Numerous empirical studies suggest that there exist trading strategies that can yield positive abnormal returns presumably because of asset pricing errors.For example, Jegadeesh and Titman(1993)report that investors can make substantial abnormal profits by buying past winners and selling past losers2.These studies have several common characteristics.First,they are based all on observed or realized prices.Naturally,the realized prices are the result of interactions among a large number of investors.Therefore, it is difficult to rely only on the empirical studies to identify the roles played by different investors in price determination.Second,the trading strategies such as the momentum trading documented in the empirical literature usually takes the price process as exogenous.This methodology is valid only if the investors who follow these strategies,in total,are price-takers.Investors cannot actively affect price processes for profit-making purpose. A distinctive feature of our model is its explicit investigation of how smart money (the manipulator)interacts with irrational traders and what profit the manipulator makes from exploiting other investors'behavioral biases.In other words,the manipulator in our model manipulates the price process to create more chances for the irrational investors to make mistakes.This is an important feature,but largely assumed away in the existing behavioral finance literature.For instance,Barberis,Shleifer,and Vishny (1998,BSV henceforth)have a representative agent model in which trading does not occur.Daniel, Hirshleifer,and Subrahmanyam(1998,DHS henceforth)consider two classes of traders, the informed (I)and the uninformed (U).However,since prices in their model are set by the risk-neutral informed traders,the formal role of the uninformed is minimal there. Hong and Stein also model two classes of traders--news-watchers and momentum traders. News-watchers only care about what news they observe,while momentum traders make 2 Lesmond,Schill,and Zhou(2003)argue that the profit of the momentum strategy documented by Jegadeesh and Titman is illusory because of transactions costs.Lesmond,Schill and Zhou's result therefore provides positive evidence for the argument of limits of arbitrage." 33 Numerous empirical studies suggest that there exist trading strategies that can yield positive abnormal returns presumably because of asset pricing errors. For example , Jegadeesh and Titman (1993) report that investors can make substantial abnormal profits by buying past winners and selling past losers2 . These studies have several common characteristics. First, they are based all on observed or realized prices. Naturally, the realized prices are the result of interactions among a large number of investors. Therefore, it is difficult to rely only on the empirical studies to identify the roles played by different investors in price determination. Second, the trading strategies such as the momentum trading documented in the empirical literature usually takes the price process as exogenous. This methodology is valid only if the investors who follow these strategies, in total, are price-takers. Investors cannot actively affect price processes for profit-making purpose. A distinctive feature of our model is its explicit investigation of how smart money (the manipulator) interacts with irrational traders and what profit the manipulator makes from exploiting other investors’ behavioral biases. In other words, the manipulator in our model manipulates the price process to create more chances for the irrational investors to make mistakes. This is an important feature, but largely assumed away in the existing behavioral finance literature. For instance, Barberis, Shleifer, and Vishny (1998, BSV henceforth) have a representative agent model in which trading does not occur. Daniel, Hirshleifer, and Subrahmanyam (1998, DHS henceforth) consider two classes of traders, the informed (I) and the uninformed (U). However, since prices in their model are set by the risk-neutral informed traders, the formal role of the uninformed is minimal there. Hong and Stein also model two classes of traders--news-watchers and momentum traders. News-watchers only care about what news they observe, while momentum traders make 2 Lesmond, Schill, and Zhou (2003) argue that the profit of the momentum strategy documented by Jegadeesh and Titman is illusory because of transactions costs. Lesmond, Schill and Zhou’s result therefore provides positive evidence for the argument of “limits of arbitrage