Why Do People Overestimate Their Own Abilities?

In a classic survey, 93% of American drivers rated themselves as above average in skill.

Think about what this means. Average is not a vague concept — it’s the midpoint of a distribution. Half of all drivers are, by definition, below average. The survey result means that roughly 43% of respondents were either wrong, or lying, or (most likely) genuinely believed something that cannot be true.

This is the above-average effect, one of the most replicated findings in social psychology, and it extends to nearly every domain of positive self-assessment: people rate themselves as better-than-average at driving, parenting, ethical judgment, leadership, intelligence, and a wide range of skills. In each case, the result is mathematically impossible if accurate.

Dunning-Kruger and the Calibration Problem

David Dunning and Justin Kruger published a paper in 1999 that identified a specific mechanism: people with less competence in a domain tend to have particularly poor insight into their own lack of competence.

Their subjects took tests on logic, grammar, and humor. The people in the lowest quartile of performance dramatically overestimated their performance — predicting scores far above their actual results. People in the top quartile, conversely, slightly underestimated their relative performance (though not their absolute performance).

The proposed mechanism: competence in a domain includes knowing what good performance looks like. You need to understand logic to recognize when your logic is poor. People who lack the skill also lack the metacognitive ability to accurately assess their own performance in that domain.

This has become known as the Dunning-Kruger effect, and it’s been widely cited as an explanation for confident incompetence.

The 2022 Controversy

In 2022, Gignac and Zajenkowski published a re-analysis that argued the Dunning-Kruger effect, as commonly depicted (the steep curve of overconfidence in the least skilled), might be partly a statistical artifact.

The specific claim: if you plot actual performance against self-assessed performance, you will tend to get a regression-to-the-mean effect that could produce apparent overconfidence at the low end and underconfidence at the high end as a mathematical consequence of correlation imperfection, not genuine asymmetric bias.

This critique is technically valid enough to have generated significant debate. The original paper’s methodology has been scrutinized.

What has not been seriously disputed: people are generally overconfident about their abilities, and calibration is poor across domains. The existence of the above-average effect is not in question. Whether the specific mechanism — that the least competent are disproportionately unaware — is a genuine psychological phenomenon or partly a statistical artifact remains an open debate.

Why Calibration Is Hard

The deeper reason overconfidence is so pervasive is that accurate self-assessment is genuinely difficult.

For most skills, the feedback loop is slow, indirect, or absent. You drive every day without getting clear, immediate feedback on whether you’re actually good at it — the absence of accidents isn’t calibrating evidence, because skill is not the only variable. A below-average driver can avoid accidents for years through luck; an excellent driver can have an accident through circumstances beyond their control.

In many domains — parenting, leadership, ethics, intelligence — the feedback is even harder to interpret. You have limited comparison information, the measures are ambiguous, and the stakes of accurately perceiving yourself as below average are socially and psychologically costly.

Motivated reasoning compounds this: we have strong incentives to believe we are competent, ethical, and capable. The cognitive system that evaluates our own performance has access to more favorable framing than an external evaluator would use.

The Illusion of Explanatory Depth

One specific version of overconfidence has its own name: the illusion of explanatory depth (IOED), documented by Leonid Rozenblit and Frank Keil.

People typically believe they understand how complex systems work — toilets, zippers, helicopter blades, policy proposals — better than they actually do. When asked to write out a detailed explanation of how a toilet works, most people discover they cannot.

The discovery typically produces a feeling of surprise. The prior sense of understanding was not recognized as shallow until the explanation was attempted.

This matters because much of our confidence is based on a feeling of understanding rather than demonstrated understanding. We can discuss topics confidently without having tested whether we could actually explain them. The feeling of familiarity — “I know what that is” — can be mistaken for the ability to deploy or explain knowledge.

What’s Accurately Assessed

Overconfidence is not uniform.

For tasks with quick, clear, unambiguous feedback — sports performance, typing speed, chess — calibration improves dramatically because the feedback loop closes fast. You know immediately whether your move was effective; over time, this produces reasonably accurate self-assessment.

In domains where feedback is ambiguous, delayed, or absent — judgment, social skills, intelligence, ethical character — overconfidence persists.

The distribution of confidence also shifts with genuine expertise. In medicine, engineering, and other technical domains with strong feedback systems, experts tend to be better calibrated than novices, though overconfidence doesn’t disappear entirely even at high expertise levels.

The above-average effect doesn’t mean people are delusional.

It means accurate self-assessment requires something most domains don’t provide: clear, immediate, comparative feedback on performance. Without that, the brain fills in with a version of self that is generally more capable than the evidence would support.

Which is to say: 93% of drivers believe they’re above average, and almost none of them think they’re being irrational.