

0.05 0.21 0.00 GradHonors

0.17 0.00 Ivyleague

-0.02

58

Table 4 Extraversion and CEO Compensation

This table reports estimates from the following panel regression: 𝐿𝑜𝑔(𝐶𝑜𝑚𝑝)𝑖𝑡 = 𝛽1𝐸𝑥𝑡𝑟𝑎𝑣𝑒𝑟𝑠𝑖𝑜𝑛𝑖 + 𝛾𝑭𝒊𝒓𝒎𝑪𝒉𝒂𝒓 + 𝛿𝑷𝒆𝒓𝒇𝒐𝒓𝒎𝒂𝒏𝒄𝒆 + ω𝐂𝐄𝐎𝐂𝐡𝐚𝐫 + 𝑌𝑒𝑎𝑟𝑡 + 𝐹𝐸𝑖 + 𝜀𝑖𝑡.. Comp is total compensation and is comprised of salary, bonus, value of restriction stock granted, value of options granted, long-term incentive payout, and other compensation (TDC1 as reported in Execucomp). Extraversion is the residual extraversion of CEOs based on Specification 1 of Table 2. FirmChar is a vector of firm characteristics, Performance is a vector of firm performance measures, and CEOChar is a vector of CEO individual characteristics. All continuous independent variables are standardized to have mean 0 and variance 1. Appendix A provides detailed variable definitions. Specifications 1-4 (5-8) include industry (firm) and year fixed effects. Standard errors are clustered by firm, and t-statistics are reported below the coefficients for Extraversion. In the interest of brevity, we delegate t-statistics for all other variables to Internet Appendix Table IA.7. ***, **, and * denote statistical significance at the 1, 5, and 10% levels respectively. The sample includes 10,925 observations. Industry and Year Fixed Effects Firm and Year Fixed Effects [1] [2] [3] [4] [5] [6] [7] [8] Extraversion 17.90*** 6.10*** 5.70*** 5.70*** 6.32*** 5.74*** 6.00*** 7.57*** (8.41) (4.53) (4.27) (3.65) (2.98) (2.70) (2.83) (3.35) Ln (Sales) 25.40*** 20.40*** 18.50*** 5.21 17.64** 17.87** Ln (Assets) 48.70*** 55.20*** 50.60*** 34.15*** 36.84*** 36.31*** Ln (Q) 17.60*** 12.70*** 11.60*** 15.93*** 11.78*** 11.56*** Ln (Vol) 3.60*** 3.30** 3.40** -2.03 -0.94 -1.09 Ln (Firm Age) -2.20* -1.60 -1.40 7.75** 7.84** 5.76 Ln (Sales Growth) 5.60*** 5.50*** 5.05*** 4.99*** Fiscal Ret 5.30*** 4.90*** 4.48*** 4.45*** Lag Fiscal Ret 4.10*** 4.00*** 3.38*** 3.31*** Profitability 3.40** 3.40** 3.98*** 3.99*** Prof. Growth -0.20 -0.20 -0.51 -0.48 Loss Dummy -4.90* -4.70* -11.14*** -10.63*** Log (CEO Tenure) -3.30** 2.56* Log (CEO Age) 1.70 -2.80 Male 2.80 5.24 Founder 0.40 -5.93 Chair 7.40*** 3.12 GAI 7.60*** 5.78*** Rolodex 2.00 0.75 Percent CEO Text -0.20 -0.07 Optimism 4.10*** 1.16 Overconfidence -6.60*** -3.56** MBA 2.80 5.03 Doctorate 12.50 -9.28 Ivy League 2.00 -0.68 Grad with Honors 4.80 0.24 Emotional Stab 1.80 -2.66 Openness -0.70 -2.02 Agreeableness -1.30 -0.83 Conscientiousness 0.70 -0.51 R-squared 8.95% 55.24% 57.13% 58.60% 79.41% 80.29% 81.06% 81.26%

59

Table 5 Extraversion and Age of First Appointment to CEO or CFO

This table reports estimates from the following panel regression: 𝐹𝑖𝑟𝑠𝑡_𝐴𝑔𝑒𝑖 = 𝛽1𝐸𝑥𝑡𝑟𝑎𝑣𝑒𝑟𝑠𝑖𝑜𝑛𝑖 + 𝛾𝑭𝒊𝒓𝒎𝑪𝒉𝒂𝒓 + ω𝐂𝐄𝐎𝐂𝐡𝐚𝐫 + 𝑌𝑒𝑎𝑟𝑡 + 𝐹𝐸𝑖 + 𝜀𝑖𝑡.

First_Age is the age at which the executive was first appointed to CEO. Extraversion is the residual extraversion of CEOs based on Specification 1 of Table 2. FirmChar and CEOChar are vectors of firm and manager characteristics detailed in Appendix A. FE indicates an industry (Specifications 1, 3, and 4) or firm (Specification 2) fixed effect. All independent variables are standardized to have mean 0 and variance 1. Standard errors are clustered by firm, and t-statistics are reported below the coefficients for Extraversion. In the interest of brevity, we delegate t-statistics for all other variables to Internet Appendix Table IA.8. ***, **, and * denote statistical significance at the 1, 5, and 10% levels respectively. All Hires External Hires Internal Hires

[1] [2] [3] [4]

Extraversion -0.79*** -1.12*** -0.98*** -0.65***

(-4.67) (-3.28) (-3.40) (-2.96)

Ln (Sales) 0.36 1.32 0.20 0.34 Ln (Assets) -0.01 -0.21 0.06 -0.08 Ln (Q) -0.81** -1.26 -0.44 -1.10** Ln (Vol) 0.20 1.35** 0.16 0.45 Ln (Firm Age) 0.81*** -1.29 1.11*** 0.95*** Male 1.85*** 0.33 1.93* 2.42*** Founder 3.54*** -0.27 3.12* GAI 1.30*** 1.39*** 1.72*** 0.80*** Rolodex 0.00 -0.10 -0.24 0.18 Optimism -0.78*** -0.56 -1.14*** -0.49** Overconfidence -0.99** -0.17 -0.03 -0.42 MBA -0.46 -0.92* -1.57*** 0.11 Doctorate -0.13 -2.48 -2.31 -0.71 Ivy League -0.65*** -1.06** -0.12 -0.71* Grad with Honors -0.90** -0.51 -0.39 -1.18** Emotional Stability -0.49*** -0.24 -0.76** -0.58** Openness 0.23 0.37 0.21 0.31 Agreeableness -0.27 -0.08 0.30 -0.75*** Conscientiousness -0.48** -1.22*** -0.82** -0.19 Fixed Effects Industry & Year Firm & Year Industry & Year Industry & Year R-squared 27.50% 82.76% 32.22% 22.01% Observations 1,773 1,773 569 963

60

Table 6 Extraversion and CEO Tenure and Turnover

This table reports estimates from the following panel regression:

𝐶𝐸𝑂 𝑇𝑛𝑖𝑡 = 𝛽1𝐸𝑥𝑡𝑟𝑎𝑣𝑒𝑟𝑠𝑖𝑜𝑛𝑖 + 𝛾𝑭𝒊𝒓𝒎𝑪𝒉𝒂𝒓 + 𝛿𝑷𝒆𝒓𝒇𝒐𝒓𝒎𝒂𝒏𝒄𝒆 + 𝜔𝑪𝑬𝑶𝑪𝒉𝒂𝒓 + 𝑌𝑒𝑎𝑟𝑡 + 𝐹𝐸𝑖 + 𝜀𝑖𝑡. 𝐶𝐸𝑂 𝑇𝑛𝑖𝑡 denotes CEO Turnover in Specifications 1-3 and is 1 if firm i changes its CEO in year t, and 0 otherwise. In Specifications 4-6, 𝐶𝐸𝑂 𝑇𝑛𝑖𝑡 denotes CEO Tenure and is the log of CEO tenure in months. Extraversion is the residual extraversion of CEOs based on Specification 1 of Table 2. FirmChar, Performance, and CEOChar are vectors of firm, performance, and manager characteristics detailed in Appendix A. FE indicates industry fixed effects. All independent variables are standardized to have mean 0 and variance 1. The coefficients from the logistic regressions represent odds ratios. Standard errors are clustered by firm, and z-scores (in Specifications 1-3) and t-statistics (in Specifications 4-6) are reported below the coefficients for Extraversion. In the interest of brevity, we delegate test statistics for all other variables to Internet Appendix Table IA.9. ***, **, and * denote statistical significance at the 1, 5, and 10% levels respectively. The sample includes 10,925 observations.

Turnover Tenure

[1] [2] [3]

[4] [5] [6]

Extraversion 0.81*** 0.78*** 0.88*

5.15** 7.68*** 5.72***

(-4.40) (-4.94) (-1.75)

(2.27) (3.36) (3.15)

Ln (Sales)

1.04 0.83

-5.86 -10.88***

Ln (Assets)

1.23 1.56***

-12.35** -11.73*** Ln (Q)

1.11 1.29***

-2.97 -2.95

Ln (Vol)

1.13** 0.99

-0.88 2.91* Lag (Age)

1.08* 0.93

6.02*** 21.88***

Ln (Sales Growth)

1.00 1.00

5.12*** 2.55*** Fiscal Return

0.73*** 0.90

-2.79*** -1.42*

Lag Fiscal Return

0.85 0.92

-0.02 -0.09 Profitability

1.02 1.07

0.31 -0.52

Profitability Growth

0.93 0.92

-1.42 0.29 Loss Dummy

1.31* 0.99

-18.76*** -5.67**

Log (CEO Age)

0.68 1.55***

22.82*** Log (CEO Tenure) 1.45*** Male

1.45***

19.02**

Founder

1.08

106.18*** Chair

-0.64**

38.65***

GAI

-0.66***

-8.65*** Rolodex

1.20***

8.07***

Percent CEO Text

1.07

3.07* Optimism

1.10

-0.13

Overconfidence

-0.96

1.24 MBA

-0.86

-4.59

Doctorate

-0.99

-24.70*** Ivy League

1.38

9.82***

Grad with Honors

-0.88

-0.10 Emotional Stability

-0.77

4.05**

Openness

-0.83***

-3.24 Agreeableness

1.17**

-0.36

Conscientiousness

1.07

3.20 Fixed Effects Ind. &Year Ind. &Year Ind. &Year

Ind. &Year Ind. &Year Ind. &Year

R-squared 3.24% 3.84% 12.47%

2.28% 5.11% 51.37%

61

Table 7

CEO Extraversion and Outside Directorships This table reports the estimates from the following panel regression:

𝑌𝑖𝑡 = 𝛽1𝐸𝑥𝑡𝑟𝑎𝑣𝑒𝑟𝑠𝑖𝑜𝑛𝑖 + 𝛾𝑭𝒊𝒓𝒎𝑪𝒉𝒂𝒓 + 𝛿𝑷𝒆𝒓𝒇𝒐𝒓𝒎𝒂𝒏𝒄𝒆 + 𝜔𝑪𝑬𝑶𝑪𝒉𝒂𝒓 + 𝑌𝑒𝑎𝑟𝑡 + 𝐹𝐸𝑖 + 𝜀𝑖𝑡. In specification 1, Y is the log of one plus the number of directorships. In specification 2, Y is a dummy variable that equals 1 if the CEO sits on an outside board and zero otherwise. In Specifications 3 through 5, Y reflects the size of the CEOs largest outside directorships as measured by Sales, Assets, or Market Equity, and the sample is limited to the sample of CEOs with outside directorships. Extraversion is the residual extraversion as in Specification 1 of Table 2. Controls are defined in Appendix A, and all independent continuous variables are standardized to have mean 0 and variance 1. Standard errors are clustered by executive, and t-statistics (in Specifications 1, 3-5) and z-scores (for the odds ratio in Specification 2) are reported below the coefficients for Extraversion. In the interest of brevity, we delegate test statistics for all other variables, and the coefficients for other personality dimensions, to Internet Appendix Table IA.10. ***, **, and * denote statistical significance at the 1, 5, and 10% levels respectively.

