报告题目：ChatGPT, Stock Returns and Links to the Macroeconomy
报告人：周国富 华盛顿大学圣路易斯奥林商学院
报告时间：2024年6月25日(周二) 16:30-18:00
报告地点：中国科学院数学与系统科学研究院南楼N204
腾讯会议ID：638 203 071
内容摘要
We study three questions on ChatGPT which has a increasing applications across all areas First, we examine whether ChatGPT can identify useful news content for the aggregate stock market and macroeconomy. Using the news and headlines of Wall Street Journal, we find that the information extracted by ChatGPT is highly related to macroeconomic conditions, and that the good news ratio predicts future returns both in- and out-of- sample. We explore various economic reasons why there are underreactions to good news. Second, we devise a prompting strategy for using ChatGPT to identify and analyze unusual aspects of financial communications, focusing on earnings calls of S&P 500 firms across 25 dimensions. We find that the stocks reacts negatively to unusual communications. Third, we explore the power of specific learning and find that it can potentially outperform general LLM ones.
主讲人简介
Guofu Zhou is Bierman and Spears Professor of Finance at Olin Business School of Washington University in St. Louis. He has a BS degree from Chengdu College of Geology, China, a MS from Chengdu Branch, Academia Sinica, and a PhD in economics from Duke University. His research interest covers a wide range in finance, with current work primarily in AI, big data and machine learning such as exploring new factor models, proposing new machine learning methods for forecasting expected asset returns, estimating announcement risk premia, and studying the market impact of the ESG and various sentiment measures.
