Why CEOs Aren't Normal

This module teaches one uncomfortable fact and one discipline that follows from it.

The fact: the people who become CEOs are not a random sample of capable adults. They are the output of three stacked filters — the people who want the job, the people promoted toward it, and the people boards finally choose. Each filter prefers a particular temperament: more optimistic, more risk-tolerant, more execution-driven, more confident that their own actions determine outcomes. RESEARCH FINDING The best direct evidence, from psychometric surveys and structured assessments of thousands of executives, confirms that CEOs differ from both the general population and their own CFOs on exactly these dimensions (Graham, Harvey & Puri, 2013; Kaplan & Sorensen, 2021).

The discipline: once you know the base rate, you stop mistaking it for a prescription. "CEOs tend to be optimistic" describes who survives the filters. It is not evidence that optimism makes companies win. Confusing the two is the founding error of most leadership writing, and the error this curriculum exists to undo.

In the Effectiveness Equation — Traits × Behaviors × Organizational Context × Current Moment — this module works on the Traits term: not by telling you which traits to have, but by showing what the typical CEO's profile looks like, how it got that way, and why you should judge yourself (and any CEO you hire) against that unusual population rather than the population at large.

CEOs are a filtered population, and the filter selects for temperament long before it selects for results.

Every study of CEO personality is, whether it says so or not, a study of people who passed through selection gates that reward confidence, optimism and appetite for risk. That means the "typical CEO" is unusual by construction, and it means a trait that is common among CEOs may be common because it helps people get the job, not because it helps them do it. The distinction sounds academic. It is the difference between hiring for what wins and hiring for what looks like winning.

Direct psychometric evidence: CEOs really are different

RESEARCH FINDING Graham, Harvey and Puri (2013) administered validated psychometric instruments — a standard dispositional-optimism scale and lottery-style risk-aversion items among them — to a large survey sample of CEOs and CFOs of public and private firms in the United States, Europe and Asia. CEOs scored markedly more risk-tolerant and more optimistic than population norms. Roughly 80% of US CEOs were classified "very optimistic," against roughly 65% of CFOs, and US executives were more optimistic than non-US ones. The traits were associated with corporate behavior: risk-tolerant CEOs' firms initiated more M&A and were more often growth firms; optimistic CEOs used more short-term debt; risk-averse CEOs received pay with a higher fixed-salary share.

What this can support: CEOs, as a population, sit well to the optimistic and risk-tolerant end of the human distribution, and those traits travel with observable corporate choices. What it cannot support: any causal claim. The design is cross-sectional and self-reported, executives self-selected into responding, and the population comparison relies on external benchmarks. Traits are associated with policies. We will keep saying that.

RESEARCH FINDING Kaplan and Sorensen (2021) analyzed 2,603 structured assessments of candidates for CEO, CFO, COO and other top roles conducted by a single assessment firm between 2000 and 2013. Four factors explained more than half the variance in thirty rated characteristics: general ability; execution versus interpersonal; charisma versus analytical; strategic versus managerial. CEO candidates scored higher on all four; CFO candidates lower. Patterns held across public, PE-backed and VC-backed firms. Candidates with stronger interpersonal skills were more likely to be hired, while the factors — including execution — predicted later advancement to CEO, which the authors read as boards possibly over-weighting interpersonal skill at the point of hire. Gender differences on the factors were small, yet women were less likely to become CEO.

What this can support: the CEO candidate pool is measurably distinct from other executive pools. What it cannot support: firm-performance claims — the outcome is hiring and career progression, not results — and it comes from one assessor's interview-based ratings of people already selected for assessment.

Managers carry persistent "styles"

RESEARCH FINDING Bertrand and Schoar (2003) tracked top executives who moved between firms and found that manager fixed effects explained a meaningful share of variation in investment, financial and organizational policies — acquisitions, leverage, dividends, cost-cutting — after controlling for firm fixed effects. Managers carried consistent styles across employers; older cohorts were more conservative, MBA holders more aggressive. Because the design follows the same person into different firms, this is the closest thing in the library to causal evidence that individual executives matter. It cannot support any specific personality claim: "style" is a statistical residual, not a measured trait, and executives observed at two or more firms are themselves a selected group.