Ln

(1 + Dir.) Logit

(Dir. =1) Dir. Size: Ln(Sales)

Dir. Size: Ln(Assets)

Dir. Size: Ln(Equity)

[1] [2]

[3] [4] [5] Extraversion 1.99** 1.16**

11.38%* 21.58%*** 17.20%**

(2.19) (2.29)

(1.68) (2.68) (2.49)

Ln (Sales) 2.76% 1.22

56.93%*** 59.10%*** 38.55%** Ln (Assets) -0.54% 1.03

29.97%* 35.45%* 51.06%***

Ln (Q) -1.39% 0.91

3.93% 2.06% 16.68%** Ln (Vol) -1.31% 0.94

7.92% 18.05%** 9.10%

Ln (Firm Age) 0.45% 1.04

5.64% 7.97% 7.31% Ln (Sales Growth) 0.67%* 1.06**

-2.81% -5.02% -2.57%

Fiscal Return -0.29% 0.97

0.49% 0.93% 5.90% Lag Fiscal Return -0.39% 0.96

-3.94% -3.76% -3.13%

Profitability -0.07% 1.03 19.95%** 20.91%** 20.72%** Profitability Growth 0.01% 0.98 -7.92%* -6.18% -7.93% Loss Dummy -2.63%* 0.83* 1.65% 1.76% 5.75% Log (CEO Tenure) 2.03%*** 1.12* 2.50% -2.47% -9.65% Log (CEO Age) 6.56%*** 1.50*** 2.07% 11.21% 14.59%** Male -15.00%*** 0.33*** -64.66%*** -79.32%*** -76.00%*** Founder -8.22%*** 0.55*** 30.42% 24.51% 25.37% Chair 3.81%*** 1.29*** 18.93%* 26.81%** 23.60%** GAI 13.58%*** 2.23*** 6.92% 14.72%* 11.23%* Percent CEO Text -0.97% 0.93 27.46%*** 31.51%*** 27.16%*** Optimism -1.42%* 0.94 5.44% 7.59% 3.32% Overconfidence -2.96%* 0.76*** 3.85% 6.03% 2.74% MBA 3.75%** 1.24* -12.77% -3.78% -1.11% Doctorate -1.04% 1.11 2.68% -9.13% -2.63% Ivy League 0.01% 1.01 -47.36% -50.16% -28.55% Grad with Honors -0.89% 0.95 10.53% 21.49%** 16.12%** Observations 9,637 9,637

2,222 2,222 2,222

R2/Psuedo R2 27.56% 26.61%

45.86% 43.71% 46.01%

62

Table 8 Extraversion and CFO Promotion to CEO

This table reports the odds ratios for the following logistic regression: 𝑃𝑟𝑜𝑚𝑜𝑡𝑖𝑜𝑛𝑖𝑡 = 𝛽1𝐸𝑥𝑡𝑟𝑎𝑣𝑒𝑟𝑠𝑖𝑜𝑛𝑖 + 𝛾𝑭𝒊𝒓𝒎𝑪𝒉𝒂𝒓 + 𝛿𝑷𝒆𝒓𝒇𝒐𝒓𝒎𝒂𝒏𝒄𝒆 +

ωCFOChar+ 𝜁𝑪𝒖𝒎𝑷𝒆𝒓𝒇𝒐𝒓𝒎𝒂𝒏𝒄𝒆 +𝑌𝑒𝑎𝑟𝑡 + 𝐹𝐸𝑖 + 𝜀𝑖𝑡. Promotion is a dummy variable equal to one if the internal CFO is promoted to CEO following the departure of the CEO. The sample consists of all CEO departures from 2006-2013 for which we have an extraversion score for the CFO before the transition. Extraversion is the residual extraversion of the CFOs, as in Specification 4 of Table 2. Controls include firm, recent performance, and manager characteristics, as well as the cumulative performance of the firm during the CFOs tenure (definitions in Appendix A). All regressions include industry and year fixed effects, and all independent variables are standardized to have mean 0 and variance 1. Specifications 1-3 examine all promotions, 4-5 partition the sample into short and long-tenured CFOs based on median tenure, and 6-7 split the sample into firms with departing CEOs above or below median Extraversion. Standard errors are clustered by firm, and z-scores are reported below the odds ratio for Extraversion. In the interest of brevity, we delegate z-scores for all other variables to Internet Appendix Table IA.11. ***, **, and * denote statistical significance at the 1, 5, and 10% levels respectively. CFO Tenure CEO Extraversion

All Promotions

Short Long

Extravert Introvert

[1] [2] [3] [4] [5] [6] [7] Extraversion 1.52*** 1.41** 1.31* 2.61** 1.23 1.90** 1.03

(3.02) (2.35) (1.79)

(2.32) (1.09)

(2.29) (0.07)

Ln (Sales) 0.85 0.85 0.92

0.80 0.98

1.07 0.85

Ln (Q) 1.18 1.15 1.17

1.86** 0.92

1.02 0.57 Lag Fiscal Ret 0.91 0.83 0.82

0.56 0.84

0.94 1.17

Lag Profitability 0.94 0.93 0.95

0.66 1.05

0.82 1.04 Tenure 1.08 1.09 1.08

0.81 1.16

1.60* 1.07

Exec Age 0.51 0.51 0.46*

2.20 0.25**

0.15* 0.54 Optimism 0.89 0.89 0.90

1.15 0.67

0.99 0.61

Emotional Stability 1.06 1.06 1.11

0.79 1.14

1.14 1.46 Openness 1.15 1.11 1.09

1.17 0.94

1.07 1.04

Agreeableness 1.05 1.07 1.09

1.04 1.33

0.97 2.24 Conscientiousness 0.90 0.90 0.91

0.86 0.99

0.85 0.86

Cumulative Returns 1.19 1.18

0.93 2.60***

1.21 1.64 Relative For. Error 1.12 1.12

1.45 0.82

0.29** 1.42*

Guidance Dummy 0.88 0.89

1.38 0.82

0.58 0.81 CFO Percent Text 1.18 1.17

0.98 1.21

1.62** 0.96

Relative Salary 1.81***

0.65 3.22***

2.39*** 6.84* Observations 1171 1171 1171

439 547

453 379

Pseudo R-squared 5.33% 5.71% 6.89%

11.47% 13.17%

15.46% 14.65% Obs. CFO Promotion = 1 93 93 93 24 54 36 18 Prob. of CFO Promotion 7.94% 7.94% 7.94%

5.47% 9.87%

7.95% 4.75%

63

Table 9 CEO Extraversion around Turnovers: Compensation, Investor Recognition, and Firm Performance

This table reports estimates from the following panel regressions:

𝛥𝑌𝑖𝑡+3,𝑡−1 = 𝛽1𝛥𝐸𝑥𝑡𝑟𝑎𝑣𝑒𝑟𝑠𝑖𝑜𝑛𝑖𝑡+3,𝑡−1 + 𝛽2𝛥𝐶𝐸𝑂𝐶ℎ𝑎𝑟𝑖𝑡+3,𝑡−1 + 𝜀𝑖𝑡 . Y denotes several different variables for firm i. The dependent variable is the change in average level in the three years after a CEO transition (years t+1 to t+3) relative to the level in the year prior to the transition (year t-1). Extraversion is the residual extraversion of CEOs based on Specification 1 of Table 2. All measures are industry adjusted by subtracting the median Y from a control group of firms. The control group consists of firms in the same Fama-French 12 industry group and the same quintile ranking of Y in the year prior to the executive transition. Specification 1 includes Extraversion alone. Specification 2 includes CEOChar, a vector of manager characteristics that is detailed along with the dependent variables in Appendix A. Standard errors are clustered by firm, and t-statistics are reported below each Extraversion coefficient. ***, **, and * denote statistical significance at the 1, 5, and 10% levels respectively.

[1] [2] Panel A: CEO Compensation Industry-Adj. Log (Total Compensation) 6.76*** 7.15***

(2.90) (2.75)

Panel B: Investor Recognition Industry-Adj. Log (Analyst Coverage) 5.60*** 4.41** (3.59) (2.50) Industry-Adj. Log (Conf. Presentations) 6.22*** 5.15** (3.09) (2.22) Industry-Adj. Log (Media Articles) 3.07 5.06 (0.88) (1.25) Industry-Adj. Log (Media Words) 18.66 24.90* (1.57) (1.77) Industry Adj. Log (Turnover) 4.19** 5.12** (2.38) (2.52) Industry-Adj. Log (Amihud Illiquidity) -13.24*** -14.38***

(-3.69) (-3.53)

Panel C: Firm Performance Industry-Adj. Log (Sales Growth) 2.57*** 2.47*** (3.10) (2.60) Industry-Adjusted Log (Market Share) 3.49** 4.24** (2.40) (2.38) Industry-Adjusted Firm Efficiency 1.21** 1.39** (2.25) (2.37) Industry-Adjusted Profitability (OCF) 0.37 0.57 (1.27) (1.75) Industry-Adjusted Profit Margin 0.72 0.68 (1.44) (1.20) Industry-Adjusted ROA 0.24 0.37 (0.86) (1.16) Industry-Adjusted Log (Q) 0.84 1.46 (0.74) (1.12) Industry-Adjusted Return 0.91 0.76 (0.76) (0.56)

64

Table 10 CEO Extraversion and Departure Announcement Returns

This table reports the estimates from the following regression: 𝐶𝐴𝑅𝑖𝑡 = 𝛽1𝐸𝑥𝑡𝑟𝑎𝑣𝑒𝑟𝑠𝑖𝑜𝑛𝑖𝑡 + 𝜷𝟐𝑭𝒊𝒓𝒎𝑪𝒉𝒂𝒓𝑖𝑡 + 𝜷𝟑𝑪𝑬𝑶𝑪𝒉𝒂𝒓𝑖𝑡 + 𝐼𝑛𝑑𝑖 + 𝑌𝑒𝑎𝑟𝑡 + 𝜀𝑖𝑡 .

