Sentiment co-movement and stock returns

Investor sentiment is widely associated with stock mispricing, yet firm-level sentiment measures often have limited predictive power for individual stock returns when considered in isolation. We argue that what matters more is not simply the level of sentiment itself, but whether extreme sentiment is likely to co-move across stocks. To capture this, we propose Comovement Sentiment Measures (CoSENT), a set of five daily firm-level metrics that quantify the probability that a stock enters an extreme sentiment state conditional on extreme sentiment among other stocks in the cross-section. The framework distinguishes between concordant and divergent comovement, and between downside and upside sentiment dynamics, thereby allowing for a structured analysis of asymmetric spillovers. Using Thomson Reuters MarketPsych sentiment data and CRSP returns for U.S. common stocks from 1998 to 2019, we find pronounced asymmetry in sentiment spillovers: downside comovement is stronger and more pervasive than upside comovement. In cross-sectional return forecasting over horizons of one to five trading days, downside CoSENT measures, especially divergent pessimism, robustly predict returns and outperform conventional sentiment level measures as well as upside comovement indicators. Portfolio sorts and Fama–MacBeth regressions further show that strategies based on downside spillover likelihood generate economically meaningful risk-adjusted returns. Overall, the results suggest that sentiment contagion risk, rather than sentiment level alone, is an important state variable for short-horizon asset pricing and for understanding behavioural amplification during periods of market stress.