Construction of Investor Sentiment Factor from Social Media Big Data and Its Spillover Effects on Stock Market Volatility

Abstract

This study investigates the complex association between social-media-based investor sentiment and stock market volatility, with a specific focus on the mediating role of trading activity in contemporary financial markets. As digital platforms increasingly influence financial decision-making, understanding these dynamics is crucial. Using a comprehensive dataset of user posts extracted from the East Money Stock Forum for selected major CSI 300 constituent stocks, we systematically construct robust investor sentiment measures and a composite sentiment index. These metrics are subsequently matched with empirical market turnover and volatility indicators to capture real-time market reactions. The empirical results demonstrate that while the aggregate sentiment direction exhibits only a weak direct association with market volatility, sentiment intensity is significantly and positively related to volatility spikes. Furthermore, comprehensive mediation analysis suggests that the market turnover rate serves as an essential transmission channel linking digital sentiment to observed volatility. To provide a deeper understanding, quantile regression analysis further indicates that this relationship exhibits significant heterogeneity, varying considerably across different market volatility conditions, although the overall economic magnitude of this effect remains relatively limited in scope. Ultimately, these findings suggest that social media sentiment can provide valuable supplementary information for understanding and forecasting market fluctuations, especially when it is considered in conjunction with underlying trading activity and broader behavioral finance paradigms.