CAR is the three-day market-adjusted returns around the announcement of a CEO departure for firm i, and Extraversion is the extraversion score of the departing CEO. FirmChar and CEOChar are vectors of the firm and CEO characteristics controls as described in Appendix A. All independent variables are standardized to have mean 0 and variance 1, and Specifications 1-6 include industry and year fixed effects. Specifications 1-3 examine all departures, 4-6 focus on voluntary departures, and 7 examines unexpected departures, as defined in Section 5.1. Standard errors are clustered by firm, and t-statistics are reported below each Extraversion coefficient. In the interest of brevity, we delegate t-statistics for all other variables to Internet Appendix Table IA.12. ***, **, and * denote statistical significance at the 1, 5, and 10% levels respectively. All Departures Voluntary Departures Unexpected [1] [2] [3] [4] [5] [6] [7] Extraversion -0.19% -0.18% -0.21% -0.57%** -0.56%** -0.19% -2.42%***

(-0.59) (-0.57) (-0.49) (-2.20) (-2.17) (-0.61) (-2.74)

Ln (Sales)

1.31% 1.67%*

0.42% 0.35%

Ln (Assets)

-1.17% -1.13%

-0.25% -0.18% Ln (Q)

-0.46% -0.34%

-0.30% -0.21%

Ln (Vol)

-0.25% 0.08%

0.15% 0.24% Ln (Age)

-0.46% -0.62%

-0.26% -0.45%

Lag Fiscal Ret

-0.15% -0.21%

-0.23% -0.28% Log (CEO Tenure)

-0.32%

-0.02%

Log (CEO Age)

0.86%**

0.56% Male

-0.57%

-3.43%*

Founder

-0.94%

-0.68% Chair

-0.93%

0.06%

GAI

0.16%

-0.16% Rolodex

-0.36%

0.15%

Percent CEO Text

-0.49%

-0.54% Optimism

0.07%

-0.20%

Overconfidence

-0.45%

-0.25% MBA

-1.10%

-0.81%

Doctorate

2.24%*

2.34% Ivy League

0.37%

0.47%

Grad Honors

-0.72%

-1.93%** Emotional Stability

1.21%**

0.12%

Openness

0.06%

-0.01% Agreeableness

-0.64%

-0.51%

Conscientiousness

0.25%

0.64% Observations 757 757 742 531 531 520 15

R-squared 2.92% 4.24% 7.84% 4.73% 5.54% 8.86% 18.83%

65

Table 11

CEO Extraversion and M&A Announcement Returns This table reports estimates from the following panel regression:

𝐶𝐴𝑅𝑖𝑡 = 𝛽1𝐸𝑥𝑡𝑟𝑎𝑣𝑒𝑟𝑠𝑖𝑜𝑛𝑖𝑡 + 𝜷𝟐𝑫𝒆𝒂𝒍𝑪𝒉𝒂𝒓𝑖𝑡 + 𝜷𝟑𝑭𝒊𝒓𝒎𝑪𝒉𝒂𝒓𝑖𝑡 + 𝜷𝟒𝑪𝑬𝑶𝑪𝒉𝒂𝒓𝑖𝑡 + 𝐼𝑛𝑑𝑖 + 𝑌𝑒𝑎𝑟𝑡 + 𝜀𝑖𝑡 . CARit is three-day cumulative abnormal (market-adjusted) return for acquiring firm i centered at the announcement date t of the acquisition. Extraversion is the extraversion score of the CEO of the acquiring firm. DealChar is a vector of deal characteristics that are known to influence announcement returns. FirmChar and CEOChar are the vectors of the firm and CEO characteristics included as controls in Equation 3 (and described in Appendix A). All independent variables are standardized to have mean 0 and variance equal to 1. Standard errors are clustered by firm, and t-statistics are reported below each Extraversion coefficient. In the interest of brevity, we delegate t-statistics for all other variables to Internet Appendix Table IA.13. ***, **, and * denote statistical significance at the 1, 5, and 10% levels respectively. The sample includes 1,864 observations.

[1] [2] [3]

Extraversion 0.22 0.29** 0.38**

(1.45) (1.97) (2.09)

Tender

1.84*** 1.75**

Equity Finance

-1.72* -1.80* Mixed Finance

-0.56 -0.60

Public Target

-2.71*** -2.61*** Private Target

-0.96*** -0.90***

Ln (Sales)

-0.06 -0.03 Ln (Assets)

-0.31 -0.13

Ln (Q)

-0.26 -0.33* Ln (Vol)

0.03 0.10

Ln (Age)

-0.01 -0.06 Lag Fiscal Ret

-0.03 -0.10

Log (CEO Tenure) 0.21 Log (CEO Age) -0.04 Male 0.74 Founder -0.39 Chair -0.10 GAI 0.05 Rolodex -0.20 Percent Ceo Text 0.03 Optimism 1.37*** Overconfidence -0.09 MBA -0.20 Doctorate -0.06 IvyLeague -0.09 GradHonors 0.19 Emotional Stability -0.19 Openness 0.18 Agreeableness -0.18 Conscientiousness -0.18 Fixed Effects Industry & Year Industry & Year Industry & Year R-squared 2.91% 6.70% 8.48%

IA.1

Executive Extraversion: Career and Firm Outcomes T. Clifton Green, Russell Jame, and Brandon Lock*

Internet Appendix

This appendix consists of three parts. Section IA.1 provides methodological details for the

linguistic algorithms we use to measure executive extraversion. Section IA.2 further describes

our analysis which compares the linguistic algorithms to observer-assessed ratings of

extraversion. Section IA.3 presents additional analysis and robustness checks.

IA.1 Linguistic Algorithms

Researchers in psycholinguistics and artificial intelligence have developed personality

models based on linguistic outputs (e.g., Argamon et al., 2005; Oberlander and Nowson, 2006;

Mairesse et al., 2007). Developing these models involves four steps.1 First, a training dataset is

collected that contains written or spoken language of individuals along with the predictive

variables of interest (e.g., survey or observer ratings of personality). Second, a feature set is

chosen that defines the linguistic characteristics that will be used to develop the prediction

model. A common approach is to use Linguistic Inquiry and Word Count (LIWC) categories,

which includes both syntactic features (e.g., ratio of pronouns) and semantic information (e.g.,

positive emotion words) (Pennebaker et al., 2001).

In the third step, statistical learning algorithms are chosen to maximize out-of-sample

predictive power. Common supervised learning approaches include support vector machines and

decision trees. These algorithms take as input the observed features in the training corpus and

* Green is from Goizueta Business School, Emory University, clifton.green@emory.edu. Jame is from Gatton College of Business and Economics, University of Kentucky, russell.jame@uky.edu. Lock is from Kellogg School of Management, Northwestern University, b-lock@kellogg.northwestern.edu. 1 See Manning and Schütze (1999) and Hastie, Tibshirani, and Friedman (2009) for references on natural language processing and statistical learning.

IA.2

develop models according to a specified loss function. Lastly, models are evaluated and selected

based on their estimated out-of-sample predictive power. A common evaluation method is k-fold

cross validation, where the training corpus is randomly divided into k equal subsamples and each

subsample takes a turn acting as the out-of-sample data, with the k-1 remaining subsamples used

as the training data.

Mairesse et al. (2007) estimate personality ratings using training data from the

conversation extracts of 96 participants (97,468 words and 15,269 utterances) recorded using

Electronically Activated Recorders (EAR) (Mehl et al., 2001). Personality ratings (on a 7-point

scale) are obtained for each participant from 18 independent observers who have access to both

sound extracts and conversation transcripts. Mairesse et al. (2007) characterize each participant’s

communication style using 88 linguistic features from the LIWC database (Pennebaker, Francis,

and Booth, 2001) and 14 features from the MRC Psycholinguistic database (Coltheart, 1981).2

They provide personality recognition models fitted to the speech conversations using statistical

learning models that maximize out-of-sample predictive ability (10-fold cross-validation), with

feature counts standardized within sample. We average across the estimates of the four trained

models provided by Mairesse et al. (2007) to measure executive extraversion for our sample

conference call dialogue.3 The algorithms consist of two linear regression methods and two tree-

based approaches.

The four linguistic algorithms each implement regression models where the dependent

variable is the extraversion score of the individual and the explanatory variables are word

categories from the LIWC (Pennebaker et al., 2001) and MRC linguistic databases (Coltheart,

2 More information on the LIWC and MRC categories can be found at http://www.liwc.net/descriptiontable1.php and http://websites.psychology.uwa.edu.au/school/MRCDatabase/mrc2.html. 3 Our findings are qualitatively similar using any of the four algorithms. Trained model files for the Weka toolkit (Witten and Frank, 2005) are available at: http://farm2.user.srcf.net/research/personality/recognizer. All algorithms are fit using Weka default parameters.

IA.3

1981). Features include variables like the frequency of the use of different pronouns (i.e., “I”

versus “We”), the use of articles, the use of assent words (e.g., agree, ok, yes), the total number

of words used, the ratio of unique words to total words (i.e., the type/token ratio), the total

number of words used longer than six letters, the tendency to use negations, the tendency to use

positive or negative emotion words, the use of imagery words, etc.4 Since there are a large

number of LIWC and MRC features, the algorithms employ feature selection methods to select

informative linguistic categories. For example, the linear regression algorithm iteratively

removes features with the smallest standardized coefficients based on the Akaike information

criterion (AIC), while tree-based algorithms use a simplification procedure (“pruning”) to

mitigate over-fitting (see Appendix B). The algorithms use feature counts that are standardized

within the provided set of texts. This improves generalizability to language domains with

different linguistic distributions.

We describe each of the four estimation methodologies in more detail below.

IA1.1 Linear Regression Methods

IA1.1.1 Linear Regression (LR)

The linear regression model performs a least-squares regression with feature selection.

The selection algorithm iteratively removes features with the smallest standardized coefficient

until no improvement is observed with the error, as given by the Akaike information criterion

(AIC).

IA1.1.2 Support Vector Machine Regression (SVR)

4 A complete list of the LIWC and MRC features can be found in Table 6 of Mairesse et al. (2007).

IA.4

The support vector regression is based on the classical support vector machine classifier

that finds the hyperplane with the greatest separating margin between itself and the nearest data

point of each class. It bears the main features that characterize the maximal margin algorithm: a

(possibly non-linear) function is learned by a linear learning machine that reduces weight on

outliers, while the capacity of the system is controlled by a parameter that does not depend on the

dimensionality of the space.

Consider the problem of approximating the set of training data 𝐷 = {(𝑥𝑖, 𝑦𝑖)| =

1,2, . . . , 𝑁} of input vectors 𝑥𝑖 ∈ ℝ𝑝 and targets 𝑦𝑖 ∈ ℝ with a linear function,

𝑓(𝑥) = 𝑤 ⋅ 𝑥 + 𝑏

The most common form of support vector regression uses an ϵ-insensitive loss function (Vapnik

et al., 1997),5 which has the following form:

𝐿𝜖(𝑦) = {0 if |𝑓(𝑥) − 𝑦| < 𝜖|𝑓(𝑥) − 𝑦| − 𝜖, otherwise

The regression algorithm minimizes the norm ‖𝑤‖2 = 𝑤 ⋅ 𝑤 to fit as “flat” of a function as

possible. The optimal regression function is given by the following optimization problem.

min 12

‖𝑤‖2 + 𝐶 ∑(𝜉𝑖 + 𝜉𝑖∗)

𝑖

subject to {𝑦𝑖 − 𝑤 ⋅ 𝑥𝑖 − 𝑏 ≤ 𝜖 + 𝜉𝑖 𝑤 ⋅ 𝑥𝑖 + 𝑏 − 𝑦𝑖 ≤ 𝜖 + 𝜉𝑖

∗

𝜉𝑖 , 𝜉𝑖∗ ≥ 0

This is a convex optimization problem where 𝜉𝑖, 𝜉𝑖∗ are slack variables and the constant C

> 0 determines the trade-off between the flatness of f and the amount up to which deviations

larger than ϵ are tolerated. For computational reasons, the optimization problem is often solved 5 This loss function is closely related to the Huber (1964) loss function used in robust regression in statistics, which has optimal properties when the underlying distribution of the data is unknown. However, the ϵ-insensitive loss function is more computational tractable as it enables a sparse set of support vectors to be obtained.

