We develop a new emotion-based market-level sentiment indicator to measure the emotional state of the market. Using this aggregate series, we compute firm-level sensitivity to shifts in market-level emotions and find that stocks with high-emotion betas outperform low-emotion-beta firms. This performance differential is corrected in about six months. A trading strategy that takes a long (short) position in high- (low-) emotion beta stocks generates an annualized alpha of more than 6%. This evidence of emotion-based predictability is distinct from the known pricing effects of mood, traditional sentiment measures, economic and policy uncertainty, and tone.