Abstract:

This study investigates the influence of hedge fund managers’ moods on fund performance before and during the COVID-19 pandemic by analyzing approximately 560,000 English- language tweets from 230 active managers. Using BERT-based deep learning models and LDA topic modeling, we extract sentiment indices from both original tweets and interactive engagements (retweets, replies, and quotes). Results show that sentiment derived from interactive tweets has a significant and positive impact on hedge fund returns, particularly during the pandemic, while sentiment from original tweets exhibits a negative relationship with performance. Cointegration tests confirm a long-term relationship between managers’ moods and fund performance, with stronger short-term effects observed during the COVID-19 period. The findings suggest that social interaction amplifies the predictive power of sentiment, possibly by mitigating biases associated with overconfidence or emotional contagion in isolated self-expression. This research highlights the role of digital discourse as a proxy for managerial sentiment and contributes to the growing literature at the intersection of behavioral finance, social media analytics, and fund management.