IA.5

in its dual form (rather than the above primal form).6 Non-linear functions can also be estimated

by using a kernel function to map training data into a high-dimensional feature space, and then

performing a linear regression in this space.

IA.1.2 Tree-based Methods

IA.1.2.1 M5’ Model Tree (M5P)

The M5’ model tree, based on the M5 learning algorithm (Quinlan, 1992) 7, is part of a

broad class of tree-based methods that are known for their simplicity and efficiency when

dealing with domains with large numbers of features.8 A standard regression tree partitions the

feature space 𝑋 into disjoint regions 𝐷𝑖 and provides a fitted value 𝑘𝑖 = 𝐸(𝑦|𝑥 ∈ 𝐷𝑖) within each

region. This can be expressed as a simple additive model of the form,

𝑚(𝑥) = ∑ 𝑘𝑖 × 𝐼(𝑥 ∈ 𝐷𝑖)𝑖

where 𝐼(⋅) is an indicator function for whether an observed vector of features 𝑥 ∈ 𝑋 is in

partition 𝐷𝑖. To construct a tree, M5P follows a splitting procedure at each node that maximizes

the expected error reduction,

∆𝑒𝑟𝑟𝑜𝑟 = 𝑠𝑑(𝑇) − ∑|𝑇𝑗||𝑇|

× 𝑠𝑑(𝑇𝑗)𝑗

where 𝑇 is the set of training cases that reach the node, 𝑇1, 𝑇2, … are the sets that result from

splitting the node according to the chosen feature, and 𝑠𝑑(. ) is the standard deviation of the

dependent variables in the set.

6 See Smola and Scholkopf (2004) and Keerthi et al. (2001) for details on the sequential minimal optimization (SMO) algorithm, which is the implementation used to solve the SVM regression problem for our estimates. 7 The M5’ extension makes improvements to the original M5 algorithm such as modifying how enumerated attributes and missing values are treated. See Wang and Witten (1997) for details. 8 See Breiman et al. (1984) for a reference on classification and regression trees.

IA.6

Regression trees can be conveniently represented as tree diagrams where the path at each

node is a logical test on a feature that leads to a final value at the tree leaves. The M5P constructs

tree-based models, but whereas regression trees have constants in the tree leaves, the model tree

extends this to have leaf nodes containing regression models. In other words, the M5’ model tree

is analogous to piecewise linear functions.

After a tree has been constructed, it undergoes a simplification process called “pruning”

to mitigate overfitting. During this process, the algorithm examines each node from the bottom

up and selects as the final model for each node either the simplified linear model above or the

model subtree, depending on which has the lower estimated error. A final stage is to use a

“smoothing” procedure to reduce sharp discontinuities among the leaf node predictions. M5P

uses the smoothing calculation

𝑝′ =𝑛𝑝 + 𝑘𝑞

𝑛 + 𝑘

where 𝑝′ is the prediction passed up to the next higher node, 𝑝 is the prediction passed to this

node from below, 𝑞 is the value predicted at this node, 𝑛 is the number of training instances that

reach the node below, and 𝑘 is a smoothing constant.

IA.1.2.2 M5’ Regression Tree (M5R)

The M5’ regression tree is a special case of M5P that restricts the leaf nodes to be

constant values as in traditional regression trees. It shares the same splitting, pruning, and

smoothing procedures as the M5P model.

IA.2 Comparing Linguistic Algorithms to Observer-Assessed Ratings of Extraversion

We help bolster the validity of our linguistic measure by comparing our textual algorithm

to observer assessments for a subset of the sample. In particular, we gather audio excerpts for

IA.7

100 conference calls from Earningscast.com. We choose calls for the 50 most and least

extraverted CEOs in our initial sample that have conference call audio files available from

Earningscast. For each CEO, we select the most recent call available during our sample period,

and from within each call we create an audio excerpt from the CEOs’ response to the first

question during the Q&A portion of the call. We require CEOs to speak at least 25 words during

the response, and we truncate longer audio excerpts at roughly one minute (e.g., near the end of a

sentence, etc.). The average audio response length is 42 seconds (minimum of 9 seconds and

maximum of 64 seconds) and contains 112 spoken words.

We ask BBA students to evaluate the level of extraversion for each executive using a 1 to

7 scale, based on the audio excerpt. Observers also have access to the written transcripts of the

responses to the questions. We inform the evaluators that the audio responses are “executives

responding to a question during earnings conference calls (these are public Q&A sessions that

firms hold after announcing earnings),” and we provide them with the following definition of

extraversion from Wikipedia:

Extraversion is the act, state, or habit of being predominantly concerned with obtaining gratification from what is outside the self. Extraverts tend to enjoy human interactions and to be enthusiastic, talkative, assertive, and gregarious. Extraverts are energized and thrive off of being around other people. They take pleasure in activities that involve large social gatherings, such as parties, community activities, public demonstrations, and business or political groups. They also tend to work well in groups. An extraverted person is likely to enjoy time spent with people and find less reward in time spent alone. They tend to be energized when around other people, and they are more prone to boredom when they are by themselves.9

Each student evaluated 20 executives (10 extraverts and 10 introverts in random order), and we

obtained three separate evaluations for each executive. Table IA.1 reports examples of executive

responses rated most and least extraverted by the observers with links to corresponding audio

9 https://en.wikipedia.org/wiki/Extraversion_and_introversion

IA.8

files, along with whether the linguistic algorithm classifies the executive as extraverted or

introverted.

IA.3 Additional Analysis and Robustness Evidence

The remainder of the internet appendix tabulates a large number of robustness tests

discussed in the paper.

x Table IA.2 tabulates the results discussed in Section 3.2 of the text.

x Table IA.3 tabulates the results discussed in footnote 20 of Section 3.3.

x Table IA.4 tabulates the results discussed in Section 4.2

x Table IA 5. tabulates the results discussed in Section 4.5

x Table IA.6 tabulates the results discussed in Section 5.3

x Tables IA.7 – IA.13 present expanded versions (with t-statistics for all control variables) of tables 4, 5, 6, 7, 8, 10, and 11, respectively.

IA.9

Internet Appendix References: Argamon, S., Dhawle, S., Koppel, M., Pennebaker, J., 2005. Lexical predictors of personality

type. In Proceedings of the Joint Annual Meeting of the Interface and the Classification Society of North America.

Breiman, L., Friedman, J., Stone, C. J., Olshen, R. A., 1984. Classification and regression trees.

CRC press. Coltheart, M., 1981. The MRC psycholinguistic database. Quarterly Journal of Experimental

Psychology 33, 497–505. Hastie, T., Tibshirani, R., Friedman, J., Hastie, T., Friedman, J., Tibshirani, R., 2009. The

elements of statistical learning (Vol. 2, No. 1). New York: Springer. Huber, P. J. 1964. Robust estimation of a location parameter. The Annals of Mathematical

Statistics 35, 73-101. Keerthi, S., Shevade, S. K., Bhattacharyya, C., & Murthy, K. R. K., 2001. Improvements to

Platt's SMO algorithm for SVM classifier design. Neural Computation 13, 637-649. Mairesse, F., Walker, M. A., Mehl, M. R., Moore, R.K., 2007. Using linguistic cues for the

automatic recognition of personality in conversation and text. Journal of Artificial Intelligence Research 30, 457-500.

Manning, C. D., Schütze, H., 1999. Foundations of statistical natural language processing. MIT

press. Mehl, M. R., Pennebaker, J., Cros, M., Dabbs, J., and Price, J., 2001. The Electronically

Activated Recorder (EAR): A device for sampling naturalistic daily activities and conversation. Behavior Research Methods, Instruments, and Computers 33, 517-523.

Oberlander, J., Nowson, S., 2006. Whose thumb is it anyway? Classifying author personality

from weblog text. In Proceedings of the 44th Annual Meeting of the Association for Computational Linguistics (ACL).

Pennebaker, J. W., Francis, M. E., Booth, R. J., 2001. Inquiry and word count: LIWC. Lawrence

Erlbaum, Mahwah, NJ. Quinlan, J. R., 1992. Learning with continuous classes. In Proceedings of the 5th Australian

joint Conference on Artificial Intelligence 92, 343-348. Smola, A. J., Schölkopf, B., 2004. A tutorial on support vector regression. Statistics and

Computing 14, 199-222.

IA.10

Witten, I. H., Frank, E., 2005. Data Mining: Practical machine learning tools and techniques. Morgan Kaufmann.

IA.11

Table IA.1: Transcripts and Audio Excerpts of Extraverted and Introverted CEOs

The table reports examples of CEO responses to the first question during the Q&A portion of a conference call with links to corresponding audio files. Panel A lists the five executive responses deemed most extraverted by human listeners (due to ties in listener rankings, we report responses for eight executives). Panel B lists the five executive responses deemed least extraverted by human listeners (due to ties in the rankings, we report responses for six executives). The last column reports whether the executive is classified as extraverted or introverted based on the average of four linguistic algorithms described in Appendix B. An executive is classified as extraverted (introverted) if the algorithm score (computed using the text from the first column) is in the top (bottom) half of the distribution.

