Sequential social-learning model
Information Cascade
In a sequence of public decisions, early actions can reveal enough apparent consensus that later people rationally ignore their own private signals and follow the crowd.
public action history can outweigh one private signal
A cascade is informational, not simply imitation. Individuals may be rational given what they observe even while the group outcome is fragile or wrong.
Agents receive hidden private signals, observe earlier actions, and then choose. Once the public likelihood crosses a threshold, conflicting signals stop changing behavior.
(%)
The animation runs automatically, pauses on the conclusion, and then repeats. The main control changes the scenario rather than scrubbing the timeline.
- CHANGE
- Public lead size
- WATCH
- private-signal override
- MEANING
- Agents receive hidden private signals, observe earlier actions, and then choose. Once the public likelihood crosses a threshold, conflicting signals stop changing behavior.
Actions reveal less information after a cascade begins.
A sequential decision line displays each private signal behind the chosen public action and marks the point where later signals are suppressed.
What it actually says
Bikhchandani, Hirshleifer, and Welch modeled how sequential decision makers infer information from predecessors' actions. Once a sufficiently strong action imbalance forms, later choices can add no new public information.
Because the underlying private evidence is thin, cascades can be brittle: a small disclosure, trusted expert, or change in the observed sequence may reverse the crowd.
"A useful law compresses a pattern. It does not erase the conditions that make the pattern true."
How the idea developed
The modern form emerged through observation, argument, and later refinement. The timeline separates the first insight from the version now used in textbooks and practice.[1]
Bikhchandani, Hirshleifer, and Welch formalize informational cascades.
Laboratory experiments test cascade formation and fragility.
Models extend to networks, recommender systems, markets, and online platforms.
How the pattern works
The relation becomes useful only when its mechanism, measurement process, and operating range are visible.
Later agents see earlier actions but not all signals.
Public evidence can dominate one private draw.
Following actions cease to reveal the signal underneath.
A cascade is informational, not simply imitation. Individuals may be rational given what they observe even while the group outcome is fragile or wrong.
Where it earns its keep
Applications are strongest when the law changes a decision, measurement, model, or experiment rather than merely providing an analogy.
Separate choice counts from evidence
ApplicationMany followers may reflect the same early information.
Expose independent signals.
Delay popularity cues
ApplicationEarly counts can lock in arbitrary paths.
Test randomized or private evaluation.
Where it stops working
Simultaneous choice, rich communication, heterogeneous preferences, networks, strategic signaling, and visible outcomes change the classic model.
"Crowds are irrational"
Better: Following can be individually rational under limited information."Popularity proves shared evidence"
Better: A long sequence may rest on only the first few signals.Sources and further reading
Original publications and serious secondary scholarship are prioritized over summaries.
- Bikhchandani, Hirshleifer, and Welch - A Theory of Fads, Fashion, Custom, and Cultural ChangeFoundational 1992 model.https://doi.org/10.1086/261849
- Anderson and Holt - Information Cascades in the LaboratoryExperimental evidence.https://doi.org/10.1257/aer.87.5.847
- Bikhchandani, Hirshleifer, and Welch - Learning from the Behavior of OthersReview of informational cascades.https://doi.org/10.1257/jep.12.3.151