Beginning with Blackwell and Dubins (1962) and Aumann (1976) economists have focused on circumstances that make disagreement difficult in a rational setting. Aumann (1976) shows if agents share a common prior and have common knowledge of each other’s posterior beliefs, they cannot agree to disagree. More recent work, however, shows that disagreement can arise even when agents have common priors and observe the same time series of public information. Acemoglu et al. (2006) argue, there are many empirical scenarios where permanent disagreement arises. In fact Acemoglu et al. (2006) demonstrate that agreement may be impossible, despite the fact that all agents update their beliefs as Bayesians. Several empirical papers in finance have studied the effect of disagreement among investors on asset prices. There are three primary hypotheses associated with this stream of research. The first, to which we refer hereafter as the ‘‘information asymmetry hypothesis,’’ posits that high levels of investor disagreement regarding a stock imply a high level of information asymmetry. This in turn suggests an initial discounting of a stock’s price, i.e., low initial returns, followed by high subsequent returns to compensate for its high level of risk (see Williams 1977; Varian 1985; Merton 1987; Kraus and Smith 1989; Wang 1993; Harris and Raviv 1993; He and Wang 1995; Naik 1997). The second hypothesis, which was originally developed by Miller (1977), posits that high levels of disagreement lead to ‘‘optimistic prices’’ when there exist restrictions against short-selling.1 Because only the pessimistic investors are restricted from participating in the market for that stock, the stock’s price will therefore reflect the opinion of the more optimistic investors. We designate this the ‘‘sidelined investor hypothesis.’’ Morris (1996), Chen et al. (2002), and Viswanathan (2002) offer alternative explanations for optimistic investors’ participation in stock markets. Naturally, the greater the level of shorting constraints or the greater the level of disagreement, the more upwardly biased the price and hence the initial return will be. When this overvaluation reverses, we will observe negative abnormal returns for the stock. Thus, the first two hypotheses come to opposite conclusions; the first suggests low initial and high subsequent returns, while the second suggests high initial and low subsequent returns. It is quite plausible that the aforementioned competing hypotheses may be at play simultaneously in the market, neutralizing the effect of each other or through another mechanism that results in unbiased prices. In fact, there is a third hypothesis, which has not received as much attention in the empirical asset pricing literature, which leads to similar conclusions albeit employing different arguments. This hypothesis is attributed to Diamond and Verrecchia (1987) and Hong and Stein (2003). The central prediction of their models is that prices remain unbiased despite the presence of disagreement among investors. Consequently, future abnormal returns will, on average, be close to zero. These models rely on the existence of influential rational agents. While Diamond and Verrecchia (1987) require a perfectly rational market maker with unlimited computational abilities and access to all public information, Hong and Stein (2003) depend on the presence of perfectly rational arbitrageurs to eliminate any mispricing. We refer to this hypothesis hereafter as the ‘‘unbiased prices hypothesis.