Panel A: Listener-rated Extraverts

Text Algorithm

Classification Joe, to change the subject slightly, we borrowed $3.1 billion, and if we were to take it all down today, we would be paying $84 million in interest for $3.1 billion. I personally never thought I'd live to see something like that. It's a wonderful time to borrow money. It's a shame that we're all done doing it. It's been such fun, and it's been so terrific for the company's balance sheet and for our future. We're very proud of that. Matt, Dr. Maddox, has done a remarkable job, and we're going to refer to you as Dr. Maddox from now on. Having done this skilled surgery on our balance sheet, we should call you Dr. Maddox. Next question, please. https://sites.google.com/site/ceoextraversion/extravert1

Extravert

I think it's applying across all product and price lines, to be frank with you. And I don't mean to date myself, but as I sit around and look at the table here, no one around this table can remember mortgage rates being higher than 6% or 7%. And I think one of the factors that we are dealing with, quite frankly, is most analysts and most young buyers, especially first time home buyers in the market today, have been accustomed to low rates for all their lives. So I do know one thing: rates are going to go up. And we're going to have to deal with those on a go-forward basis, but I think rates going up will indicate that the economy is doing better; we're creating jobs and I think that will be good for the economy and it'll be good for us. https://sites.google.com/site/ceoextraversion/extravert2

Extravert

Yeah. Thank you, Peter. First and foremost, we have a very strong and deep bench. And as you know, over the years, we've taken pride in moving our leaders into functional roles and different business roles to round out their experience and get them ready for bigger opportunities. And so as we approach this year, Kiran and I have been speaking for the last couple years about his personal true north. And at that point in time, we decided and Kiran was very interested in talking about what he wanted to do in the next chapter. And our hope was that that next chapter was to be with Intuit forever and ever. But at some point he reached the point where he said, it's time for me to retire. So we celebrated his birthday a week ago and he said, now's the time. And it set up the opportunity for us to make some key organizational changes. In the case of Sasan Goodarzi, which is where your specific question was, Sasan has actually been the general manager of many of many of our businesses. He led our verticals businesses several years ago in the early 2000s. https://sites.google.com/site/ceoextraversion/extravert3

Extravert

And, Bob, this is Scott. Our Chief Banking Officer, Dave Kennedy, he has continued throughout 2012 and 2013 to look at branch profitability and his team that runs our entire network. They're very focused on making sure that each office is contributing. But as you know, with a low interest rate environment and once again with a sluggish economy, we're going to continue to look for more contribution and more leverage out of that network. https://sites.google.com/site/ceoextraversion/extravert4

Introvert

Sure. Thanks, George. Let me start by talking about Roofing. I would tell you, I guess I've been with the company now 21 years so I've been around the analysis of our results for a long time. This was really the first quarter that we dug into kind of state-by-state shipment data and state-by-state market share data at the level we did because we saw some trends in the quarter that did surprise us in terms of the size of the volume swings that we saw in some parts of the country. So at least in my history I would say this was probably the biggest change in the overall

Extravert

IA.12

geographic mix of the market. Now I think this is a bit of an outgrowth of the change we've seen in the overall market over the course of the last four or five years where we are seeing more inventory put into the market early in the year, and then we're seeing distribution customers primarily manage that inventory through the summer. https://sites.google.com/site/ceoextraversion/extravert5 So let me give you a little more detail on that, John. So what we said was the growth of specific products, so some of it was Network Performance Monitor, which is still a product that's – it's a largest product we have in our portfolio, some of it was Network Traffic Analyzer, some of it was Network Configuration Monitor, really the three flagship products. And our network management product portfolio grew faster than they have at any time in the last six quarters. https://sites.google.com/site/ceoextraversion/extravert6

Introvert

Yeah. It's a fact that I was requested and I went and met with the Minister in August, and I told him and had the lawyers present and had all of his lawyers and I basically said, hey, you know the situation is that it's not our plant and it's not our workers. They're part of the CGT. So the first thing is, I'm not talking to the CGT until you've arranged that they and Goodyear at least come into some sort of a tentative agreement and then if Goodyear brings us into it, then fine. And so then he talked and he wanted to know, he appreciated that. And so we started talking about – he wanted to know, well, how many people. https://sites.google.com/site/ceoextraversion/extravert7

Extravert

Well, that's quite a philosophical question, Michelle. I'll just give you a couple of thoughts; not sure I'll do it justice. But first and foremost, I think our company is still here because we've had very, very strong people in the leadership of the company. I'll admit myself in that. But I think we've had good people that the leadership of the company who thought long, and who were appropriately conservative at appropriate times. We believe in funding our pension plan. We believed in investing in our facilities, and that has stood us well over time. We do have some legacy benefits that we carry, certainly our iron ore resources would be among those. We have some legacy obligations we carry as well, but there're things that we've tried to treat seriously and deal with properly and do the right thing to insure our long-term vigor and long-term prospects for the company. https://sites.google.com/site/ceoextraversion/extravert8

Extravert

Panel B: Listener-rated Introverts

Text Algorithm

Classification Kevin, I'd say that yes, the increased spending is what's driving our improvement. A lot of the increased spending has been focused on the international markets, the areas where we – we've traditionally had less presence, and I think that's paying off. It's helping us in traditional markets as well as to broaden our product line. So, we're pleased with the progress. It's – the increased spending is necessary in this market, because the demands being placed on our customers' products continue to increase. https://sites.google.com/site/ceoextraversion/introvert1

Introvert

I think the margins that we're recognizing now are probably a pretty good indicator of what you should expect independent of significant changes in the market. https://sites.google.com/site/ceoextraversion/introvert2

Introvert

Heather, it's a bit early, the state estimate just was announced last week, which calls for the crop to come out about 4% smaller than last year, but will be the second-largest crop in history. The crop harvest is going to be very late, which is an issue because that reduces the amount of the fall season in which to market the crop and the currency versus last year dollars are bit stronger. It's still too early to make any judgments. So, we'll be watching that market closely, and we'll try to give more color next quarter on that. https://sites.google.com/site/ceoextraversion/introvert3

Introvert

Yes, that's something that we're considering and it could be a possibility. We're going to continue to evaluate those opportunities and if something that makes sense comes along, we won't be afraid to pull the trigger. https://sites.google.com/site/ceoextraversion/introvert4

Introvert

Okay, let me try to answer your question, Louise. First of all, I think when we say the FDA did Introvert

IA.13

not give us any additional enforcement action in the meeting and that's what we mean exactly, what we say in the press release. In the meeting, that they did not mention anything about additional enforcement action. Now, of course, you understand that this does not guarantee the FDA will not give us any interaction in the future, okay? And second question on the 2014, move up to 2013. I think Bryan might be able to give you details, but at this point, we are moving urgently as we can. We have moved a lot of people both internally and externally try to accelerate the program as much as possible. https://sites.google.com/site/ceoextraversion/introvert5 I think you better look at two pieces of that and maybe the smaller piece is the aftermarket. But we have seen in the U.S. the early stages of correction are fundamentally going to over correction to get parts inventories down and to stretch out some of the rebuilds and then as things settled in those aftermarket order rates begin to move back up to a level that's in line with the production change. So, what we're seeing in the U.S. is the bottoming that is starting to see modest quarter-over-quarter improvement. As we go through the corrections in Australia and China, we're seeing the drop at the front-end and so that's becoming a drag on our third quarter bookings for aftermarket. https://sites.google.com/site/ceoextraversion/introvert6

Introvert

IA.14

Table IA.2 Persistence in Conference Call Extraversion: Firm vs. Managers

This table reports the estimates of tests of persistence in extraversion across conference calls. Extraversion of CEOs (CFOs) is estimated using the residuals from Specification 1 (4) of Table 2. Column 1 of Panel A reports the correlation between a CEO’s extraversion score based on conference calls over the 2006-2009 period and their extraversion score based on conference calls over the 2010-2013 period. Column 2 reports the correlation in the extraversion score of two different CEOs working for the same company. Column 3 tests whether the difference in the correlation between columns 1 and 2 is statistically significant. Panel B reports analogous results for CFOs. T-statistics are reported below each estimate. Panel A: CEOs

Same Manager, Same Firm Different Manager, Same Firm Difference

[1] [2] [1] -[2]

ρ(Extraversiont, Extraversiont+1) 0.75 0.25 0.48

(38.14) (6.44) (12.37)

Obs. 1145 609 1772 Panel B: CFOs

Same Manager, Same Firm Different Manager, Same Firm Difference

[1] [2] [1] -[2]

ρ(Extraversiont, Extraversiont+1) 0.63 0.30 0.34

(28.47) (8.75) (8.59)

Obs. 1209 782 1991

IA.15

Table IA.3: Determinants of Hiring Extraverted Executives This table reports the estimates of the following regression:

Extraversioni = β1Ln(Sales)it-1 + β2Ln(Q)it-1 + β3Ln(Firm Age)it-1 + β4Ln(Vol)it-1 + Indi + εit. Extraversion is the residual extraversion of CEOs (CFOs) based on Specification 1 (3) of Table 2. Definitions of the independent variables are provided in Appendix A. INDi captures industry fixed effects as measured using the Fama and French (1997) 12 industry classification. All continuous variables are standardized to have mean 0 and variance 1 each year. Standard errors are clustered by firm, and t-statistics are reported below each estimate.

CEOs CFOs

[1] [2]

[1] [2]

Ln (Sales)t-1 0.27 0.24

0.35 0.36

(7.36) (6.29)

(12.67) (12.93)

Ln (Q)t-1 0.21 0.16

0.06 0.06

(6.35) (4.71)

(2.54) (2.64)

Ln (Vol)t-1 0.14 0.06

0.00 0.00

(3.88) (1.44)

(-0.08) (-0.00)

Ln (Firm Age)t-1 -0.03 -0.02

-0.05 -0.03

(-0.82) (-0.53)

(-2.08) (-1.11)

Non-Durables

-0.02

-0.08

(-0.13)

(-1.01)

Durables

-0.15

0.10

(-0.90)

(1.06)

Manufacturing

0.10

-0.22

(1.48)

(-3.55)

Energy, Oil, Gas, and Coal

0.21

-0.31

(1.38)

(-3.09)

Chemicals

0.22

-0.11

(1.71)

(-1.10)

Computers, Software, Electronics

0.10

0.05

(1.53)

(0.98)

Telecom

0.20

0.17

(0.87)

(0.88)

Utilities

-0.73

-0.15

(-5.24)

(-1.73)

Wholesale and Retail Shops

0.04

0.04

(0.54)

(0.59)

Healthcare

0.04

0.11

(0.36)

(1.21)

Financials

-0.09

0.29

(-0.85)

(3.44)

Other

0.09

0.06

(1.02)

(0.77)

Obs. 1,121 1,121

1,851 1,851 R2 7.90% 11.62%

10.98% 13.15%

IA.16

Table IA.4: Extraversion and Executive Compensation: Additional Analysis This table reports estimates from the following panel regression:

𝐿𝑜𝑔(𝐶𝑜𝑚𝑝)𝑖𝑡 = 𝛽1𝐸𝑥𝑡𝑟𝑎𝑣𝑒𝑟𝑠𝑖𝑜𝑛𝑖 + 𝛾𝑭𝒊𝒓𝒎𝑪𝒉𝒂𝒓 + 𝛿𝑷𝒆𝒓𝒇𝒐𝒓𝒎𝒂𝒏𝒄𝒆 + ωExecChar + 𝑌𝑒𝑎𝑟𝑡 + 𝐹𝐸𝑖 + 𝜀𝑖𝑡.. Comp captures a variety of executive compensation measures. Extraversion is the residual extraversion of executives based on Specification 1 of Table 2. FirmChar is a vector of firm characteristics, Performance is a vector of firm performance measures, and ExecChar is a vector of individual characteristics. In Panel A, Comp is CEO Cash Compensation defined as salary + bonus. In Panel B, Comp is CEO Equity Compensation defined as Total Compensation – Cash Comp. In Panel C, Comp is based on CEO total compensation, comprised of salary, bonus, value of restriction stock granted, value of options granted, long-term incentive payout, and other compensation (TDC1 as reported in Execucomp), and Extraversion is measured using quintiles. In Panel D, Comp is based on CFO total compensation and Extraversion is residual extraversion for CFOs. In Panel E, Comp is CEO Relative Pay defined as Log (CEO total compensation) – Log (CFO compensation), and Extraversion is Relative Extraversion defined as CEO Extraversion less CFO Extraversion. All continuous variables are standardized to have mean 0 and variance 1. Appendix A provides detailed variable definitions. The layout is the same as in Table 4 in the text, although only the coefficients on Extraversion are reported for brevity. In particular, Specifications 1&5 examine Extraversion in isolation, 2&6 add firm characteristics, 3&7 add performance measurements, and 4&8 add manager characteristics. Specifications 1-4 (5-8) include industry (firm) and year fixed effects. Industry and Year Fixed Effects Firm and Year Fixed Effects [1] [2] [3] [4]

