There has long been interest in the effect of chan ges in ‘animal spirits’. The paper uses a differences-in-differences approach to estimate the macroeconomic effect of sentiment, comparing periods of high and low sentiment, making use of cross-sectional differences. Sentiment is measured using the Baker-Wurgler (2006) sentiment index. We estimate the effect of sentiment on intertemporal tradeoffs. We obtain statistically precise estimates of the effect of sentiment. Low sentiment leads agents to discount future payoffs more heavily. The estimated effects are large (600 basis points). High sentime nt induces overinvestment for some firms; low sentiment induces underinvestment for other firms.