Herding

Herding is usually modelled as an information cascade. Investors act sequentially, each holding a noisy private signal about value and each able to observe the trades that came before. A rational participant updates on both sources. Once a run of observed buying becomes strong enough, the public evidence of others' actions outweighs any single private signal, and the investor trades with the crowd regardless of what their own research says. From that point on their trade conveys no new information, so the next investor faces the same distorted evidence base and the cascade self-perpetuates.

The critical property is that private information stops being aggregated into price. Prices continue to move, but on recycled rather than new information, which makes the cascade fragile: a small credible contrary signal, an earnings miss or a failed financing, can reverse it abruptly and produce a decline far larger than any change in fundamentals would justify.

Institutional herding is driven less by naivety than by incentives. Managers are evaluated against a benchmark over short horizons, so the practical risk is not being wrong but being wrong alone. Deviating from consensus positioning increases tracking error and, if the deviation underperforms, threatens mandates and employment. Matching the crowd produces losses that are shared and therefore forgivable. The result is portfolios that cluster around index weights and around whatever positioning peers are known to hold.

The same pressures appear in sell-side research, where forecasts bunch tightly around the consensus estimate and revisions tend to follow the first mover rather than anticipate it. Fund flows reinforce the pattern: retail money chases trailing performance, forcing managers to buy more of what has already risen. Index inclusion events, crowded factor trades and quarter-end window dressing all leave visible fingerprints of the same behaviour.

Herding is one of the behavioural mechanisms invoked to explain why prices can depart from intrinsic value for extended periods, and it maps directly onto documented return patterns. Correlated buying pushes prices past fair value and sustains short-horizon momentum; the eventual correction produces the long-horizon reversal associated with overreaction. Neither pattern is consistent with a strictly efficient market, although both survive partly because limits to arbitrage such as borrowing costs, margin calls and mandate constraints deter the traders who would otherwise correct them.

Empirically, researchers infer herding from clustering in trades, from the tendency of institutions to buy and sell the same names in the same quarter, and from unusually low dispersion of individual returns around the market return during stress. Low dispersion suggests investors are trading the index rather than the security.

At Level I you should be able to recognise herding from behavioural evidence rather than from outcomes. Look for language indicating that a decision was made because others were doing it, that private analysis was overridden, or that the discomfort of holding an unpopular position drove the trade. Herding often travels with fear of regret, availability bias and social proof, so vignettes frequently combine them; the distinguishing feature is the observation of, and deference to, other participants' actions.

You should also be able to state the consequences: reduced information content in prices, excess volatility, momentum followed by reversal, and crowded positions that become illiquid precisely when exits are needed. Standard mitigations are procedural, such as documenting an independent valuation before consulting consensus, mandating a devil's advocate view, and evaluating managers over horizons long enough that contrarian positions can be sustained.