[5] [6] [7] [8]

Panel A: CEO Cash Compensation Extraversion 7.93% 2.93% 2.81% 3.92%

2.21% 2.11% 2.00% 3.86%

(6.21) (2.91) (2.81) (3.39)

(1.55) (1.51) (1.44) (2.58)

Panel B: CEO Equity Compensation Extraversion 24.73% 9.10% 8.54% 8.24% 11.68% 10.92% 11.60% 13.84% (7.24) (3.56) (3.37) (2.82) (2.67) (2.46) (2.61) (2.73) Panel C: CEO Total Compensation Quintiles Extraversion Q2 22.05% 8.78% 7.72% 6.74% 3.61% 2.51% 3.95% 4.58%

(3.45) (2.41) (2.18) (1.92) (0.75) (0.55) (0.87) (1.01) Extraversion Q3 29.20% 16.51% 15.17% 14.19% 6.26% 5.53% 7.15% 9.45%

(4.71) (4.41) (4.21) (3.92) (1.28) (1.17) (1.52) (2.02) Extraversion Q4 41.19% 17.30% 15.93% 15.71% 9.49% 9.02% 10.60% 13.90%

(6.40) (4.28) (3.99) (3.77)

(1.67) (1.63) (1.96) (2.56)

Extraversion Q5 52.70% 19.27% 18.08% 17.67%

16.57% 13.99% 13.83% 16.29%

(7.84) (4.73) (4.53) (3.94)

(2.43) (2.06) (2.08) (2.36)

Panel D: CFO Total Compensation Extraversion 25.94% 6.86% 6.61% 5.18% 11.35% 11.26% 11.15% 9.07%

(14.61) (5.72) (5.59) (4.32) (6.55) (6.62) (6.66) (5.32) Panel E: CEO Relative Compensation Relative Extraversion 7.53% 8.79% 8.76% 7.88% 10.73% 10.77% 10.72% 9.60%

(6.17) (7.41) (7.39) (6.17) (6.60) (6.62) (6.60) (5.56)

IA.17

Table IA.5 Extraversion and CEO Turnover: The Role of Firm Performance

This table reports estimates from the following panel regression:

𝐶𝐸𝑂 𝑇𝑛𝑖𝑡 = 𝛽1𝐸𝑥𝑡𝑟𝑎𝑣𝑒𝑟𝑠𝑖𝑜𝑛𝑖 + 𝛾𝑭𝒊𝒓𝒎𝑪𝒉𝒂𝒓 + 𝛿𝑷𝒆𝒓𝒇𝒐𝒓𝒎𝒂𝒏𝒄𝒆 + 𝝎𝑪𝑬𝑶𝑪𝒉𝒂𝒓 + 𝑌𝑒𝑎𝑟𝑡 + 𝐹𝐸𝑖 + 𝜀𝑖𝑡. 𝐶𝐸𝑂 𝑇𝑛𝑖𝑡 denotes CEO Turnover and is 1 if firm i changes its CEO in year t, and 0 otherwise. Extraversion is the residual extraversion of CEOs based on Specification 1 of Table 2. FirmChar, Performance, and CEOChar are vectors of firm, performance, and manager characteristics detailed in Appendix A. FE indicates industry fixed effects. Each year, the sample is split into firms with above and below median stock returns (Specifications 1 and 2) or above and below median profitability (Specifications 3 and 4) over the previous year. All independent variables are standardized to have mean 0 and variance 1. Standard errors are clustered by firm.

High Returns Low Returns

High Prof. Low Prof.

[1] [2]

[3] [4]

Extraversion 0.96 0.80

0.91 0.82

(-0.34) (-2.28)

(-0.82) (-1.97)

Ln (Sales) 1.00 0.73

0.77 0.86

(-0.01) (-1.59)

(-0.85) (-0.87)

Ln (Assets) 1.03 2.06

1.40 1.71

(0.14) (3.24)

(1.00) (2.72)

Ln (Q) 1.08 1.47

1.18 1.43

(0.60) (4.32)

(1.40) (3.41)

Ln (Vol) 0.91 1.05

0.94 1.01

(-0.80) (0.49)

(-0.52) (0.09)

Lag (Age) 0.98 0.92

0.89 0.96

(-0.14) (-0.91)

(-1.07) (-0.46)

Ln (Sales Growth) 1.04 0.98

0.96 1.01

(0.37) (-0.30)

(-0.33) (0.16)

Fiscal Ret 0.96 1.27

0.87 0.91

(-0.46) (0.89)

(-1.08) (-0.95)

Lag Fiscal Ret 0.98 0.85

0.88 0.98

(-0.18) (-1.22)

(-0.81) (-0.23)

Prof 1.08 1.10

1.14 1.25

(0.60) (0.91)

(1.02) (1.29)

Prof. Growth 0.89 0.96

0.90 0.92

(-1.30) (-0.67)

(-1.34) (-1.18)

Loss Dummy 0.72 1.15

0.88 1.01

(-1.01) (0.74)

(-0.38) (0.04)

Log (CEO Age) 1.57 1.56

1.57 1.54

(4.20) (5.29)

(4.19) (4.74)

LN (CEO Tenure) 1.27 1.57

1.44 1.59

(2.07) (4.69)

(3.03) (4.75)

Male 4.07 0.83

1.17 1.14

(1.39) (-0.57)

(0.27) (0.38)

Founder 0.87 0.56

0.91 0.48

(-0.48) (-2.39)

(-0.35) (-2.88)

IA.18

Chair 1.08 0.49

0.66 0.66

(0.42) (-4.49)

(-2.29) (-2.43)

GAI 1.19 1.22

1.45 1.08

(1.99) (2.81)

(4.25) (1.11)

Rolodex 1.13 1.05

1.13 1.08

(1.37) (0.65)

(1.32) (1.10)

Percent Ceo Text 1.06 1.10

1.21 1.02

(0.42) (0.79)

(1.26) (0.16)

Optimism 0.99 0.92

1.09 0.86

(-0.14) (-1.02)

(0.88) (-1.88)

Overconfidence 0.89 0.83

0.95 0.79

(-0.53) (-1.00)

(-0.23) (-1.19)

MBA 1.03 0.97

0.92 0.97

(0.18) (-0.19)

(-0.42) (-0.22)

Doctorate 1.69 1.40

1.82 1.33

(0.96) (0.92)

(0.82) (1.04)

Ivy League 1.00 0.78

0.98 0.81

(-0.03) (-1.80)

(-0.15) (-1.66)

Grad with Honors 0.64 0.85

0.71 0.85

(-1.63) (-0.78)

(-1.18) (-0.78)

Emotional Stability 0.88 0.81

0.87 0.80

(-1.19) (-2.27)

(-1.28) (-2.41)

Openness 1.11 1.23

1.25 1.13

(0.88) (2.17)

(1.84) (1.35)

Agreeableness 1.11 1.04

0.92 1.18

(0.87) (0.49)

(-0.70) (1.72)

Conscientiousness 1.04 0.88

0.93 0.96

(0.33) (-1.27)

(-0.51) (-0.37)

Fixed Effects Industry & Year Industry & Year

Industry & Year Industry & Year R-squared 10.59% 14.91%

13.06% 13.06%

Observations 5,465 5,465

5,459 5,466 Unconditional Prob. 3.93% 6.11%

4.30% 5.71%

Y =1 215 334

235 312

IA.19

Table IA.6 CEO Extraversion and M&A Activity

This table reports estimates from the following logistic panel regression: 𝑀𝑁𝐴𝑖𝑡 = 𝛽1𝐸𝑥𝑡𝑟𝑎𝑣𝑒𝑟𝑠𝑖𝑜𝑛𝑖𝑡 + 𝜷𝟐𝑭𝒊𝒓𝒎𝑪𝒉𝒂𝒓𝑖𝑡 + 𝜷𝟑𝑪𝑬𝑶𝑪𝒉𝒂𝒓𝑖𝑡 + 𝐼𝑛𝑑𝑖 + 𝑌𝑒𝑎𝑟𝑡 + 𝜀𝑖𝑡 .

In Specifications 1 and 2, MNAit is a dummy variable if firm i conducted a merger or acquisition in year t. In Specifications 3 and 4, MNAit is the total value of acquisitions scaled by total assets (in natural logs). Extraversion is the extraversion score of the CEO of the acquiring firm. FirmChar and CEOChar are the vectors of firm and CEO characteristics included as controls in Equation 3 (and described in Appendix A). All independent variables are standardized to have mean 0 and variance equal to 1. Specifications 1 and 2 report odds ratios and z-scores from a logit regression, and Specifications 3 and 4 report coefficients and t-statistics. Standard errors are clustered by firm. M&A Dummy Scaled M&A [1] [2] [3] [4] Extraversion 1.13 1.07

3.20% 2.53%

(3.08) (1.56)

(2.91) (1.91)

Ln (Sales)

0.89

-5.29%

(-1.12)

(-1.91)

Ln (Assets)

1.38

4.18%

(2.87)

(1.47)

Ln (Q)

1.00

0.98%

(0.09)

(0.77)

Ln (Vol)

0.76

-4.19%

(-5.90)

(-3.27)

Ln (Age)

0.86

-4.47%

(-3.30)

(-3.32)

Lag Fiscal Ret

0.99

0.68%

(-0.24)

(0.72)

Log (CEO Tenure)

1.10

3.06%

(2.01)

(2.33)

Log (CEO Age)

0.91

-2.31%

(-2.17)

(-1.79)

Male

1.17

-1.62%

(0.70)

(-0.31)

Founder

0.93

-2.05%

(-0.59)

(-0.55)

Chair

0.93

-1.83%

(-0.96)

(-0.82)

GAI

1.06

3.00%

(1.41)

(2.91)

Rolodex

1.02

-0.84%

(0.38)

(-0.65)

Percent Ceo Text

0.93

-2.14%

(-1.18)

(-1.42)

Optimism

1.18

3.86%

(4.18)

(3.67)

Overconfidence

0.94

-2.83%

(-0.74)

(-1.03)

MBA

1.05

1.34%

(0.61)

(0.63)

Doctorate

1.20

13.83%

(0.67)

(1.19)

IvyLeague

0.93

-2.06%

IA.20

(-1.05)

(-1.03)

GradHonors

1.07

0.91%

(0.65)

(0.30)

Emotional Stability

1.01

0.73%

(0.24)

(0.57)

Openness

1.01

-0.10%

(0.24)

(-0.08)

Agreeableness

1.03

0.86%

(0.49)

(0.62)

Conscientiousness

0.95

-1.33%

(-0.96)

(-0.92)

Fixed Effects Industry & Year Industry & Year

Industry & Year Industry & Year Observations 10,925 10,925

10,925 10,925

Pseudo R2/ R2 4.59% 7.76%

5.07% 6.03%

IA.21

Table IA.7 Extraversion and CEO Compensation

(same as Table 4 in the text, with test-statistics reported for the control variables) Industry and Year Fixed Effects Firm and Year Fixed Effects [1] [2] [3] [4]

[5] [6] [7] [8]

