Investor Sentiment and Stock Market Volatility: A Study Using Social Media Data and Financial Indicators

Social-media sentiment, Twitter, investor attention, tweet volume, stock returns, stock-price movement, machine learning, financial markets. markets, Twitter.

The study analyzes the correlation between the sentiment and social-media attention toward the stock and the stock price movement by examining daily data from Twitter and the stock market for 25 stocks selected for one year starting from September 2021 to September 2022. The study accounts for the direction as well as the size of the following day's price change. Sentiment scores, proportions, and probabilities measure sentiment; tweet volume is the main indicator of attention. Descriptive statistics, correlation analysis, lagged analysis, attention-group comparisons, robust regression, stock fixed-effects models, and machine-learning methods with a strict chronological train-test procedure are used for the analysis. The results show that sentiment from social media is only marginally and inconclusively associated with later stock returns. The predictive power of classification models was found to be under the 55.12% majority class baseline and XGBoost did not perform well. By contrast, social-media attention showed a weak and positive correlation with the size of future price changes. The magnitude of absolute next-day returns rose from 2.05% in the low-attention observations to 2.38% in the high-attention observations. The relationship between tweet volume and absolute returns was positive and the relationship was robust to controlling for market variables and stock fixed effects. The results indicate that the attention paid by social media might be more indicative of the magnitude of stock-price movements than of their direction. But these findings suggest association, not causation.