CEO Personality — What the Evidence Actually Says

Module 1 established that CEOs are an unusual population. This module asks the next question: what does personality — measured properly — actually predict once someone is in the chair?

The honest answer has two halves. RESEARCH FINDING Personality reliably predicts what CEOs do: how much they acquire, how much they borrow, how much they change strategy, how much they spend on R&D (Malhotra et al., 2018; Gow et al., 2016; Harrison et al., 2019). RESEARCH FINDING It predicts whether the company wins far less reliably, and the effects that do appear are small, conditional on industry and recent performance, and frequently reverse under a different measure (Cragun, Olsen & Wright, 2020; Herrmann & Nadkarni, 2014).

The second thing you need to know is how the evidence was made. Very few CEOs sit for personality tests. So researchers infer personality from photographs in annual reports, pronoun counts in interviews, pay gaps, option-exercise behavior, earnings-call transcripts, and — in one famous case — biographies of seventeen executives coded by graduate students. Some of these proxies are validated; all of them are noisy; and the noise is not random.

This module works on the Traits term of the Effectiveness Equation by giving you the tools to read any claim about CEO personality — including claims about yourself — with the right degree of confidence. You will finish with a nine-item Skeptic's Checklist you can run on a headline in under a minute.

Personality is a reliable predictor of behavior and an unreliable predictor of results — and most of what you have read about it was measured from a distance.

Traits set a CEO's defaults: what they reach for under pressure, what kind of change they initiate, what they find easy to say. Whether those defaults produce results depends on the company, the industry, the moment and the governance around them. Any study that skips from a trait to a stock return without passing through those conditions is telling you less than it seems, and if it measured the trait from a photograph, it is telling you even less.

The general-population baseline

RESEARCH FINDING Judge, Bono, Ilies and Gerhardt (2002) meta-analyzed 222 correlations from 73 samples and found corrected correlations with leadership (emergence and effectiveness combined) of .31 for extraversion, .28 for conscientiousness, .24 for openness, −.24 for neuroticism and .08 for agreeableness, with a multiple correlation of .48. Extraversion was the most consistent correlate. But the samples were students, military and mid-level managers, and "leadership" meant peer-rated emergence or supervisor-rated effectiveness — not running a company. It is a baseline, not a CEO finding.

Trait-by-trait evidence in CEOs

RESEARCH FINDING Extraversion. Malhotra, Reus, Zhu and Roelofsen (2018), measuring extraversion from unscripted remarks by 2,381 S&P 1500 CEOs across 10,166 firm-years, found more extraverted CEOs made more and larger acquisitions, especially in less competitive industries and where managerial entrenchment was higher; their deals also earned stronger announcement returns. Gow, Kaplan, Larcker and Zakolyukina (2016; working paper) found extraversion negatively related to contemporaneous and future ROA and cash flow. Harrison, Thurgood, Boivie and Pfarrer (2020) found observed extraversion associated with higher perceived firm risk (stock volatility). Benischke, Martin and Glaser (2019) found that the usual negative relationship between option wealth at risk and strategic risk taking reversed to positive for highly extraverted CEOs. Extraversion, then, is associated with doing more — and the evidence on whether the doing pays is mixed.

RESEARCH FINDING Conscientiousness. Herrmann and Nadkarni (2014), surveying 120 Ecuadorian SMEs, found conscientiousness cut both ways: it hindered the initiation of strategic change but improved the performance of change once implemented. Nadkarni and Herrmann (2010), in 195 Indian BPO firms, found conscientiousness inhibited strategic flexibility, which in turn mediated the personality–performance link. Gow et al. (2016) found it negatively related to growth. Harrison et al. (2020) found observed conscientiousness associated with lower perceived risk and better returns. Benischke et al. (2019) found conscientious CEOs responded to equity risk-bearing with more caution. Conscientiousness is the trait of execution and restraint; it helps finishing and hurts starting.

RESEARCH FINDING Openness. Gow et al. (2016) found openness positively related to R&D intensity and negatively to leverage. Nadkarni and Herrmann (2010) found it enhanced strategic flexibility; Herrmann and Nadkarni (2014) found it related to initiating change but not to implementing it well. Benischke et al. (2019) found open CEOs, like extraverted ones, converted risk-bearing into boldness.

RESEARCH FINDING Agreeableness. Judge et al. (2002) found agreeableness related to effectiveness (.21) but not emergence (.05). Herrmann and Nadkarni (2014) found it related to both initiating change and implementing it well. An inverted-U effect of agreeableness is often attributed to Nadkarni and Herrmann (2010), but the bibliography could not confirm it from an accessible source; treat that specific claim as unverified.

