Using Generative AI to predict the weather impact on future stock returns
摘要
This study explores the use of Generative AI, specifically OpenAI’s ChatGPT, for forecasting the impacts of severe weather events on stock returns. Employing prompts that assess textual weather descriptions, ChatGPT, a powerful generative AI large language model (LLM), provides predictions incorporated into econometric models. Results show that when ChatGPT forecasts negative stock impacts from storms, larger, more profitable firms with lower leverage and higher liquidity experience lower subsequent returns, suggesting investor underreaction to weather risk. ChatGPT’s predictive abilities are stronger during favorable economic conditions like uptrends, low volatility, and robust employment growth, implying investor underreaction amid bullish sentiment.