Overconfidence is usually described as a personality trait. Some people have it, some don’t, and the ones who do trade too much.
The research points somewhere more interesting. Overconfidence appears to be dynamic rather than fixed, rising and falling in response to recent results, and the mechanism that drives that variation is how outcomes get remembered and explained rather than what the outcomes actually were.
That makes it a loop rather than a trait, and loops can be interrupted at specific points.
Three Biases That Reinforce Each Other
The psychology of investing literature identifies several tendencies that combine to sustain confidence regardless of results.
They operate in sequence:
- Self-attribution, where good outcomes are credited to skill and poor ones to circumstances
- Hindsight bias, where past events come to seem more predictable than they were
- Confirmation bias, where subsequent information is filtered toward the existing view
- Selective recall, where the successful decisions are more available than the unsuccessful ones
Each alone is manageable. Together they form a system where evidence that should reduce confidence gets reprocessed into evidence that supports it.
The Market-Level Evidence
The most striking support comes not from surveys but from aggregate trading data, where the pattern is visible at scale.
Research testing the trading volume predictions of formal overconfidence models found that share turnover is positively related to lagged returns for many months, with the relationship holding for both market-wide and individual security turnover.
The logic connecting those is direct. If investors attribute market gains to their own skill, their confidence rises after good periods, and higher confidence produces more trading. Turnover rising after returns rise is what that mechanism predicts, and it’s what the data shows.
Note what’s absent from that chain. Nothing about the investors’ actual skill changed. Only their assessment of it did, and only in response to a market-wide move they didn’t cause.
The same research interprets the individual-security version of the pattern as evidence of the disposition effect, the tendency to sell winners and hold losers. Two distinct biases showing up in the same dataset, distinguishable by whether the turnover follows the market or the individual holding.
Why the Loop Doesn’t Self-Correct
The obvious question is why experience doesn’t eventually correct the assessment.
Research examining the persistence of the pattern points at the same set of tendencies, noting that self-attribution, hindsight and confirmation bias may be to blame for the persistence of overconfidence, and grouping optimism, self-attribution and hindsight together as forms of self-deception.
The mechanism is that each bias protects the others. A losing trade attributed to bad luck doesn’t count as evidence about skill. A market move that seemed obvious afterwards doesn’t register as something that was uncertain beforehand. And information that would challenge the assessment gets weighted less than information supporting it.
Feedback that never reaches the assessment can’t correct it, however much of it arrives.
What Makes This Different From Ordinary Optimism
Overconfidence in this specific sense isn’t the same as being generally positive, and the distinction matters for what can be done about it.
The research distinguishes several components: overestimation of one’s own performance, overprecision in the form of confidence intervals that are too narrow, and overplacement, the belief that one is better than average. These respond differently to experience. Past success has been found to strongly affect the first two while leaving the third largely unchanged.
That’s a useful detail. After outperforming, investors don’t necessarily believe they’ve become better than other people. They estimate their own future returns too highly and express those estimates with too much certainty, which is the version that leads to larger positions and tighter stop assumptions.
Interrupting the Loop
The intervention points follow from where the loop is weakest:
- Record the reason before the outcome, since attribution is only testable against a written prior
- Note the confidence level at the time, which makes overprecision measurable later
- Review decisions in batches, because patterns are visible across twenty trades and invisible in one
- Log the trades you didn’t make, since selective recall omits the avoided losses and the missed gains equally
- Watch turnover after good periods, as rising activity following gains is the observable signature of the loop
The fifth is the most practical. An investor whose trading frequency rises after a strong market period has direct evidence the mechanism is operating in their own behaviour, without needing to introspect at all.
What Can Realistically Be Improved
Awareness alone does little. These tendencies operate below the level where knowing about them helps, which is why financially sophisticated investors show them too.
What does change is the availability of contrary evidence. A written record of reasoning and confidence, reviewed in batches, makes self-attribution difficult to sustain because the alternative explanations were recorded before the outcome was known.
That’s a modest intervention with a modest effect. It’s also the only one the research supports, and it costs a few minutes per decision.
