Explain herding and how it links to market anomalies
The idea
Herding is following the crowd in place of independent analysis, and it is not always irrational at the individual level: a manager who is wrong alongside everyone else keeps their job, while one who is wrong alone does not. Aggregated, it produces momentum, bubbles and crashes, since prices move on flows rather than on information. Anomalies such as momentum and post-earnings drift are consistent with herding and underreaction, though a persistent anomaly is always open to the competing explanation that it is compensation for a risk not yet identified.
A practice question
What is herding in an investment context?
- Following the actions of others rather than one's own analysis — correct
- Holding a diversified portfolioWhy not: A deliberate risk management choice, made independently of what others are doing.
- Investing only in index fundsWhy not: A considered decision about cost and efficiency, not an abdication of analysis to the crowd.
The distinguishing feature is that the crowd's behaviour, not the evidence, drives the decision.
A harder one
Why can herding be individually rational for a professional manager?
- Because consensus positions always outperformWhy not: They do not, which is what makes the behaviour costly for clients even while it is safe for the manager.
- Because the career cost of being wrong alone exceeds that of being wrong with everyone else — correct
- Because regulations require consensus positioningWhy not: No such requirement exists; managers are free to hold whatever their mandate permits.
The asymmetry is in the manager's own payoff, not the client's. It is a principal-agent problem as much as a behavioural one.
The hardest tier
A documented anomaly persists for decades. What are the two live explanations?
- Either transaction costs have fallen, or the anomaly was never realWhy not: Falling costs would erode an anomaly rather than sustain it, and the premise is that it has been documented.
- Either the market is inefficient, or the data are fabricatedWhy not: Fabrication is not a serious candidate for a result replicated across markets and decades.
- Either an unmeasured risk, or limits to arbitrage — correct
It may compensate a risk not captured by the model used to test it, or behavioural biases may be persistent while limits to arbitrage prevent their elimination. Both are serious: the behavioural explanation requires that arbitrage be constrained, since otherwise the anomaly would be traded away, and those limits are themselves observable.
In the app: 9 practice questions on this outcome, a narrated walkthrough with diagrams, spaced review that brings it back before you forget it.
Take the free diagnostic →