Financial Statement Analysis with Large Language Models
, Accounting PhD Student
, Assistant Professor of Accounting
, James H. Lorie Professor of Accounting and FMC Faculty Scholar
We (Kim, Muhn and Nikolaev) examine to what extent general-purpose large language models (LLM) such as GPT4 can make informed financial decisions based on mostly numerical financial data. We provide standardized and anonymous financial statements to a pre-trained LLM and design sophisticated chain-of-thought prompts that resemble how human analysts make earnings predictions. Our current results show that – even without additional narrative contexts or industry-specific information – the LLM outperforms the median financial analyst in its ability to forecast annual earnings. The overall prediction accuracy of the LLM is on par with the performance of a specifically trained artificial neural network model. We then show that LLM models’ prediction does not stem from its training memory. Lastly, our trading strategies based on GPT’s prediction yield a higher Sharpe ratio and alphas than strategies based on machine-learning-based models. Our preliminary results provide evidence on the potential of a general-purpose LLM in generating accurate, explainable financial forecasts. Our further analyses will more directly examine (i ) the value of incorporating textual data and (ii ) the source of GPT’s performance by testing different prompts in a factorial design.