RESEARCH FINDING Neuroticism / emotional stability. Judge et al. (2002) found neuroticism the strongest negative correlate of leadership (−.24). Nadkarni and Herrmann (2010) found emotional stability enhanced strategic flexibility; Herrmann and Nadkarni (2014) found it related to both initiation and effective implementation. Harrison et al. (2020) found observed neuroticism associated with higher perceived firm risk.

How the traits are measured

RESEARCH FINDING Harrison, Thurgood, Boivie and Pfarrer (2019) developed and validated a machine-learning tool that scores CEO Big Five traits from earnings-call speech against observer ratings, applied it to more than 3,000 S&P 1500 CEOs, and found traits predicted strategic change with effects that depended on recent firm performance. Gow et al. (2016) built a similar model from ghSMART assessments and applied it to roughly 70,000 calls by roughly 4,600 CEOs; out-of-sample correlations between predicted and assessed traits ranged from about .23 (agreeableness) to .49 (neuroticism), which means substantial measurement error. Chatterjee and Hambrick (2007) built the widely used "unobtrusive" narcissism index from photo prominence in the annual report, prominence in press releases, first-person-singular pronouns in interviews, and the CEO's pay relative to the second-highest-paid executive — later criticized because pay ratios may reflect firm size or governance. Malmendier and Tate (2005; 2008) proxied overconfidence by option-holding behavior and by press portrayal. Cragun, Olsen and Wright (2020), pooling 37 narcissism studies, concluded findings were "mixed and potentially dependent upon methods."

The cautionary tale

RESEARCH FINDING Peterson, Smith, Martorana and Owens (2003) used historiometric Q-sort ratings of archival material — interviews, biographies, writings — for 17 CEOs of 9 large US firms and reported correlations between CEO traits and top-team dynamics (conscientiousness with legalism and centralization; neuroticism with factionalism and rigidity; agreeableness with cohesion; extraversion with leader dominance; openness with risk taking), and between team dynamics and income growth. Hollenbeck, DeRue and Mannor (2006) showed the findings had low statistical power and unstable parameter estimates: results changed materially when a single CEO was dropped. The bibliography's instruction is to treat the study as illustrative, never as reliable effect sizes. It is presented here as contested.

Where the evidence is weak

Almost none of this literature measures the CEO directly; it measures speech, documents and portfolios and infers. The validated language tools correlate with assessed personality at moderate levels, so every downstream effect is attenuated and possibly biased. Samples are single-industry (Chatterjee & Hambrick), single-country SMEs (Nadkarni & Herrmann; Herrmann & Nadkarni), or large US public firms whose CEOs face very different discretion. Performance effects are small (the bibliography notes ρ ≈ .05–.15 for narcissism outcomes) and frequently non-significant on ROA and Tobin's Q. Bertrand and Schoar (2003) give the strongest evidence that who the manager is matters for policy, but it measures no trait at all. Hambrick (2007) called for direct psychological measurement of executives; the field has mostly answered with better proxies.

The Big Five, one at a time, for CEOs

FRAMEWORK The Big Five is the most replicated model of personality structure. For a CEO, each trait is best understood as a default setting with a characteristic footprint in corporate behavior.

Extraversion is the trait of assertion, energy and social dominance. In CEOs it is associated with doing more: more acquisitions, larger acquisitions, more strategic change (Malhotra et al., 2018; Harrison et al., 2019). It is also the trait most associated with emerging as a leader in general samples (Judge et al., 2002), which is why so many CEOs have it. What it is not associated with is better operating performance; Gow et al. (2016) found the opposite, and Harrison et al. (2020) found markets price extraverted CEOs as riskier. INTERPRETATION Extraversion buys activity and visibility, both of which are useful in some situations and costly in others.

Conscientiousness is the trait of order, discipline and follow-through. Its footprint is the cleanest in the library and the most instructive: it dampens initiation of change and improves execution of change (Herrmann & Nadkarni, 2014). Markets appear to like it (Harrison et al., 2020). It is the Operator's trait, and its failure mode is a company that executes beautifully toward a strategy it should have abandoned.

Openness is the trait of curiosity, imagination and tolerance for the unconventional. Its footprint is R&D, lower leverage, and the initiation — not the implementation — of strategic change (Gow et al., 2016; Herrmann & Nadkarni, 2014). It is the Visionary's trait.

