Numerous economic models suggest that divergence of opinion can affect the stock market equilibrium. However, these models produce conflicting cross-sectional implications of the effect of divergence of opinion on expected asset returns. Moreover, prior empirical research fails to identify a reliable proxy for divergence of opinion; proxies suggested by the literature give mixed results. In this dissertation, we propose a new proxy and empirically investigate the relationship between divergence of opinion and expected returns.

We first derive a new proxy for divergence of opinion among individual investors by generalizing the Mixture of Distribution Hypothesis model of Tauchen and Pitts (1983), which links volatility and volume. This new measure of divergence of opinion is not only strongly and positively related to other proxies previously used in the literature (e.g., dispersion in analysts' earnings forecasts and turnover) in cross-sectional regressions, but it also provides a closer link to the theoretical work of Miller (1977) and Hong and Stein (2003). Furthermore, this proxy has several advantages over existing ones. (1) It can proxy for divergence of opinion among all individual investors instead of merely financial analysts who may issue biased reports due to conflicts of interest and whose recommendations may not be followed by many investors. These phenomena make dispersion in analysts' earnings forecasts a noisy and imprecise measure of divergence of opinion among individual investors. (2) It isolates divergence of opinion from the joint distribution of volume and volatility, thus it is a cleaner measure of individual investor's divergence of opinion than asset turnover, which can also proxy for liquidity and information-induced price adjustment speed.

In addition to deriving a new proxy for divergence of opinion, we investigate how divergence of opinion affects cross-sectional stock returns when stocks are sorted according to their characteristics such as size, book-to-market, and momentum, which reflect the degree of difficulty associated with completing a short sale transaction. The empirical results show that stocks with higher divergence of opinion tend to be more overpriced and that the lowest-divergence portfolio outperforms the highest-divergence portfolio by 8 percent annually. This effect of divergence of opinion on equity returns is concentrated on small, low-book-to-market, and high-momentum stocks, which are more difficult and costly to short sell. In particular, we find that the low divergence of opinion stocks outperform the high divergence ones by 17 percent, 15 percent and 6 percent per year in the three small size groups, and by 17 percent, 14 percent and 7 percent after risk adjustment. On the other hand, the return difference between low- and high- divergence of opinion stocks in the two large size groups either has the wrong sign or is statistically insignificant. Among the small stocks, low-divergence growth (high-momentum) stocks outperform their high-divergence counterparts by almost 23 percent (26 percent) per year after the risk adjustment. Even for mid-size stocks, this effect is still strong with low-divergence growth (high-momentum) stocks outperforming their high-divergence counterparts by 7 percent (12 percent). The inclusion of other proxies such as turnover and dispersion in analysts' earnings forecasts in cross-sectional regressions does not drive out the effects of divergence of opinion. Further, subperiod analysis shows that the divergence of opinion has had smaller influence on subsequent returns in recent years, presumably because short-sale constraints have become less binding. The evidence supports Miller's (1977) view that given short-sale constraints, high divergence of opinion should lead to lower future returns because short-sale constraints prevent the pessimists' view from being reflected in the stock price, which contrasts with Hong and Stein's (2003) unbiased price prediction. Specifically, we find that the prediction of Hong and Stein (2003) is valid only for stocks which are relatively easy to short sell.

This dissertation also suggests several directions for future research. (1) We will study how short-sale constraints affect the relationship between returns and divergence of opinion by using a direct measure of short-sale constraints from the equity lending market. In this way, we can estimate quantitatively how much the short-sale constraints can explain the superior performance of low-divergence-of-opinion stocks. (2) The time series characteristics of the monthly average of divergence of opinion across all the stocks, defined as the market-wide aggregate divergence of opinion, shows that high divergence of opinion coincide with the 1987 stock crash and with most business cycle downturns. It is interesting to explore the relationship between market-wide aggregate divergence of opinion, market-wide liquidity, and stock returns in future work. (3) Another question for future work is to test Hong and Stein's (2003) hypothesis that high divergence of opinion should forecast more negative skewness. While leverage-effects, volatility-feedback, and bubble theories all fail to explain the underlying mechanism of negatively-skewed market returns, the new proxy for divergence of opinion can potentially lead to an answer to this puzzle.