Herding Behavior and Stock Returns: An Exploratory Investigation

We collect trading and ownership statistics for U.S. stocks between 1970 and 1989 and we study the cross-section of returns. In rational and frictionless markets, equity returns should not depend on asset turnover nor should they depend on owner identity. Yet, with market imperfections, crowd behavior may affect returns. We examine two types of herding: (i) conventional investing, and (ii) trading for non-informational reasons. Incomplete information models predict that conventional stocks command higher prices. Noise trader models predict that shares that are traded for non-informational reasons are more risky and sell for lower prices. We find evidence that supports both predictions, even if we control for beta, firm size, and the book-to-market ratio.