The filters: why the population is non-representative

RESEARCH FINDING Judge, Bono, Ilies and Gerhardt (2002) meta-analyzed 222 correlations from 73 samples — mostly students, military personnel and mid-level managers, not CEOs — and found that extraversion (.31), conscientiousness (.28), openness (.24) and emotional stability (neuroticism −.24) each correlated modestly with leadership emergence and effectiveness combined, with a multiple correlation of about .48. Critically, some traits related more to emergence (being seen as a leader) than to effectiveness: conscientiousness .33 for emergence versus .16 for effectiveness; agreeableness .05 for emergence but .21 for effectiveness. This is the cleanest demonstration in the library that the traits that get you noticed are not the traits that make you effective — which is exactly how selection distorts the CEO population.

RESEARCH FINDING Adams, Almeida and Ferreira (2009) showed that founder-CEO status is endogenous: founders step down after both very bad and very good results, so the founders in the chair at any moment are a doubly filtered set. Wasserman (2003), in a hazard analysis of 202 internet start-ups, found that hitting milestones — completing the product, raising each round — sharply raised the probability the founder would be replaced. The CEO you observe is the CEO who was not removed, and removal is not random. Lee, Hwang and Chen (2017) add, using language- and option-based proxies, that founder CEOs of S&P 1500 firms displayed more optimistic and overconfident behavior than professional CEOs — proxy evidence, but a reminder that different routes into the job produce different temperament profiles.

The theory that ties it together

FRAMEWORK Hambrick and Mason (1984) proposed that organizations reflect their top managers' experiences, values and cognitive bases; Hambrick (2007) added that executives' characteristics show up in outcomes only where they have discretion, and that heavy job demands push executives toward heuristics and disposition. A framework, not a finding — but it explains why a filtered population should worry us: if the firm reflects the CEO, a systematically unusual CEO population produces systematically unusual firms.

Where the evidence is weak

The direct psychometric evidence rests on two large but non-random samples (survey respondents; candidates one assessor evaluated). Neither links traits to firm performance in a way that survives the objection that unusual people select into unusual firms. The "CEO temperament" list in the learning objectives goes beyond what these studies measured: the library directly supports optimism, risk tolerance, general ability, execution orientation, charisma and strategic orientation; agency, achievement drive, uncertainty tolerance, internal locus of control, resilience and decisiveness should be read as INTERPRETATION consistent with, but not proven by, that evidence. Finally, how much any CEO matters is contested: Quigley and Hambrick (2015) report the share of performance variance attributable to CEOs rising across decades, while Fitza (2014; 2017) argues that much of any measured "CEO effect" is indistinguishable from chance given short tenures. Present any percentage as disputed.

Start with the base rate

If you meet a randomly chosen adult and they tell you they are confident, optimistic and comfortable betting on themselves, you learn something about that person. If you meet a CEO and they tell you the same thing, you learn almost nothing, because nearly every CEO would say it. The information content of a trait depends on how common it is in the population you are drawing from. CEOs are drawn from a population in which confidence is close to universal.

This is the base-rate problem, and it runs through every judgment you will make about a CEO, including the one in the mirror. A board impressed by a candidate's conviction is impressed by the entry ticket. An investor who reads a founder's optimism as a signal about the venture is reading a temperament that was fixed long before the venture existed. A CEO who takes their own certainty as evidence is sampling from a distribution designed to produce certainty.

Three filters, stacked

FRAMEWORK Think of the route to the CEO chair as three filters in series, each letting through a narrower slice of temperament.

Self-selection. The job attracts people who want responsibility for outcomes they cannot fully control, tolerate ambiguity, and believe their actions matter. People who find that exhausting rather than energizing do not put themselves forward. Nobody designs this filter; it operates through career choices made in someone's twenties. INTERPRETATION This is the likeliest source of the agency and internal-locus-of-control components of the temperament — not because the evidence measures them, but because it is hard to imagine anyone reaching the final filter without them.