Agreeableness is the trait of warmth, cooperation and conflict avoidance. It barely helps people emerge as leaders but modestly helps them be effective (Judge et al., 2002), and in small firms it relates to both starting and finishing change well (Herrmann & Nadkarni, 2014). INTERPRETATION The weak association with emergence is why agreeable CEOs are rarer than agreeable managers; the association with effectiveness is why the rare ones are often underrated.

Neuroticism is the trait of emotional reactivity. It is the most consistently negative trait in the leadership literature, and in CEOs it is associated with lower strategic flexibility and higher perceived risk (Nadkarni & Herrmann, 2010; Harrison et al., 2020). Its opposite, emotional stability, is the one trait that helps at both ends — starting change and executing it (Herrmann & Nadkarni, 2014). If you want a single trait to select on, it is this one, and it is also the one that is hardest to see in an interview.

Why success is not extreme extraversion or charisma

Three findings, laid side by side, dismantle the charismatic-CEO assumption.

First, Judge et al. (2002) showed that extraversion is the strongest correlate of leadership emergence. That explains why boards, search firms and the press find extraverts compelling, and why Kaplan and Sorensen (2021) found CEO candidates scoring higher on a charisma-versus-analytical factor and interpersonally strong candidates being hired more often. Charisma gets you into the room and gets you chosen.

Second, once in the chair, extraversion predicts activity: more deals, more change, more visibility (Malhotra et al., 2018). Activity is not performance. Malhotra's extraverts earned stronger short-window announcement returns on their deals — a market reaction, not a result — and Gow et al. (2016) found extraversion negatively related to ROA.

Third, the trait that markets reward and that helps execution is conscientiousness, which correlates with emergence far more than effectiveness in general samples (Judge et al., 2002) and which, in CEOs, actively suppresses the flashy initiation of change (Herrmann & Nadkarni, 2014). The trait that looks least like charisma is the one most associated with finishing.

INTERPRETATION Put together: charisma is over-selected because it is visible at hiring and looks like leadership; it is associated with doing more once hired; and doing more is not the same as doing well. This is the core of Module 14's Visibility ≠ prevalence; visibility ≠ effectiveness. It does not mean introverts make better CEOs. It means extreme extraversion is an entry-ticket trait, and the company's results depend on what sits underneath it.

Direct measurement versus proxies

FACT A direct measure asks the CEO, or someone who knows the CEO well, structured questions with known psychometric properties. The library contains three kinds: validated self-report surveys (Graham, Harvey & Puri, 2013), structured interview-based assessments by professionals (Kaplan, Klebanov & Sørensen, 2012; Kaplan & Sorensen, 2021) and other-ratings by the top team (Ou et al., 2014; 2018).

FACT A proxy infers the trait from something the CEO left behind. The main families:

• Document prominence and pay: photo size in the annual report, press-release mentions, first-person-singular pronouns, and the pay gap to the number two — the Chatterjee and Hambrick (2007) narcissism index, reused by Buyl, Boone and Wade (2019) and many others.
• Revealed preference in personal portfolios: holding deep-in-the-money options past the point a rational diversifier would exercise — the Malmendier and Tate (2005) "Longholder" overconfidence proxy, later reused for bank CEOs (Ho et al., 2016) and founders (Lee, Hwang & Chen, 2017).
• Press portrayal: how journalists describe the CEO (Malmendier & Tate, 2008).
• Language models on earnings calls: machine-learning tools trained to predict assessed Big Five scores from speech (Harrison et al., 2019; Gow et al., 2016; Malhotra et al., 2018).
• Historiometric coding: third parties rating biographies and interviews (Peterson et al., 2003).

INTERPRETATION Proxies are not worthless; several are validated, and without them there would be almost no large-sample CEO personality research. But each carries a specific contamination. Pay gaps reflect governance and firm size. Option-holding reflects inside information, liquidity and tax as well as belief. Earnings-call speech is coached, and the validation correlations of .23–.49 (Gow et al., 2016) mean that half or more of the variance in the "personality" score is not personality. Press portrayal reflects the journalist. And a headline that says "narcissistic CEOs do X" usually means "CEOs with large photographs and big pay gaps do X." The brief's rule is that every such measure is called a proxy, every time. Adopt it.

The CEO Research Skeptic's Checklist

FRAMEWORK Run these nine questions on any CEO-personality claim.

