Herding
Following what others are doing. It is frequently rational for the individual and produces outcomes that are collectively poor.
MadStockAlerts Research · Updated August 28, 2026
What to take away
- Others' behaviour is genuine information, which is why herding is not simply irrational.
- An information cascade forms when people act on others rather than on their own signal.
- Career and reputational incentives reinforce it among professionals.
- Crowded positions unwind together, which is a real and measurable risk.
- The defence is a process defined before the crowd formed.
MAD Academy Training Video · 0:44
Comfort Is Not Confirmation
Agreement feels like evidence and is not, and crowded positions are exactly the ones with the narrowest exit.
This lesson is part of a Stock Alerts + Tools plan.
Why it is not simply an error
If many people are doing something, some of them may know something. Weighting that as evidence is reasonable, and it is what makes herding difficult to dismiss as a mistake.
The failure is a cascade. Once people act on what others are doing rather than on their own information, the crowd's size stops reflecting the information behind it, and the apparent evidence is people copying each other.
- 1A few act on their own informationWhich is genuine evidence
- 2Others observe them and followReasonably, since the behaviour is informative
- 3Later entrants observe only the crowdTheir own information is never expressed
- 4The crowd's size stops carrying information
The professional version
For a professional manager the incentives are asymmetric. Being wrong alongside everyone else is survivable; being wrong alone is not, whatever the reasoning was.
That asymmetry is a rational response to how performance is evaluated, and it produces convergence in positioning that has nothing to do with any assessment of value. Keynes described the mechanism in the 1930s and it has not changed.
Crowding as a measurable risk
- A position held widely by similar investors will be exited by all of them at once under pressure.
- The exit occurs into a market where the natural buyers are the same people trying to sell.
- This is the mechanism behind momentum crashes and behind factor drawdowns.
- Ownership data provides a partial view of crowding, with the lags the 13F article describes.
The defence
A process defined before the crowd formed is the only reliable defence, because the pressure operates in the moment. That is the same conclusion the written plan, the saved layout and the price alert articles all reach from different directions.
It is also worth being explicit that going against a crowd is not itself a strategy. A crowded position can be correct, and contrarianism as a reflex is the same error with the sign reversed.
Measuring crowding
Crowding is a real risk and it is partially observable, with the lags the disclosure regime imposes.
- The number of institutional holders and the change in it, from the quarterly filings.
- Short interest, for the other side of the same question.
- Concentration of ownership among similar investors, which the filings partially reveal.
- Valuation spreads within a factor, which indicate how much has been paid for that exposure.
- Correlation between a position and others of the same type, which rises as crowding does.
Every one of these is lagged, partial, or both. They establish that a position is widely held rather than measuring how quickly it would be exited, which is the property that actually matters.
The most direct evidence arrives too late to act on: a crowded position unwinding is observable in the price, and the observation and the unwind are the same event.