Promotion selection. Organizations promote people who look like leaders to the people already in charge. Judge et al. (2002) show that the traits predicting emergence as a leader are not the traits predicting effectiveness — conscientiousness helps you emerge much more than perform; agreeableness helps you perform but barely helps you emerge. Every promotion round applies a little more of this bias, and a career is twenty of them.

Board selection. The final gate is a small group making a rare, high-stakes, low-feedback decision under time pressure. Kaplan and Sorensen (2021) found boards appeared to favor interpersonal skill at hiring even though execution better predicted later advancement. Boards are human; they hire people they find persuasive in a room.

Stack the three and you get the population Graham, Harvey and Puri (2013) measured. INTERPRETATION The CFO comparison is the most telling detail in that study. CFOs are also senior, filtered and ambitious — and still measurably less optimistic. The CEO filter selects for something specific.

The temperament, and what it is not

The recurring components — agency, achievement drive, optimism, uncertainty tolerance, calculated risk-taking, internal locus of control, resilience, confidence, decisiveness, execution orientation — describe a real cluster. Kaplan and Sorensen's four factors and Graham, Harvey and Puri's optimism and risk tolerance are the parts the library measures directly; the rest is a reasonable reading of the same picture.

Two things the cluster is not.

It is not a virtue list. Each component is a setting, not a good. Optimism gets a factory built during a downturn and keeps a doomed product alive for three extra quarters. Risk tolerance wins the acquisition and over-pays for it. The Trait Dial (Module 6) exists because the same trait has a left and a right, and context decides which is right.

It is not a prediction of success. That deserves its own section.

"CEOs tend to be X" versus "X makes a good CEO"

The first statement is about the filtered population. The second is about causation inside the job. They are logically independent, and the evidence supports the first far better than the second.

Consider optimism. RESEARCH FINDING Graham, Harvey and Puri (2013) found CEOs highly optimistic and found optimism associated with more short-term debt. That describes a tendency and its footprint. It says nothing about whether optimistic CEOs run better companies — and Module 4 will show the evidence points toward higher variance of outcomes, not a higher average.

Consider charisma. RESEARCH FINDING Kaplan and Sorensen (2021) found CEO candidates score higher on a charisma-versus-analytical factor and that interpersonally strong candidates are more likely to be hired. That is a fact about selection. Whether charisma helps once you have the job is a separate question; Module 3 will argue that execution predicts outcomes better than interpersonal polish in the one dataset that lets us compare them (Kaplan, Klebanov & Sørensen, 2012).

Sliding from "tends to be" to "makes a good" is not a beginner's mistake. Boards make it. Search firms make it. CEOs make it about themselves when they credit the traits that got them hired rather than the behaviors they chose once inside. INTERPRETATION A useful discipline: whenever you hear a trait praised in a CEO, ask whether the praise would have been offered before the results were known. If yes, it is a selection trait. If it could only be said in hindsight, it might be a performance trait. Most praise is the first kind.

Survivorship, and the illusion of uniform confidence

The CEOs you can observe are the ones who have not been removed. Adams, Almeida and Ferreira (2009) show founders leave after both very bad and very good results; Wasserman (2003) shows founders are replaced precisely when their companies hit milestones. Boards fire CEOs whose confidence visibly failed and quietly ease out CEOs whose confidence was never visible. What remains looks more uniformly confident than the population that started.

INTERPRETATION Survivorship does three things to your judgment. It makes confidence look like a requirement when it may be a residue. It hides the CEOs whose doubt served them well, because doubt is rarely narrated. And it produces the celebrity-CEO literature, where survivors' traits are back-fitted into a formula. Malmendier and Tate (2009) found that CEOs who win prestigious business-press awards subsequently underperform relative to their own prior record and to matched non-winners, earn more, and divert effort to outside activities — strongest in weakly governed firms. The award is a survivorship marker and, it turns out, a warning sign. Module 14 takes this further: Visibility ≠ prevalence; visibility ≠ effectiveness.

