Investor Sentiment as a Mediator Between EPU and Herding: Evidence from Text-based Analysis

Abstract

Research Question: How does economic policy uncertainty (EPU) influence herding behavior through investor sentiment in the Chinese stock market? Motivation: Prior studies have focused on the direct effects of EPU on market behavior, yet the behavioral transmission mechanism via investor sentiment remains underexplored. This study addresses this puzzle by examining whether investor sentiment operates as a causal intermediary rather than a mere co-movement factor, linking EPU to herding behavior. Idea: The study investigates whether investor sentiment mediates the effect of EPU on herding behavior. The independent variable is EPU, the mediator is investor-expressed textual sentiment derived from online forums, and the dependent variable is herding behavior. The central hypothesis is that EPU influences investor sentiment, which in turn affects herding behavior. Data: The analysis uses monthly data from January 2011 to December 2023 for Chinese-listed stocks. Investor-generated posts from the online stock forum are classified as positive or negative to construct stock-level sentiment indices, capturing psychological rather than purely market-based dimensions of sentiment. Traditional sentiment proxies primarily reflect market outcomes rather than investors’ psychological states. Method/Tools: The study employs a three-step stepwise panel regression framework for mediation analysis to identify the indirect effect of EPU on herding through sentiment. Findings: Results reveal a significant partial mediation effect: while EPU impacts herding behavior, part of its influence is indirectly transmitted through investor sentiment. These findings highlight a previously underexplored behavioral channel and demonstrate the value of textual sentiment in reflecting investors’ psychological responses to policy uncertainty. Contributions: This study contributes to behavioral finance by uncovering a sentiment-based transmission mechanism between EPU and herding, and by showing that investor-expressed online sentiment provides a more accurate measure of psychological drivers of market behavior compared with traditional market-based proxies.