Can ChatGPT Forecast Stock Price Movements? Return Predictability and Large Language Models
Alejandro Lopez-Lira, Yuehua Tang
TL;DR
This paper demonstrates that off-the-shelf large language models, particularly GPT-4, can forecast the immediate stock-market reaction to news headlines and anticipate short-horizon price drift, especially for small and negatively toned news. It establishes a theoretical framework linking LLM processing capacity to information frictions and limits to arbitrage, showing a threshold model where only sufficiently sophisticated LLMs generate profitable signals. Empirically, GPT-4's predictions produce high initial hit rates and meaningful drift, with model size and news complexity shaping performance; the results also reveal heterogeneous processing across news topics. The findings imply that AI-driven enhancement of information processing can improve market efficiency over time, though adoption dynamics may initially sustain some return predictability. The paper contributes both a novel empirical instrument for studying market information processing and a formal model connecting AI capabilities to price formation and efficiency.
Abstract
We document the capability of large language models (LLMs) like ChatGPT to predict stock market reactions from news headlines without direct financial training. Using post-knowledge-cutoff headlines, GPT-4 captures initial market responses, achieving approximately 90% portfolio-day hit rates for the non-tradable initial reaction. GPT-4 scores also significantly predict the subsequent drift, especially for small stocks and negative news. Forecasting ability generally increases with model size, suggesting that financial reasoning is an emerging capacity of complex LLMs. Strategy returns decline as LLM adoption rises, consistent with improved price efficiency. To rationalize these findings, we develop a theoretical model that incorporates LLM technology, information-processing capacity constraints, underreaction, and limits to arbitrage.