Assessing the Explanatory Power of Investor Sentiment on Volatility

Abstract

This paper investigates the extent to which firm‐level investor sentiment, as measured by EWMA‐smoothed FinBERT scores, can explain subsequent movements in both realized and implied volatility. We construct asymmetric sentiment variables by median‐centering and splitting into positive/negative deviations, then estimate two‐way‐fixed‐effects panel regressions on a large cross‐section of stocks. Our results show that both positive and negative sentiment shocks have statistically significant, economically meaningful impacts on volatility for realized volatility. For implied volatility, our results indicate a muted effect on negative news, while positive news has no impact. These findings highlight the value of incorporating high‐frequency sentiment measures into volatility forecasting models and underscore the asymmetry in market responses to good versus bad news.