We study two important questions regarding trading dynamics in China. How do retail and institutional investors trade, and what are the underlying factors for these behaviors? Different from the United States, China’s stock market has two prominent features. Dominance of retail investors and active participation by the government. After reviewing nearly 100 previous studies, we reach three conclusions. First, there are substantial heterogeneity in retail investors. Small retail investors have low financial literacy, exhibit behavioral biases, and not surprisingly, negatively predict future returns, whereas large retail investors and institutions are capable of processing information and positively predict future returns. Second, the macro- and firm-level information environment in China is slowly but gradually improving, which greatly affects trading behaviors of different investors, especially the more sophisticated institutional investors and large retail investors. Finally, the Chinese government actively adjusts their regulations on the stock market to serve the dual goals of growth and stability. Many regulations are effective, while some may generate unintended consequences.

Using the Boehmer, Jones, Zhang, and Zhang (2021) algorithm, we identify a broad swath of marketable retail investor orders in the U.S. market during the pandemic. The marketable retail trading volumes rapidly rise from $325 billion in 2019 to $852 billion at mid-2020, and stay high for the next two years. The retail order flows positively predict cross-sectional returns over various horizons, and are associated with wider future effective spreads and higher future volatilities, as well as less market participations by high frequency traders and short-sellers. We find supportive evidence for informed and uninformed retail hypotheses. (Presented at 2022 Plato Market Innovator (MI3) Conference, 2022 Transparency and Market Structure Conference, Tsinghua Finance Seminar Series.)

We examine whether large language models (LLMs) can extract contextualized representation of Chinese news articles and predict stock returns. The LLMs we examine include BERT, RoBERTa, FinBERT, Baichuan, ChatGLM and their ensemble model. We find that tones and return forecasts extracted by LLMs from news significantly predict future returns. The equal- and value-weighted long minus short portfolios yield annualized returns of 90% and 69% on average for the ensemble model. Given that these news articles are public information, the predictive power lasts about two days. More interestingly, the signals extracted by LLMs contain information about firm fundamentals, and can predict the aggressiveness of future trades. The predictive power is noticeably stronger for firms with less efficient information environment, such as firms with lower market cap, shorting volume, institutional and state ownership. These results suggest that LLMs are helpful in capturing under-processed information in public news, for firms with less efficient information environment, and thus contribute to overall market efficiency. (Presented at at ABFER-JFDS Annual Conference on AI and FinTech 2024, China Fintech Research Conference (CFTRC) 2024, Summer Institute in Finance (SIF) Annual Conference 2024, Seminar Series at Sun Yat-Sen University, Tsinghua University and Summer Institute in Digital Finance (SIDF) 2024.)

Using a proprietary granular database of a major Chinese stock exchange, we examine heterogenous investors’ trading dynamics around one of the most important macro announcements of the Chinese central bank, the monthly release of monetary aggregates data. Exploiting the trading heterogeneity across assets and across investor types, we find that before announcements, institutional investors reduce their aggregate stock exposure while over-weighing riskier stocks of smaller caps, whereas retail investors provide liquidity by increasing their aggregate stock exposure and avoiding the riskier stocks. Large retail and institutional investors become more informed before announcements and trade in correct directions consistent with the news surprises after announcements, while smaller retail investors trade in opposite directions. While the institutional investors accumulate positive returns with risk compensated, the market realizes sizable pre-announcement equity premium. (Presented at CFRC 2024, CIFFP 2023, Seminar Series at Central University of Finance and Economics and Tsinghua University.)

We find that anomaly returns are generally unchanged during FOMC days. The average return on the long- and short-leg, of a comprehensive set of 207 anomalies, increases by 26.3 bps and 28.8 bps, respectively, prior to the FOMC and reverses back afterwards. But for a small group of anomalies that do have substantial changes, their profitability tends to go down with absolute pricing errors greater than usual. Our evidence challenges existing studies that find the CAPM perform better during the FOMC period. Furthermore, we uncover that the less participation of retail investors contributes to the decline of profitability. (Presented at China International Conference in Finance (CICF) 2023, Tsinghua SEM Seminar Series, Renmin University Seminar Series, Tongji University Seminar Series, Nanjing University Seminar Series.)

Robot investment assistants (RIAs) are designed to help individual investors with investment decisions by providing information and advice services. Using account level data between 2020 and 2021 from the largest mutual fund investment platform in China, we examine the values of different RIA services. Higher usage of RIAs is associated with higher future raw and risk-adjusted returns, higher diversification, risk-taking, and turnover for individual investors. We further find advice services, rather than information services, are playing a more important role in investors’ future returns and trading activities. We find no definite evidence that existing RIA services alleviate behavioral biases. (Presented at China International Conference in Finance (CICF) 2022, Joint Conference on Statistics and Data Science (JCSDS) 2023, FinTech Seminar Series at Peking University.)