Skip to main content

Herding

Updated Jun 24, 2026 at 2:35pm

Research Draft High 1,234 words

Herding is the tendency of market participants to converge on the same actions — buying, selling, or holding the same securities — by imitating the behaviour of others rather than acting on their own private information or analysis. Its core tension is that following the crowd can be individually rational (others may know something you don't, and being wrong alongside everyone else is safer for one's career) yet collectively destabilising, because it suppresses the independent, contrarian information that would otherwise keep prices anchored to fundamentals. Herding is thus both a private survival strategy and a public-good problem: the mechanisms that make it sensible for the individual are the same mechanisms that inflate bubbles and deepen crashes.

How it's formed

Behavioural economists distinguish several distinct channels, and the distinction matters because they have different policy and trading implications:

  • Informational cascades (Banerjee 1992; Bikhchandani, Hirshleifer & Welch 1992; Welch 1992). Agents act sequentially and observe predecessors' choices but not their private signals. After enough agents move the same way, a rational latecomer infers that the public information (the visible actions) outweighs his own private signal, so he ignores his own information and copies. Once started, the cascade is self-reinforcing but fragile — it rests on little actual information and can reverse abruptly when a credible new signal arrives.
  • Reputational / career-concern herding (Scharfstein & Stein 1990). Money managers and analysts mimic peers because of a "sharing-the-blame" effect: being wrong with the crowd is far less damaging to a reputation than being wrong alone, so professionals rationally suppress contrarian views.
  • Compensation-based herding. When pay or assessment is benchmarked against peers or an index, deviating from the consensus is penalised regardless of being right, creating an incentive to hug the herd.

A critical separation, emphasised by Bikhchandani & Sharma (2001), is intentional ("true") herding — deliberately imitating others while overriding one's own information — versus spurious herding, where investors merely react similarly because they share the same public information or face the same constraints. Spurious herding is an efficient outcome and is not a bias at all; only intentional herding represents the inefficiency people usually mean by the word.

How it's used in practice

For practitioners, "herding" functions less as a tradeable signal than as a lens and a risk warning:

  • Sentiment and crowding analysis. Desks monitor crowding — concentration of positioning, fund flows, short interest, retail participation, social-media sentiment — to gauge how much of a move is consensus-driven and therefore vulnerable to a sharp reversal once the marginal buyer is exhausted.
  • Contrarian framing. The classic application is contrarian: extreme one-sidedness (everyone bullish, dispersion of opinion collapsing) is treated as a condition for reversal. This underlies sentiment-extreme indicators and the folk wisdom of "be fearful when others are greedy."
  • Self-discipline. Behaviourally, naming herding helps a trader recognise FOMO-driven entries and benchmark-hugging in their own process — chasing a stock because it is going up and everyone owns it.
  • Systemic risk. Regulators and risk managers watch herding as a driver of fire-sale dynamics, liquidity spirals, and contagion across correlated funds.

Adoption, debate & evidence

Herding is one of the most studied concepts in behavioural finance, but the empirical record is genuinely mixed and the measurement is contested.

The dominant statistical methods are: the LSV measure (Lakonishok, Shleifer & Vishny 1992), which gauges whether managers buy/sell the same stocks more than chance would predict — and notably found little significant herding among the US pension-fund managers in its original sample; and the dispersion-based CSSD (Christie & Huang 1995) and CSAD (Chang, Cheng & Khorana 2000) measures, which look for return dispersion that is too low during large market moves. Christie & Huang found no herding evidence in the US. Chang, Cheng & Khorana found no herding in the US or Hong Kong, partial herding in Japan, and significant herding in the emerging markets of South Korea and Taiwan — a recurring pattern: herding is more detectable in emerging, less transparent markets, and during periods of stress, crisis, or extreme volatility than in developed markets in calm conditions.

The central honest caveat, stressed by Bikhchandani & Sharma (2001), is that these statistical measures cannot cleanly separate intentional herding from spurious herding — clustered trades may simply reflect common rational reactions to public news. There is a weak link between the elegant theory (cascades, reputation) and what the regressions actually detect. So the well-evidenced claim is narrow: trades and returns cluster more than independence predicts, especially in emerging markets and crises. The stronger claim — that this clustering is irrational imitation creating exploitable mispricing — is plausible and theoretically motivated but far harder to prove, and standalone "herding indicators" have no demonstrated forecasting edge. Reputational herding has cleaner support: studies of analyst career concerns (e.g. Hong, Kubik & Solomon 2000) find inexperienced analysts are punished more for bold, off-consensus forecasts, consistent with the Scharfstein–Stein mechanism.

Strengths & limitations

Herding's strength as a concept is explanatory: it gives a coherent, microfounded account of bubbles, crashes, momentum, fads, and contagion that pure efficient-markets theory struggles with, and it correctly predicts where such episodes concentrate (opaque markets, stressed regimes, agency-laden professional settings).

Its limitations are sharp. First, it is hard to measure and even harder to attribute — the spurious-vs-intentional ambiguity means much measured "herding" may be efficient. Second, it is a poor timing tool: crowds can stay crowded far longer than a contrarian can stay solvent, and "everyone is bullish" gives no edge about when the reversal comes. The #1 misuse is treating any observed correlation or consensus as proof of irrational herding and a reason to fade it — ignoring that the crowd is frequently right (it may be a justified response to real news), and that fighting a cascade early is how contrarians get run over. Herding is a hypothesis to investigate, not a signal to trade mechanically.

Sources

Disputes flagged: The literature does not agree that intentional herding is pervasive in developed markets; LSV, Christie–Huang, and Chang et al. find weak or no herding in the US. Statistical herding measures cannot separate irrational imitation from efficient common reactions — treat magnitude and "irrationality" claims as contested.