Overconfidence Effect

People’s confidence in their own judgments and estimates systematically exceeds how accurate those judgments actually turn out to be, especially on difficult or uncertain questions.

4 min read

What Is It?

“Overconfidence” covers a few related but distinct phenomena in the research literature, overestimating one’s own ability, overplacing oneself relative to others, and overprecision, being more certain a specific judgment is correct than the evidence warrants. This entry focuses on overprecision, which is the version most relevant to everyday organizational estimates. Researchers studying judgment calibration, notably Sarah Lichtenstein, Baruch Fischhoff, and colleagues in a series of studies through the 1970s, found a consistent pattern when they asked people to estimate a fact and then state how confident they were in that estimate: across many difficult general-knowledge tasks, answers assigned very high confidence turned out correct less often than that confidence level would predict. Miscalibration is often especially visible on difficult questions, where people can remain quite confident even as their objective accuracy falls, though the size and even direction of this gap is sensitive to the specific tasks and questions used. The effect isn’t about any particular domain of knowledge, it’s about calibration, the match between how sure someone feels and how often they’re actually right. Expertise can improve calibration where people receive repeated, high-quality feedback, but experts can still become overconfident when making judgments outside the specific conditions where their expertise has actually been tested and corrected over time.

Why Does It Matter?

Organizations rely on confident-sounding estimates constantly, revenue forecasts, project risk assessments, technical judgments about whether something will work, and confidence is often the only signal available for how much weight to give an estimate, since the underlying accuracy usually can’t be checked until much later, if ever. When confidence is systematically inflated relative to accuracy, decisions get made as though estimates are more reliable than they actually are, and the miscalibration is invisible in the moment specifically because it feels like ordinary, reasonable certainty rather than like a bias.

This compounds in group settings. Because organizations often use expressed confidence as a proxy for competence or certainty, miscalibrated confidence can receive more influence than the underlying accuracy warrants, regardless of whether that confidence is actually justified by anything more than the speaker’s own miscalibration. An organization that systematically rewards certainty of presentation can accidentally reward overprecision, especially when forecast accuracy is never tracked afterward to check whether it was deserved.

What Changes Once You See It?

You start treating stated confidence as a separate piece of information from actual track record, and start checking a person’s or a team’s calibration history, how often their confident estimates have actually turned out right, rather than taking current confidence as self-validating.

You start asking explicitly for a range or a distribution rather than a single confident number, and then you track whether those ranges are actually calibrated over time, rather than assuming that expressing uncertainty in interval form automatically means the uncertainty was estimated well, people are notoriously prone to overly narrow intervals too. If someone repeatedly gives eighty percent confidence intervals, roughly eighty percent of outcomes should actually fall inside them, that comparison is a real, trackable measure.

You also get more skeptical of your own confident judgments specifically on the hardest, most uncertain questions, since that’s exactly where the gap between felt confidence and actual accuracy tends to be largest.

Common Misunderstandings

It isn’t the same as Planning Fallacy, though the two often show up together and reinforce each other. Planning Fallacy is specifically about underestimating the time, cost, and risk of a project because the estimate is built from the plan itself rather than from comparable track records, Overconfidence Effect is the broader miscalibration between stated confidence and actual accuracy across judgments generally, not only project estimates.
It isn’t the same as the Dunning-Kruger Effect, though both involve inflated self-assessment and the two are frequently confused. Dunning-Kruger concerns misjudgments of one’s own relative competence or performance in a domain, Overconfidence Effect here refers to being more certain that a specific judgment is correct than subsequent accuracy justifies, it shows up broadly, including among genuine experts, on judgments that are simply hard to calibrate well.
It isn’t the same as Optimism Bias. Optimism bias concerns expecting outcomes to turn out unusually favorably, overconfidence concerns being too certain that one’s judgment, favorable or unfavorable, is correct.
It isn’t a claim that confidence is always misplaced. Well-calibrated confidence, built from real track record and honest feedback, is genuinely informative, the concern is specifically the systematic gap that opens up between stated confidence and actual accuracy when that calibration hasn’t happened.
It doesn’t mean the fix is simply to sound less confident. Hedging language without any actual improvement in calibration doesn’t close the gap, the fix is checking confidence against outcomes over time, not performing uncertainty.

Diagnostic Question

How often, historically, has this level of stated confidence actually turned out to be right?

Explore Further

Field Notes

None yet.

Related Field Guide

Origin

Sarah Lichtenstein, Baruch Fischhoff, and colleagues conducted influential calibration studies through the 1970s, comparing people’s stated confidence with the actual accuracy of their judgments and documenting systematic overconfidence in many difficult judgment tasks. The work became a major part of the broader judgment-and-decision-making literature on calibration and overprecision, and was later given wide popular treatment in behavioral economics.