Joachim Andersen

Accelerated Learning

Overconfidence Bias: Why Feeling Sure Is Not Being Right

A small leaning house of blank black cards with gold edges on a dark walnut table at night, beside an open brass pocket compass and a brass hourglass with golden sand, in warm golden light
The higher we build on certainty, the less it takes to fall.


Overconfidence Bias is the tendency to be more confident in my knowledge, abilities and predictions than the facts justify. It’s not the same as being confident. Confidence helps me act. Overconfidence means my certainty has drifted away from my actual track record.

I notice it most when I estimate how long something will take, or how sure I am about a forecast. My gut gives me a number and a feeling of certainty, and the feeling of certainty is usually the part that turns out to be wrong.

The essence of Overconfidence Bias is to ask: “How often have I been right when I felt this sure, and what would I see if I were wrong?”


Origin: Measuring Confidence Against Reality

Psychologists began measuring overconfidence in the 1960s and 1970s by asking people questions and then asking how sure they were of each answer. If people were perfectly calibrated, answers given with 90% confidence would be correct 90% of the time. Instead, in classic studies by Sarah Lichtenstein, Baruch Fischhoff and colleagues, answers that people were 100% certain of were wrong about 20% of the time.

In 2008, Don Moore and Paul Healy published The Trouble With Overconfidence, which showed that researchers had been using the same word for three different things. Separating them helps me see which kind of overconfidence I am dealing with.

In my own experience, the three often show up together. When I am excited about an idea, I tend to think it will be done sooner than it will, that I am better placed than others to pull it off, and that the range of possible outcomes is narrower than it really is. Each of these errors is small on its own, but together they can turn a reasonable plan into a risky one.


Key Concepts

  • Overestimation
    Believing we’re better, faster or more in control than we really are, such as thinking we will finish a task sooner than we will.

  • Overplacement
    Believing we’re better than others, even when everyone in the group believes the same. I write about one version of this, where the least skilled overrate themselves the most, in Dunning-Kruger Effect.

  • Overprecision
    Being too sure that we know the truth, which shows up as ranges and forecasts that are far too narrow.

  • Calibration
    The match between how confident we are and how often we’re right. Good calibration, not maximum confidence, is the goal.

  • Confidence Feels Like Knowledge
    A clear, coherent picture in our head can feel like evidence, even when it’s built on very little information.


Some Examples

  • Executives and Their Narrow Forecasts
    In a long-running survey, chief financial officers of US companies were asked to forecast the stock market return for the next year along with a range they were 80% sure it would fall within. In a 2013 study of a ten-year panel of more than 13,300 such forecasts, Itzhak Ben-David, Graham and Harvey found that the actual returns fell inside the executives’ ranges only 36% of the time. Firms with the most mis-calibrated executives also tended to invest more and use more debt.

  • Trading Too Much
    In a 2001 study of more than 35,000 households at a US discount brokerage, Brad Barber and Terrance Odean found that men traded 45% more than women. Trading reduced men’s net returns by 2.65% points a year, compared to 1.72% points for women. The authors argue that overconfidence explains the difference, since men tend to be more overconfident than women in areas like finance, and the extra trading mostly added costs.

  • The Thesis That Took Longer
    In a 1994 study, Roger Buehler, Dale Griffin and Michael Ross asked psychology students writing an honors thesis when they expected to finish. On average they predicted 33.9 days, and even their worst-case estimate was 48.6 days. They actually took 55.5 days, and fewer than a third finished by the time they had predicted. This kind of overestimation is often called the planning fallacy.

  • Entering a Crowded Market
    In a 1999 experiment, Colin Camerer and Dan Lovallo let participants decide whether to enter a market where only the top-ranked entrants would earn money. When the ranking was random, entrants earned money on average. When it depended on their own performance on a test of skill, so many people were confident they would rank near the top that the entrants as a group lost money. The authors called this reference group neglect: each person forgot that the other entrants were confident too.


Applying Overconfidence Bias

I cannot simply decide to be less confident, but I can make my confidence easier to check and harder to inflate. This is the simple process I follow:

  1. Keep Score
    I write down important predictions along with how confident I am, and later compare the results with what actually happened.

  2. Widen My Ranges
    When I give an estimate, I give a range instead of a single number, and then make it wider than feels comfortable.

  3. Start From the Outside View
    Before I estimate how long a project will take, I look at how long similar projects have actually taken, and adjust from there.

  4. Remember the Competition
    When I believe I have an edge, I ask who else is confident in the same opportunity and what they might know that I don’t.

  5. Build in a Buffer
    For decisions where being wrong is costly, I plan for a worse outcome than I expect, using a Margin of Safety.


Pitfalls and Limitations

  • Some confidence is necessary. People who never feel sure enough to act miss opportunities, and a degree of optimism can help me persist through difficult work.

  • It’s not the same everywhere. People can be underconfident on easy tasks and overconfident on hard ones, so the direction of the error depends on the situation.

  • Experience does not cure it. The executives in the forecasting study were experienced professionals, yet their ranges were far too narrow year after year.

  • It spreads in groups. When everyone around me is sure of the same plan, our shared confidence can feel like independent confirmation, even though we may all be relying on the same assumptions.

  • Confidence is rewarded. Others often trust people who sound certain, which can push me to sound more sure than I am, and then to believe it.


Related Mental Models

Overconfidence Bias reminds me that the feeling of being sure is not evidence that I am right. By keeping score of my predictions, widening my ranges and starting from how things have actually turned out for others, I can keep the confidence I need to act while bringing it closer to what I really know.



To learn more about mental models and philosophy,
check out: Wise: How to Attain a Deep Understanding of the World, History, Our Place in It, and How to Live Well



📚 Sources and References