A recent study by the University of Chicago reveals that large language models (LLMs), such as GPT-4, can analyze financial statements with an accuracy that rivals and sometimes surpasses human analysts. The findings were published in a paper titled “Financial Statement Analysis with Large Language Models.”
The researchers tested GPT-4’s ability to predict future earnings growth from corporate financial statements. Remarkably, the model outperformed human analysts, especially in challenging situations where human predictions often falter. “Even without narrative or industry-specific information, the LLM outperforms financial analysts in its ability to predict earnings changes,” the paper notes. Moreover, the model’s accuracy is comparable to specialized state-of-the-art machine learning models.
LLMs have a significant edge due to their extensive knowledge base and rapid pattern recognition capabilities. These models can make intuitive decisions even with incomplete data, generating useful insights into a company’s future performance. The study also found that trading strategies based on GPT-4’s predictions yielded higher Sharpe ratios and alphas than those based on other models.
While human financial analysts will remain essential for the foreseeable future, this research suggests that LLMs can be valuable tools, enhancing analysts’ work by providing more informed and accurate insights. Over time, the integration of LLMs could transform the role of financial analysts, making their work more efficient and precise.
This breakthrough points to a future where LLMs play a crucial role in financial analysis, potentially reshaping how analysts approach their work and make decisions.