This research paper explores the ability of large language models (LLMs), specifically ChatGPT, to predict stock price movements using news headlines. The authors demonstrate that ChatGPT can indeed predict stock returns, particularly for smaller stocks and following negative news, outperforming traditional sentiment analysis methods. They propose a theoretical model to explain this phenomenon, which considers information processing constraints, underreaction, limits-to-arbitrage, and the capabilities of LLMs. The authors also introduce an interpretability framework to analyze the reasoning behind ChatGPT's predictions, highlighting its strengths and limitations. Finally, they investigate the potential impact of widespread LLM adoption on market efficiency, suggesting that these technologies could enhance market efficiency by reducing underreaction to news.
Paper:
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4412788
Join Robinhood