Redundant Information and Predictable Returns

Abstract

How well do investors distinguish information that already is priced from genuinely novel and exclusive private information? This paper examines whether investors misweight information that already is in stock prices (“redundant information”) in making their trading decisions. I extend the Kyle (1985) model to allow for non-Bayesian updating and transaction costs. The model predicts that price changes exhibit a state space process, in which the parameter for investors' non-Bayesian weighting of redundant information is estimable distinctly from information asymmetry, transaction costs, and serial correlation in liquidity trader demand. Using this model, I estimate a firm-quarter measure of investors' misweighting of redundant information. I find that, on average, investors behave as if over 47 percent of the information content in the immediately prior price change is private information. Overall, these results suggest one way that momentum and mean reversion in stock price returns could result from investors' misuse of information.