News Travels Slowly, Traders Chase It: Hong and Stein's Unified Theory

By 1999, two teams had already offered behavioral explanations for the great puzzle of short-run momentum and long-run reversal. Barberis, Shleifer and Vishny blamed conservatism and pattern-seeking. Daniel, Hirshleifer and Subrahmanyam blamed overconfidence and ego.

Both models rest on a specific cognitive bug. Both require you to accept that investors' brains are wired wrong in a particular, laboratory-documented way.

Harrison Hong and Jeremy Stein came at it from an angle that is, in a certain sense, more subversive. Their investors have no psychological biases at all. Nobody is overconfident. Nobody misjudges probabilities. Their investors are simply limited: each one can only pay attention to some information, and can only run a simple strategy. That is it. That is the entire behavioral assumption.

And out of that modest, almost undeniable assumption, both anomalies fall out.

The problem: maybe the bug is not in the brain

The prior models put the error inside the investor's head. Hong and Stein put it in the structure of the market, specifically in who knows what, and when.

They populate their market with exactly two types of trader, and both are what economists politely call "boundedly rational", meaning not stupid, just narrow.

Type one: newswatchers. These people do fundamental research. They look at the company, the industry, the filings. They form a view of the value and trade on it. Their limitation: they completely ignore prices. They do not look at the chart. They do not care what the stock has done. They just do their homework and trade the conclusion.

Type two: momentum traders. These people do no fundamental research whatsoever. They look only at recent price moves and trade on them. Their limitation: their strategy is simple. They can only condition on a short window of past returns. They cannot run a sophisticated rule that knows when to stop.

Neither type is irrational, exactly. Each is doing something reasonable given what they choose to look at. They are just each half-blind, and it turns out that two kinds of half-blindness combine into something worse than either.

The key idea via analogy: the rumour and the bandwagon

Here is the whole model as a story.

Stage one: the news leaks slowly. This is Hong and Stein's key assumption, and it is entirely plausible. Information about a company does not hit everyone at once. It diffuses gradually across the population of newswatchers. A few insiders and specialists learn it first, then the regional analysts, then the broader market, over weeks or months.

Now, if the newswatchers all traded on their own information and watched the price, the first few would move the price and everyone else would learn from it instantly. But by assumption, newswatchers do not look at prices. So each one trades only when the news personally reaches them. The result: the price creeps up gradually, in a slow trickle, as more and more newswatchers get the memo. The market underreacts at first, then keeps adjusting for a long time. That is post-earnings drift and the beginning of momentum, produced with zero psychology.

Stage two: the trend attracts chasers. Now the momentum traders notice. They see the price has been rising. They do not know why and do not care. They buy.

Here is the elegant bit. The momentum traders are, in a sense, doing something socially useful: they are speeding up the incorporation of the news that the newswatchers are absorbing too slowly. Their trend-following is a crude arbitrage of the underreaction, and their profits are earned honestly.

Stage three: the bandwagon overshoots, and it is mathematically unavoidable. But recall the momentum traders can only run a simple strategy. They cannot tell whether today's price rise reflects genuine news still diffusing, or simply the buying of other momentum traders who arrived yesterday. To them, both look identical. It's a price going up.

So the early momentum traders' buying pushes the price up, which triggers later momentum traders to buy, which pushes the price up further, which triggers more. Each wave is responding to the previous wave rather than to any news. The fundamental information ran out a while ago, but the bandwagon is now self-sustaining.

The price sails past fair value. Overreaction. And eventually there are no more momentum traders left to recruit, the buying stops, and the price sinks back to where the fundamentals said it should be all along. Long-run reversal.

The deep and slightly tragic point: the very traders who correct the underreaction are the ones who cause the overreaction. They cannot do the first without doing the second, because their strategy is too simple to know when to stop. The cure and the disease are the same people. And the later momentum traders, the ones who show up at the end of the trend, systematically lose money to the early ones. Being a trend follower is profitable only if you are early.

Why it mattered

• It sets a much lower bar for behavioral finance. You do not need to believe investors are cognitively broken. You only need to believe that information spreads gradually and that some people follow trends. Both of those are obviously, trivially true. This makes the model very hard to dismiss, which is exactly why it is important.
• It produces sharp, testable, un-obvious predictions. If momentum comes from slow information diffusion, then momentum should be stronger in stocks where information travels more slowly. Hong and Stein predicted exactly this, and follow-up work (Hong, Lim and Stein, 2000) found that momentum is indeed stronger among stocks with low analyst coverage, especially among the losers, where bad news travels slowest because nobody has an incentive to spread it. That is a genuine prediction confirmed out of sample, which is worth ten elegant models.
• It gives momentum traders a role in the ecosystem. They are not parasites. They are the mechanism by which slow-moving information gets into prices. They just overdo it, structurally, because their strategy is too crude to self-limit.
• It explains why momentum crashes. The unwinding at the end of a trend is not a gentle adjustment, it is a bandwagon reversing, and everyone is on the same side of it. Momentum's notorious tendency to blow up violently is a natural consequence of this structure.

The honest limitations

• The core assumption is convenient. Newswatchers who never look at prices is a strong assumption. In reality, everyone looks at prices, and a fundamental investor who noticed the price already reflected his news would simply not trade. Hong and Stein need this blindness to prevent the price from immediately aggregating everyone's information, and it is doing an enormous amount of work.
• No rational arbitrageur shows up. A smart trader who understood this model could sit and wait for the overshoot, then short it, and make money while dampening the cycle. The model has no such agent. As with BSV and DHS, you have to import the limits of arbitrage from outside to explain why the mispricing survives.
• Three good models, one set of facts. Hong-Stein, BSV, and DHS all predict momentum then reversal. They cannot all be right, and telling them apart empirically is genuinely hard. The analyst-coverage evidence is the best discriminating test anyone has produced, and it favors Hong-Stein, but it is one test.
• Momentum has been punishing lately. Whatever the mechanism, momentum strategies have suffered severe crashes (2009 being the canonical horror show) and periods of poor performance since publication. A model that explains why momentum exists is not the same as a model that lets you survive trading it.

The one-line takeaway

Hong and Stein showed you can get both momentum and reversal without anyone being psychologically defective: just let news spread slowly through a population that ignores prices, and let trend followers who cannot tell real news from other people's trend-following pile on top, and the market will underreact, then overshoot, then collapse back, entirely on its own.