Extraversion 17.90% 6.10% 5.70% 5.70%

6.32% 5.74% 6.00% 7.57%

(8.41) (4.53) (4.27) (3.65)

(2.98) (2.70) (2.83) (3.35)

Ln (Sales)

25.40% 20.40% 18.50%

5.21% 17.64% 17.87%

(7.34) (5.78) (5.22)

(1.12) (2.36) (2.41)

Ln (Assets)

48.70% 55.20% 50.60%

34.15% 36.84% 36.31%

(13.29) (14.31) (13.45)

(6.80) (6.07) (5.98)

Ln (Q)

17.60% 12.70% 11.60%

15.93% 11.78% 11.56%

(10.91) (6.73) (6.26)

(12.35) (7.62) (7.50)

Ln (Vol)

3.60% 3.30% 3.40%

-2.03% -0.94% -1.09%

(2.68) (2.46) (2.55)

(-1.29) (-0.61) (-0.73)

Ln (Firm Age)

-2.20% -1.60% -1.40%

7.75% 7.84% 5.76%

(-1.82) (-1.32) (-0.91)

(2.27) (2.30) (1.62)

Ln (Sales Growth)

5.60% 5.50%

5.05% 4.99%

(4.86) (4.76)

(4.05) (4.01)

Fiscal Ret

5.30% 4.90%

4.48% 4.45%

(4.60) (4.45)

(4.43) (4.39)

Lag Fiscal Ret

4.10% 4.00%

3.38% 3.31%

(4.64) (4.62)

(4.20) (4.14)

Profitability

3.40% 3.40%

3.98% 3.99%

(1.98) (2.02)

(2.78) (2.78)

Prof. Growth

-0.20% -0.20%

-0.51% -0.48%

(-0.28) (-0.24)

(-0.66) (-0.62)

Loss Dummy

-4.90% -4.70%

-11.14% -10.63%

(-1.90) (-1.82)

(-5.74) (-5.45)

Log (CEO Tenure)

-3.30%

2.56%

(-1.98)

(1.79)

Log (CEO Age)

1.70%

-2.80%

(1.18)

(-1.47)

Male

2.80%

5.24%

(0.48)

(0.81)

Founder

0.40%

-5.93%

(0.09)

(-0.97)

Chair

7.40%

3.12%

(2.91)

(1.55)

GAI

7.60%

5.78%

(5.41)

(3.35)

Rolodex

2.00%

0.75%

(1.29)

(0.45)

Percent CEO Text

-0.20%

-0.07%

(-0.18)

(-0.06)

Optimism

4.10%

1.16%

(3.11)

(0.66)

Overconfidence

-6.60%

-3.56%

IA.22

(-2.68)

(-2.10)

MBA

2.80%

5.03%

(1.23)

(1.57)

Doctorate

12.50%

-9.28%

(1.45)

(-1.46)

Ivy League

2.00%

-0.68%

(1.00)

(-0.24)

Grad with Honors

4.80%

0.24%

(1.25)

(0.06)

Emotional Stability

1.80%

-2.66%

(1.14)

(-1.41)

Openness

-0.70%

-2.02%

(-0.41)

(-0.90)

Agreeableness

-1.30%

-0.83%

(-0.76)

(-0.39)

Conscientiousness

0.70%

-0.51%

(0.47)

(-0.18)

R-squared 8.95% 55.24% 57.13% 58.60%

79.41% 80.29% 81.06% 81.26% Observations 10,925 10,925 10,925 10,925

10,925 10,925 10,925 10,925

IA.23

Table IA.8 Extraversion and Age of First Appointment to CEO or CFO

(same as Table 5 in the text, with test-statistics reported for the control variables) All Hires External Hires Internal Hires

[1] [2] [3] [4]

Extraversion -0.79 -1.12 -0.98 -0.65

(-4.67) (-3.28) (-3.40) (-2.96)

Ln (Sales) 0.36 1.32 0.20 0.34

(1.37) (1.44) (0.49) (0.96)

Ln (Assets) -0.01 -0.21 0.06 -0.08

(-0.04) (-0.23) (0.14) (-0.23)

Ln (Q) -0.81 -1.26 -0.44 -1.10

(-2.48) (-1.48) (-0.71) (-2.35)

Ln (Vol) 0.20 1.35 0.16 0.45

(0.58) (2.28) (0.28) (1.01)

Ln (Firm Age) 0.81 -1.29 1.11 0.95

(4.00) (-1.44) (3.36) (3.10)

Male 1.85 0.33 1.93 2.42

(3.28) (0.85) (1.83) (3.44)

Founder 3.54 -0.27 3.12

(3.97) (-0.12) (1.79)

GAI 1.30 1.39 1.72 0.80 (8.60) (4.72) (7.61) (3.53) Rolodex 0.00 -0.10 -0.24 0.18

(-0.02) (-0.30) (-1.09) (0.72)

Optimism -0.78 -0.56 -1.14 -0.49

(-4.87) (-1.46) (-4.10) (-2.21)

Overconfidence -0.99 -0.17 -0.03 -0.42 (-1.97) (-0.77) (-0.03) (-0.63) MBA -0.46 -0.92 -1.57 0.11 (-1.57) (-1.68) (-2.97) (0.27) Doctorate -0.13 -2.48 -2.31 -0.71 (-0.11) (-1.20) (-0.79) (-0.48) Ivy League -0.65 -1.06 -0.12 -0.71 (-2.59) (-2.18) (-0.29) (-1.89) Grad with Honors -0.90 -0.51 -0.39 -1.18 (-2.07) (-0.55) (-0.55) (-2.26) Emotional Stability -0.49 -0.24 -0.76 -0.58 (-2.71) (-0.63) (-2.34) (-2.43) Openness 0.23 0.37 0.21 0.31 (1.15) (0.89) (0.62) (1.10) Agreeableness -0.27 -0.08 0.30 -0.75 (-1.41) (-0.20) (0.94) (-2.83) Conscientiousness -0.48 -1.22 -0.82 -0.19 (-2.22) (-2.85) (-2.10) (-0.67) Fixed Effects Industry & Year Firm & Year Industry & Year Industry & Year R-squared 27.50% 82.76% 32.22% 22.01% Observations 1,773 1,773 569 963

IA.24

Table IA.9

Extraversion and CEO Tenure and Turnover (same as Table 6 in the text, with test-statistics reported for the control variables)

Turnover Tenure

[1] [2] [3]

[4] [5] [6]

Extraversion 0.81 0.78 0.88

5.15 7.68 5.72

(-4.40) (-4.94) (-1.75)

(2.27) (3.36) (3.15)

Ln (Sales)

1.04 0.83

-5.86 -10.88

(0.35) (-1.27)

(-1.05) (-2.63)

Ln (Assets)

1.23 1.56

-12.35 -11.73

(1.61) (2.72)

(-2.09) (-2.63)

Ln (Q)

1.11 1.29

-2.97 -2.95

(1.47) (3.41)

(-1.22) (-1.55)

Ln (Vol)

1.13 0.99

-0.88 2.91

(2.02) (-0.11)

(-0.41) (1.87)

Lag (Age)

1.08 0.93

6.02 21.88

(1.65) (-0.99)

(3.25) (12.79)

Ln (Sales Growth)

1.00 1.00

5.12 2.55

(0.02) (-0.05)

(4.33) (2.90)

Fiscal Return

0.73 0.90

-2.79 -1.42

(-3.43) (-1.42)

(-2.52) (-1.79)

Lag Fiscal Return

0.85 0.92

-0.02 -0.09

(-1.20) (-0.85)

(-0.03) (-0.13)

Profitability

1.02 1.07

0.31 -0.52

(0.23) (0.89)

(0.13) (-0.28)

Profitability Growth

0.93 0.92

-1.42 0.29

(-1.33) (-1.53)

(-1.17) (0.33)

Loss Dummy

1.31 0.99

-18.76 -5.67

(1.95) (-0.05)

(-4.70) (-2.06)

Log (CEO Age)

0.68 1.55

22.82

(-1.08) (6.30)

(14.63)

Log (CEO Tenure) 1.45 (4.94) Male

1.45

19.02

(4.94)

(2.38)

Founder

1.08

106.18

(0.25)

(25.32)

Chair

-0.64

38.65

(-2.38)

(13.24)

GAI

-0.66

-8.65

(-3.32)

(-5.03)

Rolodex

1.20

8.07

(3.15)

(4.71)

Percent CEO Text

1.07

3.07

(1.25)

(1.81)

Optimism

1.10

-0.13

(1.05)

(-0.09)

Overconfidence

-0.96

1.24

(-0.62)

(0.40)

IA.25

MBA

-0.86

-4.59

(-1.08)

(-1.49)

Doctorate

-0.99

-24.70

(-0.08)

(-3.23)

Ivy League

1.38

9.82

(1.22)

(3.69)

Grad with Honors

-0.88

-0.10

(-1.37)

(-0.02)

Emotional Stability

-0.77

4.05

(-1.50)

(2.19)

Openness

-0.83

-3.24

(-2.70)

(-1.56)

Agreeableness

1.17

-0.36

(2.19)

(-0.19)

Conscientiousness

1.07

3.20

(0.99)

(1.55)

Fixed Effects Industry &

Year Industry &

Year Industry &

Year

Industry & Year

Industry & Year

Industry & Year

R-squared 3.24% 3.84% 12.47%

2.28% 5.11% 51.37% Observations 10,925 10,925 10,925

10,925 10,925 10,925

IA.26

Table IA.10 Extraversion and Outside Directorships

(same as Table 7 in the text, with test-statistics reported for the control variables)

Ln

(1 + Dir.) Logit

(Dir. =1) Dir. Size: Ln(Sales)

Dir. Size: Ln(Assets)

Dir. Size: Ln(Equity)

[1] [2]

[3] [4] [5] Extraversion 1.99% 1.16

11.38% 21.58% 17.20%

(2.19) (2.29)

(1.68) (2.68) (2.49)

Ln (Sales) 2.76% 1.22

56.93% 59.10% 38.55%

(1.31) (1.24)

(3.52) (3.10) (2.39)

Ln (Assets) -0.54% 1.03

29.97% 35.45% 51.06%

(-0.24) (0.17)

(1.81) (1.79) (3.03)

Ln (Q) -1.39% 0.91

3.93% 2.06% 16.68%

(-1.47) (-1.39)

(0.51) (0.23) (2.18)

Ln (Vol) -1.31% 0.94

7.92% 18.05% 9.10%

(-1.57) (-1.15)

(1.14) (1.98) (1.23)

Ln (Firm Age) 0.45% 1.04

5.64% 7.97% 7.31%

(0.48) (0.55)

(0.89) (0.97) (0.99)

Ln (Sales Growth) 0.67% 1.06

-2.81% -5.02% -2.57%

(1.84) (1.96)

(-0.84) (-1.11) (-0.70)

Fiscal Return -0.29% 0.97

0.49% 0.93% 5.90%

(-0.59) (-0.89)

(0.12) (0.19) (1.39)

Lag Fiscal Return -0.39% 0.96

-3.94% -3.76% -3.13%

(-0.87) (-1.15)

(-0.92) (-0.74) (-0.67)

