Abstract
The traditional textual analysis relies on a pre-specific structure to generate sentiment analysis that leads to recommended portfolios. Generative AI, such as ChatGPT, relies on reasoning ability to generate portfolio recommendations that otherwise cannot be generated by traditional textual analysis. Based on policy announcements by the Chinese government, we show that ChatGPT can generate a portfolio with a monthly three-factor alpha of 90 basis points. As we adjust the parameter of ChatGPT to allow for a higher degree of recommendation flexibility, ChatGPT could potentially generate portfolios that outperform the markets in out-of-sample tests.
Type
Publication
Available at SSRN: https://ssrn.com/abstract=4519182
