In this study, we explore the potential of using a Large Language Model (LLM) like GPT4 for financial statement analysis compared to human analysts. Surprisingly, the LLM outperforms human analysts in predicting earnings changes, even without specific industry knowledge. The LLM’s prediction accuracy matches that of a specialized machine learning model, hinting at its potential for decision-making. Notably, the LLM generates valuable insights about a company’s future performance through narratives. Trading strategies based on GPT4’s predictions show higher Sharpe ratios and alphas, indicating promising outcomes. Overall, the study suggests that LLMs could play a significant role in financial analysis and decision-making.
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4835311