Overconfidence Effect

Category: Probability & Belief

The Overconfidence Effect is the systematic gap between how sure you feel about your judgments and how often you are actually right.

How it works

You confuse the feeling of confidence with the fact of being right, and those two are only loosely correlated. When an answer comes to mind quickly and smoothly, your brain treats that fluency as a signal of truth, so you never go looking for the counterexample that would puncture it. You also selectively recall the reasons you are right and skip the search for reasons you might be wrong, which inflates certainty without adding accuracy. Crucially, you get almost no clean feedback: life rarely tells you "you were 90% sure and right 70% of the time," so the miscalibration never gets corrected. Moore & Healy (2008) split this into three flavors, overestimation (thinking you scored higher than you did), overplacement (thinking you beat other people), and overprecision (confidence intervals that are way too narrow), and the last one is the most stubborn.

Where you'll see it

  • **Extreme certainty, ordinary error.** Fischhoff, Slovic & Lichtenstein (1977) had people answer general-knowledge questions and rate their certainty. Even answers people marked with absolute certainty (p = 1.00) turned out wrong roughly 15 to 30% of the time, and at odds of 100:1 people were right only about 73% of the time. Being 'sure' bought them nowhere near 100% accuracy.
  • **Everyone is a better-than-average driver.** In Svenson's (1981) study, 88% of a U.S. sample rated themselves safer than the median driver in their own group, which is arithmetically impossible for most of them to be. A preregistered 2022 replication (Koppel et al.) reproduced it: 91% called themselves safer than average and 93% more skilled.
  • **Overconfident CEOs torch shareholder money.** Malmendier & Tate (2008) tracked CEOs who held their own in-the-money stock options past rational exercise instead of diversifying, a marker of overconfidence. Those CEOs made more acquisitions, and more value-destroying ones, overestimating their ability to run someone else's company.
  • **90% intervals that miss most of the time.** Across calibration studies summarized by Lichtenstein, Fischhoff & Phillips (1982), people asked for a 90% confidence range around an unknown quantity produce intervals that contain the true value only about half the time. Their 'almost certainly within this range' is closer to a coin flip.

Where it comes from

The effect was pinned down empirically by Baruch Fischhoff, Paul Slovic, and Sarah Lichtenstein in "Knowing with Certainty: The Appropriateness of Extreme Confidence" (Journal of Experimental Psychology: Human Perception and Performance, 1977), which showed that people expressing total certainty were routinely wrong, correct only about 70 to 85% of the time when they claimed to be sure. Lichtenstein, Fischhoff & Phillips then wrote the landmark review "Calibration of Probabilities: The State of the Art to 1980" (1982), consolidating the finding that subjective confidence systematically exceeds objective accuracy. Don Moore and Paul Healy's "The Trouble with Overconfidence" (Psychological Review, 2008) gave the field its modern three-type taxonomy of overestimation, overplacement, and overprecision.

How to counter it

Get calibrated, do not just get humble. Vague resolutions to "be less cocky" do nothing. Write down actual probabilities on real predictions ("70% this launches on time"), then months later score how often your 70% calls came true. When your 90% predictions hit 60%, you finally have data instead of a vibe.

Attack the interval, not the point. For any estimate, don't ask "what's my best guess," ask "give me a range wide enough that I'd be shocked to be outside it," then widen it, because overprecision is the most stubborn flavor. A useful trick from calibration training: force yourself to name two concrete reasons your answer could be wrong before you lock it in.

Import outside feedback. The forecasters who are well calibrated (weather forecasters, seasoned bridge players, oddsmakers) all share one thing: fast, repeated, scored feedback. Manufacture that loop with premortems ("assume this failed, why?") and by asking someone who disagrees with you to explain their reasoning before you commit.

The tell

Catch the phrase "I'm 100% sure" or "there's no way this is wrong" leaving your mouth, because near-total certainty is exactly where the accuracy gap is widest. Also notice when your estimate range feels comfortably tight and you skipped even trying to name how you could be wrong.

Related biases

Featured in

References

  1. Fischhoff, B., Slovic, P., & Lichtenstein, S. (1977). Knowing with certainty: The appropriateness of extreme confidence. Journal of Experimental Psychology: Human Perception and Performance, 3(4), 552-564
  2. Lichtenstein, S., Fischhoff, B., & Phillips, L. D. (1982). Calibration of probabilities: The state of the art to 1980. In D. Kahneman, P. Slovic & A. Tversky (Eds.), Judgment under Uncertainty: Heuristics and Biases (pp. 306-334), Cambridge University Press
  3. Moore, D. A., & Healy, P. J. (2008). The trouble with overconfidence. Psychological Review, 115(2), 502-517
  4. Svenson, O. (1981). Are we all less risky and more skillful than our fellow drivers?. Acta Psychologica, 47(2), 143-148
  5. Malmendier, U., & Tate, G. (2008). Who makes acquisitions? CEO overconfidence and the market's reaction. Journal of Financial Economics, 89(1), 20-43