Prior Sentiment and Returns Around Earnings Announcements

We develop a stylized model in which sentiment-driven demand creates a pre-announcement price wedge that is harder for arbitrageurs to offset when fundamental uncertainty is high. The model predicts that prior sentiment shifts expected post-announcement returns separately from the earnings surprise and that the shift is larger when firm-level and aggregate sentiment align. We test these predictions using LSEG MarketPsych sentiment, I/B/E/S earnings surprises, and CRSP returns for 83,293 U.S. quarterly earnings announcements from 1998-2022. The 10-day low-minus-high cumulative abnormal-return spread is 1.38 percentage points for firm-level sentiment and 1.50 percentage points for aggregate sentiment. Sentiment enters separately from the earnings surprise, while sentiment-by-surprise interactions are generally small and statistically insignificant. The return spread is larger when analyst forecasts are noisier and when firm-level and aggregate sentiment have the same sign. Firm-level sentiment predicts a smaller spread on randomly selected non-announcement dates, while aggregate sentiment has little explanatory power on those dates. The results isolate the predictive content of daily pre-announcement sentiment from the immediate response to realized earnings news and show how it varies with valuation uncertainty and sentiment source.