Education or Manipulation? Influencers and Retail Trading in the Chinese Market.

China’s stock market is unique, driven by its dominance of retail investors. The behavioral biases of small investors have been a critical factor contributing to the underperformance of the Chinese stock market. Retail investors usually follow influencers on social media to guide their investment decisions. Theoretically, Pedersen (2022) pioneered that influencers affect the naive investors’ belief and price movements. Anecdotal evidence suggests that influencers utilize social media to manipulate stock prices. Despite their pivotal roles, little research has focused on the credibility of influencers' information and their potential involvement in market manipulation.

Rational posts by influencers could reduce information asymmetry and provide investment value to retail investors, while irrational or misleading content might harm market efficiency. This project seeks to explore the topical structure of influencers’ discussions and assess the credibility of their information. Leveraging the unique structure of the Chinese stock market and a proprietary account-level retail trading dataset covering more than 20 million accounts and 4,000 stocks, we aim to evaluate how influencers impact retail investors’ beliefs, trading behavior, and overall market efficiency, providing evidence of their potential in market manipulation.

We have four primary objectives. First, we characterize the topical structure of influencers’ discussions by employing the cutting-edge topic modelling techniques (BERTopic). Second, we assess the credibility in the influencers’ information by linking their topic allocation to firm-level news and macroeconomic dynamics. Third, using a comprehensive account-level retail trading dataset, we examine how different topics influence retail trading and whether these trading activities correspond with stock price movements, aiming to uncover evidence of influencers’ potential in market manipulation. Lastly, scrutinizing the topic allocation between influencers and naive investors, we unveil the belief spillover process in social networks. Moreover, we also plan to explore how different topics affect price informativeness and market efficiency in the Chinese market.

Our study has several contributions. Academically, we advance the application of AI and machine learning with large language models in finance by exploring the multi-dimensional topic structure of influencer discussions. The identified topics open potential avenues for further research in various markets. Practically, we assess both the credibility and potential manipulative roles of influencers, while also highlighting their possible positive contributions to market efficiency. Finally, our findings provide valuable insights for policymakers and regulators, offering recommendations to better monitor influencer activity and ensure the stable operation of China’s stock market.