An Analysis of ChatGPT as a Tool for Portfolio Stock Selection

Abstract

Background: Artificial Intelligence has had a significant impact on numerous industries including finance with the financial sector having undergone substantial changes in recent years with the rise of new technologies and increasing reliance on data-driven decision- making processes. Purpose: The purpose of this study was to explore whether ChatGPT can serve as a tool for assisting with investment strategies that feature reduced costs and are more accessible to the public with the goal being to assess if such AI models could improve stock selection and portfolio creation for the average individual investor. Method: Portfolios were constructed by prompting ChatGPT to make a decision on whether to invest or not to invest based on whether an investment into that company would meet the investment objectives of a corresponding existing fund managed portfolio. Next the returns as well as the standard deviations (measure of risk) were computed for these portfolios as well as the existing fund managed portfolios. A difference between groups methodology was used to compare the ChatGPT portfolios and the existing fund managed portfolios from a return generation and risk minimization point of view. The Mann-Whitney U test was utilized to determine whether the differences in the ChatGPT and the corresponding existing fund managed portfolios, were statistically significant. Results: ChatGPT outperformed the existing fund managed portfolios in 9 out of the 18 portfolios from a return generation perspective, indicating some potential of ChatGPT to generate returns. However, the results were not statistically significant, and this study could not conclude that ChatGPT can outperform fund managers in generating portfolio returns. The ChatGPT portfolios had much larger standard deviations and therefore risk levels than those of the existing fund managers. These results were statistically significant indicating that ChatGPT struggles from a risk minimisation point of view. Overall, the findings suggest that ChatGPT has potential as a complementary tool for individual investors and financial professionals, particularly in growth-oriented strategies. However, its inability to minimize viii risk and its reliance on generic responses underscore the need for cautious integration into investment practices. The results also emphasize that ChatGPT is not yet equipped to replace professional fund managers due to ChatGPT being unable to minimise portfolio risk and generate stable returns. Contribution: This study adds to the literature by exploring ChatGPT's ability to recommend stocks and create portfolios based on existing investment strategies of asset managers at reputable asset management companies in South Africa as the use of ChatGPT in finance increases and more individuals turn to the large language model for various tasks.