Boys Will Be Boys: Overconfidence, Gender, and the Cost of Trading

Behavioral finance has a chronic problem, and its critics never let it forget: the biases are hard to measure directly.

You want to claim that overconfidence causes investors to trade too much and lose money. Fine. But you cannot open up someone's skull and read a confidence meter. All you can observe is the trading. So the argument risks becoming circular: they trade a lot, therefore they are overconfident, and we know they are overconfident because they trade a lot. A hostile referee will tear that apart, and should.

Barber and Odean's 2001 paper is a masterclass in escaping that trap. Rather than trying to measure overconfidence, they found a group that psychology had already established was more overconfident, for reasons having nothing to do with finance, and then tested whether that group behaved as the theory predicted.

The problem: you need a variable that predicts confidence but is not confidence

The setup is a genuine piece of scientific craftsmanship, so it is worth laying out slowly.

Step one: import an established fact from another field. The psychology literature had long documented that overconfidence is not uniform. It is strongest in domains that are perceived as masculine, and finance is very much perceived that way. In such domains, men are reliably more overconfident than women. This was not Barber and Odean's finding. It was already sitting in the psychology journals, established by researchers who had never thought about the stock market. That independence is the whole point.

Step two: derive a prediction that has nothing to do with returns. If overconfidence causes excessive trading, then men should trade more than women. That is a sharp, falsifiable prediction, and note it says nothing about whether men are better or worse investors. It is purely about volume.

Step three: derive the consequence. And if excessive trading destroys returns through costs (which their 2000 paper had already established), then men's extra trading should cost them more.

Step four: check. They took the accounts of over 35,000 households at a large discount brokerage, from February 1991 to January 1997, and split them by the gender of the account holder.

The key idea via analogy: two drivers, same road

Imagine two people driving the same route to work every day. One is convinced he is an exceptional driver. He changes lanes constantly, hunting for an advantage, weaving to gain a few seconds. The other just picks a lane and stays in it.

They arrive at roughly the same time. But the weaver has burned more fuel, worn his tyres faster, and had a couple of near misses. Same destination, more cost, more risk, and a sincere belief that he was going faster.

That is exactly what the data showed.

Men traded about 45 percent more than women. That is not a subtle statistical whisper, it is a chasm. Nearly half again as much lane-changing.

And it did not help. It hurt. Trading reduced men's net returns by about 2.65 percentage points a year. For women, the drag was about 1.72 percentage points. Both groups were damaged by their own activity. Men were damaged substantially more, because they did substantially more of it.

Note carefully what this does and does not say. It does not say women are better stock pickers. Both groups' trades were roughly value-destroying. The gap is not about who picks better, it is about who trades more, and therefore who pays more of the toll. Women underperformed too. They just underperformed less, because they did less.

Barber and Odean then tightened the screw with a lovely sub-result. If overconfidence in finance is the mechanism, then the effect should be strongest among people who are single, because a married man's trading might be moderated by the influence of his spouse, and a single man's is not. The prediction: single men should trade the most and lose the most of all. That is what the data showed. Single men traded more than married men, who traded more than married women, who traded more than single women, and the return damage followed the same ladder.

That gradation is what makes the paper convincing. A single difference between two groups could be explained a hundred ways. A monotone ordering across four groups, predicted in advance by an outside theory, is much harder to dismiss.

Why it mattered

• It solved behavioral finance's identification problem, at least once. By using an exogenous, pre-established psychological fact as an instrument for overconfidence, Barber and Odean broke the circularity that plagues the field. This is the template: find a variable that predicts the bias for independent reasons, then test the bias's financial prediction. It is much more rigorous than the usual "here is a pattern, here is a bias that could explain it".
• It gave the overconfidence theories their best empirical support. Daniel, Hirshleifer and Subrahmanyam's whole model rests on overconfident traders trading too aggressively. This is the closest thing to a direct field test of that assumption.
• It became one of the most cited papers in finance and one of the most misquoted. The finding is not "women are better investors". The finding is that overconfidence causes overtrading, overtrading is expensive, and men do more of it. The prescription is not "hire women", it is "trade less", and it applies to everyone.
• It is genuinely actionable. The practical takeaway is uncomfortable and useful: your conviction that this trade is a good idea is, statistically, not evidence that it is a good idea. It is evidence that you are confident. Those are different things, and only one of them costs you a spread.

The honest limitations

• Gender is a proxy, and proxies are leaky. Gender predicts overconfidence on average, but it also correlates with income, risk tolerance, financial education, occupation, and a dozen other things that might independently drive trading frequency. Barber and Odean control for what they can, but the instrument is not clean. This is the paper's most serious weakness and its authors are reasonably upfront about it.
• Account holder gender is not necessarily who is trading. The gender comes from the name on the account. In a household, the person placing the trades may not be the person on the paperwork. This adds noise, though it would tend to blur the effect rather than manufacture it.
• The cost environment has changed enormously. These figures come from a world of real commissions and wide spreads. Today, trading is nearly free, so the cost channel is far weaker. The behavior has not changed at all, and modern research on trading apps finds the same overtrading, but it now pays for itself in different ways: bad execution, poor timing, options premium, and leverage.
• It is one dataset, one country, one period. As with all of the Barber and Odean work, it is a single discount broker in the United States in the 1990s. The finding has replicated in other countries, which helps, but the original result is not a global sample.

The one-line takeaway

Barber and Odean used an established psychological fact, that men are more overconfident than women in domains seen as masculine, to make an advance prediction about trading, and found that men traded about 45 percent more and paid for it with a larger annual drag on returns, which is about as clean a demonstration as finance offers that confidence is expensive.