What to do with the base rate

First, judge CEOs against the CEO population, not the general one. A "very optimistic" CEO is at the median of their peers. A measured, skeptical CEO is unusual in a direction the filters do not reward, which means they got through on something else — worth finding out what.

Second, treat your own temperament as a prior, not a signal. If you feel certain, ask what you would expect to feel given who gets this job. The Two-Sentence CEO Test — "We're going to do this." / "I was wrong. Change the plan." — is built for a population in which the first sentence is easy. This module explains why the second is the rare capability.

Third, when you hire, separate entry-ticket traits from the traits your situation needs. The Fit Equation (Industry × Scale × Lifecycle × Strategy × Governance × Problem) is what you use instead of the base rate. Module 7's Maturity Model puts it in order: temperament gets you to the door, and everything that matters happens after.

This is a fictional composite.

Marlow Fastening Systems is a family-owned manufacturer of industrial fasteners in Ohio: $140 million in revenue, 610 employees, three plants, founded in 1961 and now chaired by the founder's grandson. In 2024 the board hired Dana Okafor as the first non-family CEO. She had been COO of a larger public competitor, and the search committee's notes praised her "conviction," "energy" and "obvious appetite for growth." She was, in every way the filters reward, the standard CEO candidate.

Eighteen months in, Okafor brought the board a plan: a $38 million fourth plant in Tennessee, sized for a customer contract that was 70% likely to close, financed with a term loan taking net debt from 0.8x to 2.6x EBITDA. The presentation was excellent. The chair, who had run the company for twenty-two years, asked one question: "How often have you been wrong about a customer signing?"

Okafor's honest answer was "not often." The chair's reply was the base-rate move: "That's what every CEO I've ever met says. I want to know what it means in your case."

So they looked. The board asked her to list every material forecast from her prior role — eleven, over six years — and how each resolved. Seven landed within range. Four were optimistic, two badly so, and both bad ones had involved a customer commitment that was "very likely" at the time. Her optimism was real, typical for the population she came from, and had a specific, repeatable shape: she was better at operational forecasts than at forecasts that depended on someone else's decision.

The board did not kill the plant. They restructured the decision. Phase one ($19 million, financeable to 1.6x) proceeded on the existing order book; phase two was conditioned on the contract actually signing rather than being likely to. The CFO, a lower-optimism personality with fourteen years at Marlow, was asked to own the trigger criteria in writing so the phase-two decision would not depend on how anyone felt in the room.

The contract signed eight months later, at 60% of the original volume. Phase two was resized to match. Okafor, who had wanted to build the whole plant at once, later called the split the best decision the board had made on her watch — and admitted she would not have made it alone.

The point is not that the chair was right and the CEO wrong. Okafor's optimism got the plant proposed at all; the previous family management had studied Tennessee for years without acting. Her temperament was why growth was on the table. The chair's contribution was to treat that temperament as a known base rate rather than as evidence, and to design a decision that would be right whether or not her optimism was warranted this time.

Put four archetypes into Okafor's chair with the same plant proposal and the same skeptical chair.

The Visionary CEO presents the full plant and a second one, argues the customer contract is the least interesting reason to build, and treats the chair's question as a failure of imagination. Gain: if the market turns as the Visionary expects, Marlow has capacity nobody else does. Cost: board trust degrades with every forecast that does not land, and the Visionary's own track record becomes something to avoid rather than examine. The base rate is treated as an insult instead of a fact.

The Operator CEO would never have proposed a plant sized to an unsigned contract. The proposal arrives phased, with a downside case and trigger criteria. Gain: the board's question is answered before it is asked. Cost: the Operator may have under-sized the opportunity, because the discipline that protects the balance sheet also dampens the willingness to bet on a customer who is genuinely about to sign. Marlow's previous management were Operators; that is why Tennessee took a decade.

The Family Business CEO — imagine the chair's cousin had been appointed — reads the question as a signal about how much latitude the family intends to give and negotiates the plant's size against relationships rather than the forecast. Gain: alignment with owners who will outlast any single contract. Cost: the decision is made on the wrong variable, and the plant is sized to family comfort rather than demand.

