Caught in the Act: Odean Finds the Disposition Effect in 10,000 Real Accounts

Shefrin and Statman had proposed the disposition effect in 1985: the idea that investors sell winners too eagerly and hold losers too long. It was a lovely theory. It matched what every broker had ever observed. It fit prospect theory beautifully.

But it had a problem. To prove it, you need to know two things about every individual investor: what they bought each stock for, and what they did with it afterward. That data lived in brokerage back offices, and brokerages do not hand it out.

Terrance Odean got the data. And what he did with it is one of the cleanest demolitions in behavioral finance.

The problem: you cannot see the counterfactual

Here is the subtlety that makes this hard, and it is the kind of thing that separates good empirical work from bad.

Suppose you notice that investors sell more winners than losers in your data. Is that proof of anything? No. Because in a rising market, most people hold mostly winners. If 80 percent of your positions are up, then of course 80 percent of your sales are winners. That is arithmetic, not psychology.

So Odean's key methodological move is to build the right denominator. For each day an investor makes a sale, he asks: of all the positions this person could have sold today, what fraction of the available gains did they actually realize, and what fraction of the available losses did they actually realize?

This gives two clean numbers:

The proportion of gains realized. Of all the winning positions available to sell, how many got sold?
The proportion of losses realized. Of all the losing positions available to sell, how many got sold?

If investors are indifferent to gains and losses, these two numbers should be roughly equal. Any gap is the disposition effect, measured cleanly, with the market's direction controlled away.

His data: the trading records of around 10,000 accounts at a large discount brokerage, tracked over several years. Real people, real money, real decisions, no laboratory.

The key idea via analogy: the two doors

Imagine every day an investor stands in front of two doors. Behind door one are their winning positions. Behind door two, their losers. They must decide which door to open and sell from.

Odean's finding is stark: investors walk through the winners' door far more often than the losers' door, and they do it consistently, month after month, in every part of the sample. The proportion of gains they realized was substantially higher than the proportion of losses. Same investor, same day, same information, wildly different willingness to pull the trigger depending on nothing but whether the position was above or below what they paid for it.

Now Odean does the thing that turns this from a curiosity into an indictment. He asks: were they right?

Maybe the investors were being smart. Maybe they sold the winners because those winners were overvalued and about to fall, and held the losers because those losers were undervalued and about to bounce. That is a perfectly coherent contrarian strategy, and if the subsequent returns backed it up, the disposition effect would be vindicated rather than condemned.

So he tracked what happened next. And the answer is brutal.

The winners they sold went on to outperform the losers they kept. Not by a rounding error. By a meaningful margin over the following year. Investors were systematically selling their better stocks and holding their worse ones. The disposition effect is not a contrarian strategy. It is an anti-strategy, a machine for pruning the flowers and watering the weeds.

And it gets worse when you add tax. In a taxable account, selling a winner triggers a tax bill and selling a loser generates a valuable deduction. So investors were:

Paying taxes they did not have to pay.
Forgoing deductions they were entitled to.
Selling the stocks that subsequently did better.
Keeping the stocks that subsequently did worse.

Four separate own goals, all from the same reflex.

Odean found one more detail that is almost funny. The effect disappears in December. Suddenly, in the final month of the tax year, investors become perfectly willing to sell their losers, because the tax benefit becomes salient enough to override the psychology. So the capacity for rational behavior is fully present. It just needs an annual deadline and an accountant to activate it.

Why it mattered

It converted a theory into a fact. Before Odean, the disposition effect was a plausible story supported by anecdote. After Odean, it was a documented regularity in the behavior of thousands of real investors, robust to the obvious statistical objections. It has since been replicated in Finland, Israel, China, Taiwan, on individual stocks, on funds, on housing, on futures, and among professionals.
It ruled out the polite explanations. By showing the sold winners subsequently outperformed, Odean killed the "rational mean reversion" defence. By showing the effect vanishes in December, he killed the "it's about taxes" defence, since it demonstrably was not, except for one month a year.
It is the empirical bedrock under a whole literature. The unrealized-gain overhang in a stock's investor base is now used as a genuine predictor of returns. If most holders are sitting on losses, they refuse to sell, so selling pressure is suppressed and bad news enters the price slowly. That is a micro-foundation for momentum built directly on Odean's finding.
It is unpleasantly personal. Almost every reader of this paper has done exactly what it describes, and can verify it on their own statement in ten minutes. Very few papers in finance offer that.

The honest limitations

It is one broker, one country, one period. A large discount brokerage's clients in the early 1990s are a specific population: self-directed, relatively active, American. Later replication has been extensive and supportive, but the original result was a single sample.
It cannot see the whole investor. Odean observes one account. If someone also has a retirement account, an advisor-managed portfolio, or real estate, then what looks like an irrational hold in this account might be part of a coherent overall plan. Unlikely at this scale, but not impossible.
Documenting it is not the same as trading it. The disposition effect is real and it is a hint that stocks with a lot of trapped losses may have suppressed selling pressure. Turning that into a profitable, cost-surviving strategy is a completely separate problem, and this paper does not attempt it.
The subsequent-return result depends on the horizon. The finding that sold winners beat held losers is measured over a specific window. Behavioral results like this are sensitive to horizon choice, and a careful reader should want to see the full curve rather than one number.

The one-line takeaway

Odean opened up 10,000 real brokerage accounts and found that investors are dramatically more willing to sell a stock that is up than one that is down, that the winners they sold went on to beat the losers they kept, and that they were paying extra tax for the privilege, which is about as close as finance gets to catching a bias red-handed.