Intermediatebehavioral
Herd Behavior: Why We Follow the Crowd
Overview
Why do restaurants with lines attract more lines, investors pile into bubbles, and bank runs become self-fulfilling? Herd behavior: individuals rationally or psychologically following others rather than private information. Economics models it precisely — information cascades (Bikhchandani-Hirshleifer-Welch) show how rational individuals produce irrational crowds. This lesson covers the cascade model, reputational herding, the finance applications (bubbles, runs, momentum), and how to think independently in herding environments.
Learning objectives
- Define herd behavior and information cascades
- Walk through the cascade model step by step
- Distinguish informational vs. reputational vs. payoff herding
- Apply to bubbles, bank runs, and momentum
- Evaluate defenses: independent thinking under social pressure
The cascade model
Bikhchandani, Hirshleifer, and Welch (1992): people decide sequentially, each seeing predecessors' actions but not their private signals. Person 1 gets a mildly positive signal, invests. Person 2's own signal is mildly negative — but seeing Person 1 invest, the inference ('they must know something') outweighs it; she invests. Person 3 now sees two investments — her negative signal is swamped; she invests regardless of her information. A cascade: from Person 3 on, private information never enters the public pool. The crowd's behavior reflects the first movers' signals only — yet looks like unanimous conviction. Rational individuals, collectively discarding information.
Cascades are fragile: because they rest on little information, a small shock (one credible defection) can shatter them — explaining fads' sudden collapses and bank runs' reversals.
Three kinds of herding
Informational herding (the cascade): following because others' actions signal information — rational individually, fragile collectively. Reputational herding (Scharfstein-Stein): fund managers mimic others because being wrong alone gets you fired while being wrong together is excusable — 'nobody ever got fired for buying IBM.' Payoff herding: following because others' choices directly affect payoffs — network effects (everyone's on the platform), bank runs (withdraw because others withdraw, making failure self-fulfilling). Each has different welfare implications and remedies: better information for cascades, incentive reform for reputational, coordination devices (deposit insurance) for payoff herding.
Diamond-Dybvig: if others withdraw, the bank fails, so you withdraw too — the run is a self-fulfilling equilibrium. Deposit insurance breaks it by removing the payoff to running.
Herding in financial markets
Bubbles: momentum traders buy because prices rise (others buying), detaching prices from fundamentals — the cascade in asset markets, amplified by leverage and narratives (Shiller's 'irrational exuberance'). Analyst herding: forecasts cluster — reputational safety in the consensus. Fund flows: money chases past performance (return-chasing investors herd into hot funds, then underperform). Bank runs and crises: payoff herding in wholesale funding (2008's repo runs). Crypto manias compress all three: informational cascades on social media, reputational FOMO, and payoff reflexivity. Markets are herding's natural habitat — prices aggregate others' actions visibly and continuously.
Trend-following profits from herding's persistence — but cascade fragility means momentum crashes violently. Riding herds pays until the cascade breaks; timing the break is the hard part.
Beyond markets: the social cascade
Herding shapes more than finance. Restaurants: the full one must be good (informational cue) — sometimes it's just the cascade's first movers. Technology adoption: network-effect herding locks in standards (QWERTY, VHS). Politics: preference falsification (Kuran) — people hide true views, so public consensus cascades on false unanimity until a trigger reveals it (revolutions surprise). Science: researchers herd into fashionable paradigms (reputational). Social media supercharges all of it: visible metrics (likes, shares) are cascade fuel, and algorithmic amplification picks the first movers. Understanding herding is now civic literacy.
Kuran: when expressing dissent is costly, public opinion cascades on apparent consensus — which can collapse overnight when the falsification becomes common knowledge. The mechanism of sudden revolutions.
Thinking independently
Defenses for individuals: seek base rates and outside views (the crowd's confidence isn't evidence); ask what the first movers actually knew (cascade fragility check); pre-commit to decision criteria before seeing others' choices; diversify information sources (break the sequential-observation structure); and distinguish 'others chose this' (weak signal) from 'others profited from this' (stronger). For institutions: secret ballots and simultaneous (not sequential) decisions prevent cascades; devil's advocates and red teams institutionalize dissent; prediction markets aggregate information rather than actions. The goal isn't contrarianism — it's making your private signal count instead of discarding it into the cascade.
Key points
- Herd behavior: following others' actions over private information — modeled precisely by information cascades.
- Cascade logic: sequential observation lets early movers' signals swamp later private information — rational individually, fragile collectively.
- Three types: informational (cascades), reputational (career safety), payoff (bank runs, network effects) — different remedies each.
- Finance applications: bubbles, analyst herding, fund-flow chasing, repo runs; social media amplifies all of it.
- Defenses: base rates, pre-commitment, simultaneous decisions, devil's advocates, prediction markets — make private signals count.
Common misconceptions
✓ Truth: Cascade herding can be individually rational — inferring information from others' actions is sensible. The pathology is collective: information gets discarded, not that individuals are stupid.
✓ Truth: Cascades can persist long past 'obvious' mispricing (Keynes: markets stay irrational longer than you stay solvent). Independence means using your signal, not reflexively opposing crowds.
✓ Truth: Cascades are about the structure of observation, not information quantity. Social media gives more information yet more herding — visible actions are the cascade's fuel.
Practice questions
Q1. In the Bikhchandani-Hirshleifer-Welch cascade model, a cascade begins when:
- Everyone gets the same signal
- An individual's private signal is outweighed by the inference from predecessors' actions, so they follow regardless of their own information
- The government intervenes
- Prices are fixed
Sequential observation with hidden signals: once the weight of observed actions exceeds your private signal, you herd — and your action then adds to the weight crushing the next person's signal.
Q2. Reputational herding (Scharfstein-Stein) occurs because:
- Managers love crowds
- Being wrong together is career-safe while being wrong alone gets you fired — so managers mimic each other
- Regulations require it
- Clients demand it
Career incentives punish solo errors more than consensus errors. The result: correlated 'mistakes' that look like shared conviction but are really shared self-protection.
Q3. Bank runs are best classified as:
- Informational cascades
- Payoff herding — withdrawing is rational if others withdraw, making the run self-fulfilling
- Reputational herding
- Random panics with no logic
Diamond-Dybvig: the payoff to withdrawing depends on others' withdrawals. It's a coordination failure with multiple equilibria — solved by deposit insurance, not by 'calming people down.'
Click a question to reveal the answer and explanation.
Frequently asked questions
What triggers an information cascade?
Sequential decisions with visible actions but hidden private signals. Once predecessors' actions outweigh your signal, you follow regardless of it — and everyone after you does too.
Why are cascades fragile?
They aggregate little private information — just the first movers' signals. One credible public defection can flip the inference chain, collapsing the unanimity suddenly.
How does deposit insurance stop bank runs?
Runs are payoff herding: withdrawing is rational if others withdraw. Insurance guarantees deposits regardless, removing the payoff to running — the equilibrium flips to calm.
Is momentum investing just herd exploitation?
Largely yes — momentum profits from herding's persistence (underreaction then overreaction). The risk is cascade fragility: momentum crashes when the herd turns.