They construct an index of sentiment based on factors like share turnover on the NYSE, the dividend premium, the volume and first-day returns from IPOs, discounts on closed-end funds, and the equity share of news issues. Here's the conclusions section from their paper (normally I post the abstract, but this section gives a better feel for their results):
The paper is downloadable from SSRN here.Investor sentiment affects the cross-section of stock returns. For practitioners, the main takeaway is that the cross-section of future stock returns varies with beginning-of-period investor sentiment. The patterns are intuitive and consistent with economic theory. When sentiment is high, stocks that are prone to speculation and difficult to arbitrage, namely stocks of young, small, unprofitable, non-dividend-paying, highly volatile, distressed, and extreme growth firms, tend to earn relatively lower subsequent returns. When sentiment is low, the reverse mostly holds. Most strikingly, several characteristics that exhibit no unconditional predictive power actually exhibit predictive power once we condition on beginning-of-period investor sentiment. These results suggest there is much to be done in terms of understanding more about investor sentiment and its effects.
HT: CXO Advisory Group