1. Endogeneity. Could something unmeasured drive both the trait score and the outcome? A firm's size drives both the pay-gap proxy and its acquisition capacity.
2. Selection. Who is in the sample? Kaplan and Sorensen's candidates were assessed because a PE firm chose to assess them. Chatterjee and Hambrick's CEOs were all in tech.
3. Reverse causality. Does the outcome shape the trait score? A firm already doing well gives its CEO a bigger photo and a bigger pay gap. Chatterjee and Hambrick (2007) themselves note narcissists may select into dynamic firms.
4. Measurement and proxy error. How far is the measure from the construct, and is the error random or systematic? A .3 validity correlation is not a small problem.
5. Industry effects. Does the finding hold outside the industry sampled? Hirshleifer, Low and Teoh (2012) found the overconfidence–innovation link only in innovative industries.
6. Firm effects. Does the trait matter over and above what the firm is? Bertrand and Schoar (2003) controlled firm fixed effects; many trait studies cannot.
7. Governance effects. Does the effect depend on how constrained the CEO is? Malhotra et al. (2018) found the extraversion–acquisition link stronger under entrenchment; Buyl et al. (2019) found board monitoring dampened narcissism's effect on risk.
8. Survivorship. Are we only seeing CEOs who lasted? Adams, Almeida and Ferreira (2009) showed founder-CEO status is endogenous to performance.
9. Small samples and many tests. How many CEOs, and how many correlations were run? This is the Peterson problem.

RESEARCH FINDING Peterson et al. (2003) is the checklist's worked example. Seventeen CEOs; personality coded by third parties from archival documents; dozens of correlations between five traits, several team-dynamics variables and firm performance. Hollenbeck et al. (2006) demonstrated that removing a single CEO materially changed the results. Items 4 and 9 fail outright; items 1, 2, 3 and 8 are unaddressed. The study is still cited in leadership books as evidence that "CEO personality shapes team dynamics." It may — but not because of this study. INTERPRETATION The lesson is not that the authors did anything wrong; the design was reasonable for its time. The lesson is that a finding can enter the practitioner canon on the strength of its narrative rather than its numbers, and only a checklist will catch it.

This is a fictional composite.

Halvorsen Building Products is a listed manufacturer of windows and doors: $610 million revenue, 2,900 employees, four plants in the upper Midwest, founded in 1948, 22% still owned by the founding family. In 2025 its CEO of nine years announced retirement. The board's search committee — three independents and the family's representative — narrowed to two finalists.

Candidate one, Marcus Delacroix, was president of a $1.4 billion division of a public conglomerate: fluent, expansive, energetic, with a record of six acquisitions in five years. Candidate two, Priya Venkataraman, was COO of a private competitor: precise, quiet in the room, with a record of margin improvement and one small acquisition that she had integrated flawlessly.

The night before the final vote, one independent director circulated an article summarizing Malhotra et al. (2018): extraverted CEOs make more and larger acquisitions and their deals earn stronger announcement returns. Halvorsen's strategy called for consolidation of a fragmented regional market. The director's note said: "The research supports Marcus."

The committee chair, a retired auditor, ran the checklist out loud.

Selection and industry: the study was S&P 1500 CEOs across industries; Halvorsen was a $610 million manufacturer with a family blockholder. Measurement: extraversion was inferred from earnings-call speech, and Delacroix had never given an earnings call; the committee's impression of his extraversion came from three interviews. Governance: the effect was strongest where entrenchment was higher; Halvorsen's family stake and active board were the opposite. Outcome: "stronger announcement returns" was a short-window market reaction, not integration success or five-year ROA. Reverse causality: CEOs of firms already positioned to acquire may speak more expansively.

The chair then did something more useful than the checklist: he asked what Halvorsen's actual problem was. Consolidation required doing deals, yes — but Halvorsen had never integrated a plant, its ERP was fifteen years old, and its last acquisition (2016) had lost a third of the acquired revenue in two years. The binding constraint was not deal initiation; it was deal execution. Herrmann and Nadkarni's (2014) duality — conscientiousness hinders initiation and helps implementation — pointed, if anything, the other way from the article.

They hired Venkataraman, with an explicit board mandate to build integration capability first and a commitment to revisit the acquisition pace in eighteen months. Delacroix went to a larger company, made four acquisitions in two years, and was profiled admiringly. Halvorsen made two acquisitions in three years, kept 94% of acquired revenue, and expanded margin by 210 basis points.

INTERPRETATION Neither outcome proves anything about extraversion. Both CEOs did what their traits predicted. The point is that the board used the research for what it could support — a prediction about behavior — and did not let it decide what it could not support: which behavior Halvorsen needed.

Give the same headline — "extraverted CEOs acquire more and the market rewards their deals" — to four sitting CEOs contemplating a consolidation strategy.

The Public Company CEO reads it as validation and as pressure. Investors have read it too; the sell side is asking about M&A. The gain is momentum on a strategy the market already wants. The cost is that the CEO's temperament and the market's appetite now reinforce each other with no skeptic in the loop, and short-window announcement returns become the internal definition of success. Six deals in, the integration debt comes due.

