Information Cascade, explained.
An information cascade can occur when people following earlier actions stop acting on their own private evidence, causing later choices to convey less new information.
Why it happens
Earlier actions may be useful signals, but many copied actions need not represent many independent observations. Once behavior becomes sufficiently persuasive, private clues can be suppressed and even a mistaken cascade can persist.
Bikhchandani, Hirshleifer and Welch showed how observing earlier actions can lead later decision makers to set aside their own private signals.
Read the result
Compare public choices with private signals. A long run of identical actions is more informative if it reflects independent evidence than if later players are copying the same earlier decisions.
A worked example
A queue forms outside a restaurant
Two early visitors enter based on their own clues about quality.
Later visitors see the queue and join despite weak private impressions, assuming the earlier visitors knew more.
The crowd grows, but the additional bodies may contain little additional independent evidence about quality.
OPTIONAL DEEPER DETAILGo deeper: inside the model
Inside this model
A hidden state is equally likely red or blue. Each independent clue is correct with the chosen probability. At each turn an agent compares the public likelihood from earlier actions with their clue likelihood, chooses the more likely state, and follows their clue on a tie. The public belief is updated from the probability of that action under each state; once both possible clues produce the same choice, the action conveys no new clue.
Where this idea is useful
A practical use
Early reviews can influence later buyers so strongly that later choices add little fresh evidence about product quality.
A common misconception
“Ten people agreeing means ten independent confirmations.”
Their decisions can be dependent if each inferred quality from the same earlier actions.
What this explanation leaves out
- Agents have common knowledge of clue accuracy, act to maximize correctness, and never communicate the clues themselves. Real social learning is more varied.
Can a rational individual join a mistaken cascade?
Yes. Given limited private information and persuasive public actions, following can be sensible locally even when the collective conclusion is wrong. Revealing new independent information may change it.
How many independent observations lie behind the apparent consensus?
Associated thinkers
Further reading
Explore the original research or the teaching reference behind this experiment.