Large Language Models (LLMs) like GPT-4 have demonstrated proficiency in text analysis, interpretation, and generation, with their scope of effectiveness stretching to various tasks within the financial sector. However, doubts persist about their applicability for complex financial decision-making, especially involving numerical analysis and judgement-based tasks.
A key question is whether LLMs can perform financial statement analysis (FSA), a field that heavily emphasizes on numerical data and human judgment. FSA involves assessing a company’s financial situation and forecasting its future results using financial statements, alongside understanding financial ratios, trends, and company-specific details.
To explore this hypothesis, researchers from the University of Chicago conducted a study, assigning GPT-4 to analyze anonymized, standardized financial statements and forecast future earnings with only numerical data from the financial records, withholding any narrative or industry-specific information.
Remarkably, GPT-4 demonstrated superior ability to anticipate changes in earnings compared to human financial professionals, particularly in situations challenging to humans. These findings indicate that LLMs exhibit a unique advantage in managing intricate financial data, even without any narrative context. Additionally, GPT-4’s predictive power was observed to be equal that of specialized machine learning models trained specifically for this task. It could analyze and interpret financial data with high accuracy.
Contrary to expectations, the accuracy of GPT-4’s predictions was found not to depend on its training memory. Instead, the model utilizes the data it examines to produce insightful narrative about the company’s future performance. The researchers also investigated the usefulness of GPT-4’s forecasts in trading strategies and found that such strategies yielded higher alphas and Sharpe ratios, suggesting more successful outcomes and better risk-adjusted returns.
The study concluded that LLMs like GPT-4 could play a significant role in decision-making within the financial sector due to their robust performance in analyzing financial statements and providing insightful predictions. In the future, LLMs could entirely replace certain tasks currently undertaken by human analysts. Therefore, the role LLMs play in the financial industry is set to increase in the future, as their capabilities continue to evolve and improve.
