Informational cascades in the laboratory: Do they occur for the right reasons?

Recently, the theory of informational cascades has been tested in an experiment by Anderson and Holt (1997) who report that their data support the theory amazingly well. In this note we report on an experiment designed to find out whether observed cascades are indeed due to rational Bayesian updating. However, we find little support for rational updating. The simple heuristic "follow your own signal" does much better in explaining our data than Bayesian rationality.