The PE-Backed CEO, dropped into a family company, treats the question as due diligence, produces the eleven-forecast track record unprompted, and proposes phasing as a way to earn phase two. Gain: fast credibility and an auditable decision. Cost: a family board may read the transactional framing as being managed rather than led, and the instinct to hit milestones fast may pull phase two forward before the contract is real.

INTERPRETATION None of the four is wrong in the abstract. The Operator's answer is safest, the Visionary's has the highest ceiling, the PE-Backed CEO's process is most defensible. Which one Marlow needed depended on whether its problem was under-investment (it was) and how much governance the board could exercise (a lot). The base rate does not tell you which archetype to be; it tells you what your own default probably is.

The CEO temperament fails in a specific way: the traits that passed the filters get mistaken for the traits that produce results, and the CEO stops calibrating.

FRAMEWORK On the Overuse Ladder, the base-rate error lives on three rungs at once. Confidence → arrogance: the CEO's certainty, which is population-typical and therefore uninformative, is treated by the CEO as evidence about the world. Optimism → delusion: the forecasting bias that Graham, Harvey and Puri (2013) found across the population becomes a personal blind spot that the CEO's history would reveal if anyone looked. Vision → fantasy: the CEO's story about why they succeeded — usually a story about temperament — becomes the plan for succeeding again, in a company or a moment that needs something else.

The mechanism is survivorship turned inward. A CEO who has been right several times in a row has, by definition, survived. They do not see the equally confident peers whose bets failed, and they credit their confidence rather than the mix of confidence, context and luck that actually produced the result. Chatterjee and Hambrick (2011) found that highly narcissistic CEOs respond less to objective performance feedback and more to social praise; you do not need to be narcissistic for a milder version to take hold once your track record feels like proof.

Early warning signs a CEO or board can notice:

• Forecasts are never audited against outcomes, and nobody has proposed it.
• Doubt from the CFO or board is described as "not understanding the opportunity" rather than as information.
• The CEO's account of past success is a story about character, not conditions.
• Decisions are sized to conviction rather than to what would be right under both the optimistic and pessimistic case.
• The CEO has begun to be profiled or given awards, and enjoys it (Malmendier & Tate, 2009).
• "I was wrong. Change the plan." has not been said out loud in the last year.

The correction is not to become less confident; confidence is the entry ticket and you need it to act. The correction is to know the base rate, audit your history against it, and build decisions robust to the possibility that this time your temperament is doing the talking.

1. List your last ten material forecasts — revenue, hires, timelines, customer decisions. How many resolved above, within and below your estimate? What shape does the miss have?
2. Which of the three filters (self-selection, promotion, board selection) did you pass most easily, and what does that say about which traits got you here rather than which you needed?
3. Think of a CEO you admire. Would you have praised the same traits before their results were known? If yes, what are you actually admiring?
4. When did your CFO, chair or a direct report last doubt a decision and turn out to be right? How did you describe their doubt at the time?
5. If you were removed tomorrow, what would the story be? Which peers were removed with that story, and what did they have in common with you?
6. Which temperament component — optimism, risk tolerance, decisiveness, execution drive — is highest in you, and where has it cost you money?
7. What decision are you currently making on conviction that could be restructured to be right whether or not the conviction is warranted?

• CEOs are a filtered population — self-selection, promotion selection and board selection each favor optimism, risk tolerance, confidence and execution drive — and direct psychometric evidence confirms CEOs differ from both the general population and their own CFOs (Graham, Harvey & Puri, 2013; Kaplan & Sorensen, 2021).
• A trait common among CEOs is uninformative about any individual CEO and is not evidence that the trait produces results. "CEOs tend to be X" and "X makes a good CEO" are different claims.
• The traits that get people noticed as leaders differ measurably from those that make them effective (Judge et al., 2002); every promotion and board decision applies that bias.
• Survivorship makes the CEO population look more uniformly confident than it is; observed CEOs are the ones not removed, and removal is not random (Adams, Almeida & Ferreira, 2009; Wasserman, 2003).
• Treat your temperament as a prior, audit your forecasting history against it, and build decisions that are right whether or not your conviction is warranted this time.