Herd Behavior: When Other People's Actions Become the Signal

Herd behavior occurs when people treat other people's actions as information and those actions become the signal for the next decision. The loop can amplify a sound discovery, a mistake, or a forced sale.

Herding is a feedback mechanism, not a personality type

An investor may follow others for a reasonable reason. Their trade can reveal information the investor does not possess, or it may reduce the cost of acting under uncertainty. Herding begins when the observed action changes later actions, and those responses become new evidence for still later participants. The group can then move farther than any participant would have moved using private information alone.

The information-cascade literature formalizes this problem: early actions are visible while private signals are hidden, so later people may rationally copy the apparent consensus even when their own information points elsewhere. Bikhchandani, Hirshleifer, and Welch's model (Journal of Economic Perspectives article) is a model of a mechanism, not proof that every market movement is a cascade.

What turns a movement into a loop?

Four conditions make amplification more likely:

• Visibility: prices, flows, rankings, or withdrawals are easy to observe.
• Uncertainty: participants have incomplete information and need a signal.
• Response: the signal changes trading, funding, production, or allocation.
• Constraint: leverage, liquidity, deadlines, or rules force similar actions at the same time.

A loop can be positive or negative. Rising prices attract trend followers, which adds buying. Falling collateral values trigger margin calls, which add selling. A bank's withdrawals reduce its liquid resources, which can make further withdrawals more rational. A loop does not determine whether the first action was correct; it determines how much the response feeds back into the next action.

A documented withdrawal cascade

The March 2023 failure of Silicon Valley Bank shows herding through funding rather than stock prices. The Federal Reserve's review describes concentrated uninsured deposits, rapid electronic withdrawals, and risk-management failures. Each depositor's decision was shaped by the possibility that others would withdraw first. Once withdrawals accelerated, the bank's remaining liquidity and the public signal of distress made further withdrawals more defensible. The case does not show that all depositors acted irrationally; it shows how individually understandable actions can create a collective failure before a slower balance-sheet adjustment is possible.

Other forms of market imitation

Momentum strategies use recent price movement as an input. If enough capital follows the same signal, the strategy can reinforce the movement it measures. That does not establish that momentum returns are “just herding”; risk premia, information diffusion, and changing fundamentals are alternative explanations.

Flows and crowded positions can create liquidity that is available while many participants want the same exposure and scarce when they all seek the exit. A quoted price remains an observation of the last transaction, not a guarantee that the entire position can be sold at that price.

Corporate and policy imitation can also create loops. Companies copy a popular technology, accounting metric, or acquisition strategy because peers have adopted it. The visible adoption becomes evidence of legitimacy, while the shared investment changes supplier prices, labour availability, and competitive behaviour. Similarity may reflect common information rather than imitation, so the causal claim needs evidence about sequence and dependence.

When the loop stops

Feedback weakens when private information becomes visible, funding constraints are relaxed, prices become unattractive, or a participant with a different mandate enters. A value investor may buy a falling asset; a central bank may provide liquidity; a bank may raise capital; a company may halt a project. These interventions change the mapping from signal to action.

Loops can also reverse without a new fact. A crowded position has a narrow exit when many holders must sell at once. Leverage and redemption rules can turn a discretionary decision into a forced one. The reversal often appears abrupt because the same visibility and imitation that built the position now operate in the opposite direction.

Questions for an investor

• Identify the private information. Are participants acting on independently observed fundamentals, or mainly on other people's positions and prices?
• Map the forced actors. Which holders face margin calls, redemptions, collateral rules, index changes, or deadlines?
• Measure liquidity under stress. How much can be sold or refinanced without moving the price, and who supplies the other side?
• Look for independent checks. Do cash flows, customer use, solvency, and physical capacity support the narrative, or is the narrative itself the main evidence?
• Separate correlation from causation. Similar trades may reflect shared information rather than a feedback loop. Seek the sequence that made one action an input to the next.

Herd behavior is therefore a property of an information and funding system. It can spread accurate information, amplify an error, or accelerate a necessary adjustment. The investor does not need to label a crowd rational or irrational in advance. The useful work is to identify which signals are endogenous, which participants are constrained, and what would break the loop.