The Founder CEO reads it and shrugs, because a founder's acquisition style is set by conviction, not literature. The gain is immunity to headline-driven strategy. The cost is that founders are also measurably more optimistic on proxies (Lee, Hwang & Chen, 2017) and may be the CEOs the checklist protects least — nobody in the room applies it to them.

The PE-Backed CEO reads it and asks the sponsor what the model assumes. The board is the constraint the study says weakens the extraversion effect; the deal thesis is written down; the CEO's personality is one input among many, and the sponsor has seen the announcement-return story before. The gain is disciplined use of a real finding. The cost is that a sponsor's hold period can force acquisition pace regardless of what the CEO's traits or the checklist say.

The Operator CEO reads it and worries that they are the wrong personality for the strategy. The gain is self-awareness — the Operator knows conscientiousness suppresses initiation. The cost is over-correction: an Operator who tries to perform extraversion makes worse deals than either a natural extravert or a disciplined Operator. INTERPRETATION The right move for the Operator is usually not to change personality but to build the initiation capability into the team — a corporate-development lead with the trait the CEO lacks — and keep the integration discipline that is the Operator's actual edge.

The concept fails in two directions: treating personality as destiny, and treating research as instruction.

FRAMEWORK On the Overuse Ladder, the personality-as-destiny error runs rigor → bureaucracy. A board or CEO who discovers personality research starts assessing everyone, builds trait profiles into succession, and begins explaining every decision through the lens of somebody's Big Five. Assessment is useful; assessment as a substitute for judgment is not. The traits explain small shares of variance in behavior and smaller shares in results, and a company run by its personality reports has replaced one kind of over-confidence with another.

The research-as-instruction error is subtler and more common. A finding about behavior ("extraverts acquire more") is read as a finding about results ("hire an extravert"). A proxy ("CEOs with large photographs") is read as a construct ("narcissists"). A single-industry sample becomes a universal law. A 17-CEO study becomes a bullet point in a keynote. Each step loses information and gains confidence.

For the CEO personally, the failure mode is the one Module 1 introduced: the trait that got you selected — usually extraversion or charisma — is treated as the trait that makes you effective, and the trait that makes you effective — usually emotional stability and conscientiousness, or a team that supplies them — goes unexamined. Dominance → intimidation is the ladder rung for the extravert who never learned that emergence and effectiveness are different lists.

Early warning signs a CEO or board could notice:

• Personality language ("she's not a visionary," "he's too cautious") is doing the work that a diagnosis of the company's problem should do.
• A single study, article or assessment vendor is cited as settling a hiring or strategy question.
• Nobody in the room can say how the personality claim was measured.
• The CEO's visible traits are praised in board discussions; the CEO's execution record is not examined with the same care.
• The company is doing more — deals, launches, reorganizations — and nobody has asked whether more is what the situation needs.

1. Which Big Five trait would your executive team say is your most extreme? Which would your spouse say? If they differ, which one shows up under pressure?
2. Name a corporate decision in the last two years that your temperament made easy. Was it the right decision, or the easy one?
3. Have you ever been assessed directly — structured interview, validated instrument, 360? If not, everything you believe about your own profile is self-report or inference. How confident should you be?
4. Find one personality claim you have repeated about CEOs (yours or others'). Run the nine-item checklist on it. Which items fail?
5. If extraversion buys activity rather than results, what activity in your company exists because of your temperament rather than because of the strategy?
6. Who on your team supplies the trait you lack — the initiation you suppress, or the discipline you skip? What happens if that person leaves?

• The Big Five predict what CEOs do — acquisitions, R&D, leverage, strategic change — more reliably than whether the firm wins; performance effects are small, conditional and measure-dependent (Malhotra et al., 2018; Gow et al., 2016; Cragun et al., 2020).
• Conscientiousness hinders starting and helps finishing; extraversion and openness help starting; emotional stability helps both (Herrmann & Nadkarni, 2014).
• Extraversion and charisma are entry-ticket traits: they predict leadership emergence and hiring, and CEO activity, not operating results (Judge et al., 2002; Kaplan & Sorensen, 2021).
• Most CEO-personality research uses proxies — photographs, pronouns, pay gaps, option-holding, press portrayal, earnings-call language — with validation correlations that leave half or more of the variance unexplained. Call them proxies, every time.
• Run the nine-item Skeptic's Checklist on any headline; Peterson et al. (2003) and Hollenbeck et al. (2006) show how a well-told finding can fail it.