Herding Is Rational for the Individual and Dangerous for Everyone
Following the crowd is often the sensible personal choice, especially when your job depends on it. That is precisely why it produces collective outcomes nobody wanted.
Two Kinds of Herding
Informational herding occurs when people infer that others know something they do not. Seeing many participants buy an asset is treated as evidence about the asset rather than as evidence about the participants.
This produces an information cascade. Once enough people have acted, later participants rationally weight the observed behaviour above their own private information, and stop contributing new information to the market. The crowd grows while the evidence underneath it does not.
Reputational herding is different and, in professional settings, more powerful. It occurs when the consequences of being wrong depend on whether others were wrong at the same time.
The Career Calculation
Consider a portfolio manager choosing between a consensus position and a contrarian one.
| Position | If right | If wrong |
|---|---|---|
| With consensus | Performed in line | Everyone was wrong, tolerable |
| Against consensus | Substantial credit | Alone and exposed |
The payoff is asymmetric in a way that has nothing to do with the merits. Being wrong alongside everyone else is survivable. Being wrong alone is a career event.
It is better for reputation to fail conventionally than to succeed unconventionally. That observation is decades old and the incentive structure it describes has not changed.
Benchmark measurement sharpens this further. A manager evaluated against an index takes on real risk by deviating from it, so the largest constituents get held largely because not holding them is the riskier professional choice.
How It Shows Up
Crowded positions form where many funds hold the same names for the same reasons. The positions look diversified across institutions and are in fact identical, which means the exit is far narrower than the entrance.
Analyst estimates cluster tightly, and revisions arrive in waves after a company reports rather than before. Forecasting away from consensus carries the same asymmetric risk as positioning away from it.
Sector rotations run further than fundamentals justify in both directions, because participation is partly a function of participation.
The Unwind
The danger is not the accumulation, it is the reversal. When a crowded position turns, everyone attempts to reduce simultaneously, and liquidity that appeared adequate when measured position by position turns out to have been the same liquidity counted many times.
Forced selling by one participant moves the price against the others, triggering their risk limits, producing more selling. The mechanism is mechanical rather than emotional, and it runs faster than discretionary decision making.
This is why measures of crowding are watched as risk indicators. A position held by many similar investors carries a risk that does not appear in its own volatility history.
Why It Cannot Simply Be Fixed
Telling professionals to think independently ignores the incentives they face. The manager who deviates and underperforms for two years loses the mandate before any long term judgement is possible.
The structural fixes address the incentive rather than the behaviour: longer evaluation periods, absolute rather than relative return objectives, and capital that will not withdraw during a period of divergence. All three are difficult to arrange, which is why herding persists in an industry populated by people who know exactly what it is.
The Bottom Line
Herding arises from inference about what others know and from career incentives that punish being wrong alone far more than being wrong together. Both are individually rational, which is why the aggregate outcome is so persistent. The risk it creates is concentrated in the unwind, when a position held identically across many investors has to be reduced by all of them at once.