This study investigates the association between individual investors' sentiment and short-term stock returns in the Chinese stock market. Employing the Albert deep learning algorithm, it analyses stock forum texts to develop a customized investor sentiment index tailored for the unique characteristics of the emerging market in China. The study demonstrates the superiority of this index over the naive Bayes method and support vector machines in constructing an accurate measure of individual investor sentiment. By employing ranking comparison, regression analysis of investment portfolios based on company characteristics, and assessing the performance of long-short portfolios, the study examines the impact of company characteristics on the significance of investor sentiment in influencing stock returns.
The empirical findings of this study reveal a positive correlation between sentiment and short-term stock returns in China, which stands in stark contrast to the long-term relationship observed in developed markets. Furthermore, the results demonstrate that sentiment has a more pronounced impact on the returns of small-cap, highly volatile, high book-to-market ratio, unprofitable, and financially risky companies that pose challenges in terms of evaluation and hedging. The performance of sentiment-based portfolios indicates that the observed effects are not solely driven by the concentration of high sentiment in these hard-to-hedge stocks. Additionally, by employing alternative sentiment indices and text analysis methods, the study consistently supports the reliability of the Albert method through repeated empirical evidence.
Keyword: Investor sentiment, Firm characteristics, Conditional stock returns, Sentiment analysis, ALBERT
07 91΄σΙρ 2024