Profitability -0.07% 1.03 19.95% 20.91% 20.72% (-0.09) (0.40) (2.26) (2.05) (2.04) Profitability Growth 0.01% 0.98 -7.92% -6.18% -7.93% (0.03) (-0.52) (-1.82) (-1.26) (-1.59) Loss Dummy -2.63% 0.83 1.65% 1.76% 5.75% (-1.92) (-1.83) (0.17) (0.14) (0.53) Log (CEO Tenure) 2.03% 1.12 2.50% -2.47% -9.65% (2.58) (1.88) (0.33) (-0.27) (-1.18) Log (CEO Age) 6.56% 1.50 2.07% 11.21% 14.59% (7.23) (6.05) (0.30) (1.29) (2.10) Male -15.00% 0.33 -64.66% -79.32% -76.00% (-3.25) (-3.57) (-2.73) (-2.86) (-3.72) Founder -8.22% 0.55 30.42% 24.51% 25.37% (-3.10) (-3.23) (1.59) (1.00) (1.20) Chair 3.81% 1.29 18.93% 26.81% 23.60% (2.47) (2.48) (1.72) (2.10) (2.16) GAI 13.58% 2.23 6.92% 14.72% 11.23% (14.35) (12.50) (1.07) (1.81) (1.75) Percent CEO Text -0.97% 0.93 27.46% 31.51% 27.16% (-1.19) (-1.19) (6.03) (5.42) (5.65)

IA.27

Optimism -1.42% 0.94 5.44% 7.59% 3.32% (-1.71) (-1.20) (0.95) (1.04) (0.55) Overconfidence -2.96% 0.76 3.85% 6.03% 2.74% (-1.81) (-2.62) (0.69) (0.85) (0.49) MBA 3.75% 1.24 -12.77% -3.78% -1.11% (2.23) (1.94) (-1.23) (-0.30) (-0.10) Doctorate -1.04% 1.11 2.68% -9.13% -2.63% (-0.18) (0.24) (0.24) (-0.69) (-0.24) Ivy League 0.01% 1.01 -47.36% -50.16% -28.55% (0.01) (0.05) (-1.37) (-1.13) (-0.73) Grad with Honors -0.89% 0.95 10.53% 21.49% 16.12% (-0.39) (-0.35) (1.13) (2.00) (1.97) Emotional Stability -0.21% 0.99 16.40% 9.65% 17.76% (-0.22) (-0.14) (1.08) (0.51) (1.11) Openness -0.42% 0.99 1.26% -2.37% -7.06% (-0.40) (-0.08) (0.19) (-0.30) (-1.04) Agreeableness -0.54% 0.96 0.66% -1.75% 1.07% (-0.55) (-0.67) (0.09) (-0.18) (0.14) Conscientiousness 0.47% 1.02 -10.68% -7.47% -6.96% (0.44) (0.31) (-1.52) (-0.87) (-1.00) Observations 9,637 9,637

2,222 2,222 2,222

R2/Psuedo R2 27.56% 26.61%

45.86% 43.71% 46.01%

IA.28

Table IA.11 Extraversion and CFO Promotion to CEO

(same as Table 8 in the text, with test-statistics reported for the control variables) CFO Tenure CEO Extraversion

All Promotions

Short Long

Extravert Introvert

[1] [2] [3] [4] [5] [6] [7] Extraversion 1.52 1.41 1.31 2.61 1.23 1.90 1.03

(3.02) (2.35) (1.79)

(2.32) (1.09)

(2.29) (0.07)

Ln (Sales) 0.85 0.85 0.92

0.80 0.98

1.07 0.85

(-1.10) (-1.13) (-0.57)

(-0.71) (-0.08)

(0.25) (-0.41)

Ln (Q) 1.18 1.15 1.17

1.86 0.92

1.02 0.57

(1.25) (1.06) (1.20)

(2.09) (-0.38)

(0.09) (-1.35)

Lag Fiscal Ret 0.91 0.83 0.82

0.56 0.84

0.94 1.17

(-0.57) (-1.11) (-1.20)

(-1.41) (-0.57)

(-0.25) (0.26)

Lag Profitability 0.94 0.93 0.95

0.66 1.05

0.82 1.04

(-0.41) (-0.42) (-0.33)

(-0.91) (0.21)

(-0.77) (0.10)

Tenure 1.08 1.09 1.08

0.81 1.16

1.60 1.07

(0.54) (0.63) (0.52)

(-0.20) (0.53)

(1.74) (0.15)

Exec Age 0.51 0.51 0.46

2.20 0.25

0.15 0.54

(-1.42) (-1.42) (-1.68)

(0.72) (-2.03)

(-1.85) (-0.36)

Optimism 0.89 0.89 0.90

1.15 0.67

0.99 0.61

(-0.79) (-0.80) (-0.75)

(0.56) (-1.58)

(-0.05) (-0.88)

Emotional Stability 1.06 1.06 1.11

0.79 1.14

1.14 1.46

(0.39) (0.41) (0.76)

(-0.91) (0.69)

(0.37) (0.85)

Openness 1.15 1.11 1.09

1.17 0.94

1.07 1.04

(0.83) (0.63) (0.49)

(0.46) (-0.27)

(0.19) (0.06)

Agreeableness 1.05 1.07 1.09

1.04 1.33

0.97 2.24

(0.34) (0.48) (0.56)

(0.18) (1.28)

(-0.10) (1.40)

Conscientiousness 0.90 0.90 0.91

0.86 0.99

0.85 0.86

(-0.63) (-0.66) (-0.57)

(-0.50) (-0.05)

(-0.45) (-0.35)

Cumulative Returns 1.19 1.18

0.93 2.60

1.21 1.64

(1.32) (1.19)

(-0.31) (2.63)

(0.67) (1.28)

Relative For. Error 1.12 1.12

1.45 0.82

0.29 1.42

(1.19) (1.21)

(1.46) (-0.64)

(-2.35) (1.75)

Guidance Dummy 0.88 0.89

1.38 0.82

0.58 0.81

(-0.52) (-0.43)

(0.51) (-0.53)

(-1.14) (-0.24)

CFO Percent Text 1.18 1.17

0.98 1.21

1.62 0.96

(1.28) (1.24)

(-0.06) (0.96)

(1.96) (-0.11)

Relative Salary 1.81

0.65 3.22

2.39 6.84

(4.14)

(-1.55) (3.66)

(2.59) (1.73)

Observations 1171 1171 1171

439 547

453 379 R-squared 5.33% 5.71% 6.89%

11.47% 13.17%

15.46% 14.65%

Prob. of CFO Promotion 7.94% 7.94% 7.94%

5.47% 9.87%

7.95% 4.75%

IA.29

Table IA.12 CEO Extraversion and Departure Announcement Returns

(same as Table 10 in the text, with test-statistics reported for the control variables) All Departures Voluntary Departures Unexpected

[1] [2] [3] [4] [5] [6] [7]

Extraversion -0.19% -0.18% -0.21% -0.57% -0.56% -0.19% -2.42%

(-0.59) (-0.57) (-0.49) (-2.20) (-2.17) (-0.61) (-2.74)

Ln (Sales)

1.31% 1.67%

0.42% 0.35%

(1.43) (1.66)

(0.46) (0.34)

Ln (Assets)

-1.17% -1.13%

-0.25% -0.18%

(-1.25) (-1.08)

(-0.28) (-0.17)

Ln (Q)

-0.46% -0.34%

-0.30% -0.21%

(-1.09) (-0.75)

(-0.80) (-0.52)

Ln (Vol)

-0.25% 0.08%

0.15% 0.24%

(-0.37) (0.13)

(0.30) (0.43)

Ln (Age)

-0.46% -0.62%

-0.26% -0.45%

(-1.59) (-1.38)

(-0.93) (-1.12)

Lag Fiscal Ret

-0.15% -0.21%

-0.23% -0.28%

(-0.56) (-0.78)

(-0.83) (-1.03)

Log (CEO Tenure)

-0.32%

-0.02%

(-0.65)

(-0.04)

Log (CEO Age)

0.86%

0.56%

(2.09)

(1.50)

Male

-0.57%

-3.43%

(-0.33)

(-1.77)

Founder

-0.94%

-0.68%

(-0.82)

(-0.62)

Chair

-0.93%

0.06%

(-1.26)

(0.08)

GAI

0.16%

-0.16%

(0.47)

(-0.47)

Rolodex

-0.36%

0.15%

(-0.89)

(0.40)

Percent CEO Text

-0.49%

-0.54%

(-1.04)

(-1.26)

Optimism

0.07%

-0.20%

(0.18)

(-0.49)

Overconfidence

-0.45%

-0.25%

(-0.66)

(-0.44)

MBA

-1.10%

-0.81%

(-1.52)

(-1.45)

Doctorate

2.24%

2.34%

(1.75)

(1.28)

Ivy League

0.37%

0.47%

(0.71)

(0.96)

Grad Honors

-0.72%

-1.93%

(-0.70)

(-2.16)

Emotional Stability

1.21%

0.12%

IA.30

(2.37)

(0.33)

Openness

0.06%

-0.01%

(0.16)

(-0.03)

Agreeableness

-0.64%

-0.51%

(-1.48)

(-1.38)

Conscientiousness

0.25%

0.64%

(0.60)

(1.63)

Observations 757 757 742 531 531 520 15 R-squared 2.92% 4.24% 7.84% 4.73% 5.54% 8.86% 18.83%

IA.31

Table IA.13 CEO Extraversion and M&A Announcement Returns

(same as Table 11 in the text, with test-statistics reported for the control variables)

[1] [2] [3]

Extraversion 0.22 0.29 0.38

(1.45) (1.97) (2.09)

Tender

1.84 1.75

(2.59) (2.46)

Equity Finance

-1.72 -1.80

(-1.68) (-1.75)

Mixed Finance

-0.56 -0.60

(-1.04) (-1.10)

Public Target

-2.71 -2.61

(-5.27) (-5.13)

Private Target

-0.96 -0.90

(-3.25) (-2.97)

Ln (Sales)

-0.06 -0.03

(-0.12) (-0.06)

Ln (Assets)

-0.31 -0.13

(-0.57) (-0.24)

Ln (Q)

-0.26 -0.33 (-1.49) (-1.80) Ln (Vol)

0.03 0.10

(0.13) (0.44)

Ln (Age)

-0.01 (-0.06)

(-0.08) (-0.29)

Lag Fiscal Ret

-0.03 -0.10 (-0.20) (-0.75) Log (CEO Tenure) 0.21 (1.28) Log (CEO Age) -0.04 (-0.22) Male 0.74 (0.90) Founder -0.39 (-0.73) Chair -0.10 (-0.32) GAI 0.05 (0.30) Rolodex -0.20 (-1.23) Percent Ceo Text 0.03 (0.22) Optimism 1.37 (4.02) Overconfidence -0.09 (-0.65)

IA.32

MBA -0.20 (-1.00) Doctorate -0.06 (-0.47) IvyLeague -0.09 (-0.51) GradHonors 0.19 (1.13) Emotional Stability -0.19 (-1.18) Openness 0.18 (0.82) Agreeableness -0.18 (-0.86) Conscientiousness -0.18 (-0.81) Fixed Effects Industry & Year Industry & Year Industry & Year Observations 1864 1864 1864 R-squared 2.91% 6.70% 8.48%