Psychology › Thinking, memory & biases › Heuristics and biases
Overconfidence
1977· Finding or experiment
Well replicatedStrong, repeated evidence.
People consistently believe their knowledge, skills and forecasts are more accurate than they really are.
See it on the Psychology map →What it says
In 1977 Baruch Fischhoff, Paul Slovic and Sarah Lichtenstein found people were often wrong on answers they called near-certain. Asked for ranges they are 90% sure contain a true figure, people typically capture it only about half the time. In Ola Svenson's 1981 study, about 93% of American students rated themselves above-average drivers. Brad Barber and Terrance Odean found the most active traders trailed the market by about 6 percentage points a year (1991–1996).
Why it matters
It drives excess trading, overbudget projects and missed deadlines. Useful fixes: forecast in ranges, track past forecast accuracy, and ask what would prove you wrong.
The evidence
Well replicated across calibration tests, surveys and market data, including Barber and Odean's study of 66,000 US brokerage households. Its size varies: on very easy questions people can be underconfident.
Open questions
Whether overconfidence sometimes pays, by helping entrepreneurs start firms or leaders persuade, is debated; the Dunning–Kruger effect claims a contested special case.
Inside this (1)
Connections (2)
▸Overconfidence is one of the most robust Heuristics and biasesWell replicated
The connection
Overconfidence research grew alongside the heuristics program in Eugene, Oregon, where Lichtenstein, Fischhoff and Slovic worked and Kahneman and Tversky visited. Shortcuts like availability and confirmation help produce it: supporting reasons come easily to mind, so people feel surer than the evidence justifies.
Who showed it and when
Eugene, Oregon, 1970s, at the Oregon Research Institute and Decision Research; collected in the 1982 book Judgment under Uncertainty, edited by Kahneman, Slovic and Tversky.
The evidence
Calibration studies repeat across students, doctors, executives and forecasters; a 1980 study by Asher Koriat and colleagues found listing reasons against an answer reduces overconfidence.
Open questions
Some argue it partly reflects how test questions are chosen; with questions sampled representatively from real life, the effect shrinks.
▸Dunning–Kruger effect is a claimed special case of OverconfidenceMixed evidence
The connection
Overconfidence is the broad tendency to overrate our knowledge, skill and precision. Dunning and Kruger argued for a specific version: the least skilled overrate themselves most, because the skill needed to perform is the same skill needed to judge performance. In their tests, low scorers near the 12th percentile placed themselves around the 62nd.
Who showed it and when
Justin Kruger and David Dunning, Cornell University, 1999, testing undergraduates on logic, grammar and humor.
How solid is it?
The original four studies and many repeats show low performers overestimate most; but statistics alone, through regression to the mean, explain part of the pattern.
Open questions
Whether novices' large overestimates and experts' slight underestimates reflect different kinds of thinking, or the same general bias plus measurement noise.