Dispersion of opinion and stock returns

We use a panel of more than 100,000 investor accounts in US stocks over the period 1991-1995 to construct an investor-based measure of dispersion of opinion, unlike the analyst-based measure used in the literature. We use this measure to test two competing hypotheses: the sidelined investors hypothesis and the uncertainty/asymmetric information hypothesis. We find evidence that supports the sidelined-investors hypothesis. We show that the dispersion of opinion of the investors in a stock is positively related to the contemporaneous returns and trading volume of the stock and negatively related to its future returns. Moreover, dispersion of opinion aggregates across many stocks, providing additional explanatory power in a standard asset pricing model. We also show that dispersion of opinion among retail investors Granger-causes dispersion of opinion among analysts.