Heads I'm a Genius, Tails I Was Unlucky: Overconfidence and Self-Attribution

In the same year that Barberis, Shleifer and Vishny published their model of investor sentiment, a rival team took a swing at the same problem with a completely different pair of biases. The puzzle was identical: why do markets underreact to news in the short run (momentum) and overreact in the long run (reversal)?

BSV blamed conservatism and pattern-seeking. Daniel, Hirshleifer and Subrahmanyam blamed something far less flattering and, if you are honest, far more familiar: ego.

The problem: a bias that gets stronger the more you use it

DHS build their model on two psychological facts, and the second one is the killer.

Bias one: overconfidence. People overrate the precision of their own private information. Not their information in general, specifically the information they dug up themselves. If you spent the weekend building a model of a company, you believe that model far more than you believe the same conclusion handed to you by someone else. Your work feels solid. Everyone else's feels like guesswork.

Crucially, DHS assume investors are overconfident about private signals (their own analysis) but not about public signals (the earnings release everybody sees). This asymmetry is not arbitrary, it is exactly what the psychology literature finds. You do not overrate a number printed in the newspaper. You overrate the thing you figured out yourself.

Bias two: biased self-attribution. This is the one that turns a static error into a spiral. When outcomes confirm you were right, you credit your skill. When outcomes prove you wrong, you blame bad luck, or the market being irrational, or an unforeseeable shock.

Notice what this does to your confidence over time. Good news raises it a lot. Bad news lowers it only a little, or not at all. Confidence ratchets upward. It is a one-way valve. This is why the same trader can survive years of mediocre results and emerge more certain of himself than when he started.

The key idea via analogy: the poker player on a heater

Picture a poker player who is convinced he can read faces.

He decides, based on his own private read, that his opponent is bluffing. He calls. He wins. "I knew it. I read him perfectly." His confidence in his face-reading soars. It never occurs to him that the opponent might have simply had a worse hand by chance.

Next hand, he reads a bluff again. He calls. He loses. "Terrible luck, he hit his card on the river. My read was right, the deck betrayed me." Confidence barely dented.

Now run this for an hour. His confidence has climbed steadily, because wins are counted as skill and losses are written off as variance. He is now betting enormous sums on reads that were mediocre to begin with, and he is more certain than he has ever been. And then the correction arrives all at once, when reality finally cashes him out.

That is the DHS market, and here is how it produces the two anomalies:

Overconfidence gives you overreaction. An investor gets a private signal (his research says the stock is worth more). He overrates it, and trades too aggressively on it, so the price overshoots past fair value immediately. Eventually the truth comes out through public information, the overshoot is corrected, and the price falls back. Overreaction and long-run reversal, straight out of the box. It also generates excess volatility, because prices are being pushed around by overrated private signals.

Self-attribution gives you momentum, and this is the beautiful part. After the investor buys on his private signal, some public news arrives. Suppose the public news confirms his view. What does he do? He credits his own genius. His confidence in his private signal goes up. So he trades even more aggressively in the same direction and pushes the price even further in the direction it was already going.

Read that again, because it is the paper's crown jewel: public news that confirms an existing private view causes the price to continue moving, rather than to settle. That is momentum. It is continued overreaction, driven by the confirming news feeding the ego of the person who was already positioned.

And if the public news contradicts him? He shrugs it off as noise. Confidence barely falls. So the price does not fully correct on bad news either, which is the same asymmetry that produces post-earnings drift.

So DHS get momentum not as an underreaction, as BSV did, but as a continuing overreaction driven by self-flattery. The price keeps drifting in the direction of the initial move because confirming news keeps inflating the confidence of the people who caused the move. And then, in the long run, cold hard reality eventually asserts itself and the whole edifice reverses.

Why it mattered

• It reframes momentum as a psychological chain reaction. BSV say momentum is the market being slow. DHS say momentum is the market being pumped up. These are very different mental pictures with different implications: in the DHS world, momentum crashes should be violent, because the reversal is unwinding an overshoot rather than completing an adjustment. Momentum crashes are, in fact, notoriously violent, which is a point for DHS.
• It explains why trading volume and confidence move together. Overconfident investors trade too much. This links the model directly to Odean's empirical work showing that the more people trade, the worse they do, and to Barber and Odean's finding that men, who psychology says are more overconfident in finance, trade far more than women and lose more to costs.
• It survives the "professionals aren't dumb" objection better than most. You can tell a professional analyst that he is stubborn, and he will deny it. Tell him that he trusts his own model more than a stranger's, and takes credit when he is right, and he will nod, because everyone does this and it feels like competence rather than bias. That is exactly what makes the bias durable.
• It predicts a public-event pattern nobody had looked for. The model implies specific price behavior around public announcements that depends on whether the news confirms or contradicts prior private trading. That is a sharp, testable prediction, not a story fitted after the fact.

The honest limitations

• Same fundamental critique as BSV: the biases were selected to fit. Overconfidence and self-attribution are both real and well-documented. They are also two biases out of many, chosen because they generate the known facts. There is no independent measurement in the paper telling you these particular biases are the ones driving prices.
• The competing models are hard to tell apart. BSV, DHS, and Hong-Stein all predict momentum and reversal. Distinguishing them empirically requires finding a situation where they disagree, and the differences are subtle enough that twenty-five years of empirical work has not delivered a knockout. Three elegant explanations for the same fact is two too many.
• There are no arbitrageurs. As with BSV, the model has no rational, well-capitalized trader to lean against the overconfident crowd. That such traders exist and do not fully correct the price requires the limits of arbitrage, which is a separate paper and a separate argument.
• It does not tell you when confidence pops. The reversal arrives when public information cumulatively overwhelms the private delusion. When is that? The model does not say. Which means, as usual, that behavioral finance can tell you the trade is wrong and cannot tell you when it will stop being wrong.

The one-line takeaway

Daniel, Hirshleifer and Subrahmanyam showed that investors who overrate their own research and then take credit for every win while blaming luck for every loss will build an ever-inflating confidence spiral, pushing prices further and further in the direction they already moved (momentum), until reality eventually forces the whole thing to unwind (long-run reversal).