David Hirshleifer

David Hirshleifer is an economist whose work on information cascades helps explain how groups can make a poor decision even when each person is trying to learn from available evidence. In the account presented in the book, Hirshleifer, Sushil Bikhchandani, and Ivo Welch model how people’s observations of earlier choices can outweigh their own imperfect information—and how that process can ultimately prevent a group from aggregating what its members know.

The Information-Cascade Model

The model begins with a choice between a new Indian restaurant and a new Thai restaurant. The Indian restaurant is objectively better, but each person receives only an imperfect signal about which is preferable. A signal can be wrong, so people supplement it by watching what others do; in the economists’ setup, everyone understands that others also receive information that is useful, though not infallible.

The decisive complication is that choices occur in sequence. If the first diners happen to receive misleading signals and choose the Thai restaurant, those who arrive later see it filling up. A later diner may privately receive evidence favoring the Indian restaurant, yet reasonably infer from the crowd that the Thai place is more likely to be good. Once people begin making the same choice, their visible behavior can overwhelm their private evidence, and the group may converge on the inferior restaurant by chance.

Rational Following, Collective Error

The cascade is not simply conformity or a fashion adopted under peer pressure. In the model, people follow earlier choices because they believe those choices convey information. Given uncertainty about one’s own signal, and observing that everyone before one has chosen the same way, imitation can seem rational.

That reasoning creates a reversal. At first, choices appear to transmit information about what earlier diners have learned. But after enough people stop relying on their private knowledge, later decisions no longer add independent evidence. Each person is responding to what they think earlier people knew, rather than directly contributing what they know. The observed consensus therefore looks informative while the process producing it has ceased to gather the group’s information.

The book distinguishes this mechanism from simple trend-following: the people in the model are not choosing the Thai restaurant merely because it is popular. Its popularity is treated as evidence. The danger lies precisely in that plausible inference: a run of choices that began with a few signals, possibly wrong, can make a mistaken conclusion seem increasingly well supported.

The Model’s Broader Use

The book connects the model to the spread of plank roads. Early roads appeared to solve real problems: they cut travel time between towns in half, could be used in any weather, and helped small farmers reach wider markets. Each new road seemed to confirm the approach and made alternatives appear less promising. Yet the roads lasted closer to four years than the eight years George Geddes had promised, making them too costly to maintain; by the end of the Civil War, almost all had been abandoned. The example shows how genuine early benefits and the visible choices of others can reinforce a mistaken judgment before a system’s underlying weakness becomes clear.

The model also clarifies a distinction between aggregating information and following a sequence of visible decisions. A market or voting system can, in principle, combine individual information. An information cascade instead risks replacing that aggregation with imitation: people assume earlier choices reflect knowledge, even as each new choice contributes little or none of the chooser’s own evidence. Hirshleifer’s work thus supplies the book with a mechanism for understanding how collective error can emerge not from a lack of intelligence, but from the way people use one another’s actions as information.