We report on an experiment that uses revealed preference to distinguish between rational social learning and behavioral bias. Subjects are asked to correctly guess the current binary state of the world. They must choose between receiving a private, noisy signal about the current state or observing the past guesses of other subjects in the prior period. The design varies the persistence of the state across time, which determines whether choosing social or private information is optimal, enabling us to separate subjects who choose optimally from those who excessively use either social information (“herd animals”) or private information (“lone wolves”). Aggregate behavior appears unbiased only because the number of lone wolves and herd animals is approximately equal. Our findings cannot be explained by existing behavioral models, with the possible exception of rational inattention.

(Joint work with John Duffy and Tatiana Kornienko)