Modern financial markets contain many investors. In this context, we study the role of information in investor decision-making, and the informational efficiency and liquidity of the market. An equilibrium is characterized in closed-form for a continuous-time economy with many market participants and imperfect competition, in which investors receive private information with varying quality, and are heterogeneous in their misperception of the information quality. In equilibrium, investor heterogeneity in their misperception generates return predictability by investors’ trading, and trading of different investors follows a simple factor structure with weak factors. To conduct empirical analysis that builds on these equilibrium implications, we develop a new big-data econometric method that utilizes the factor structure to accommodate the high-dimensionality of these implications. Applying the framework to price and institution holding data of the US stock market, we document that individual institution’s trading with impotent predictive power can collectively generate significant return predictability that persists for about a quarter. We estimate dynamic price impact of around 0.25 at quarterly frequency, a moderate misperception of institutions on their information quality, and institutions’ contributions to the informational efficiency of the market.