Yi Cao, George Mason University – A.I. and Earnings Calls

Earnings calls can tell us a lot about a firm.

Yi Cao, assistant professor of accounting at George Mason University, wonders whether AI can help us learn even more.

Yi Cao is an assistant professor of accounting at the George Mason University School of Business. His research interests include corporate voluntary disclosure, product market competition, financial disclosure quality, and textual analysis. Yi has also held academic positions at the School of Management and Economics of the Chinese University of Hong Kong, Shenzhen as an assistant professor of accounting. He holds her Ph.D. in accounting from the University of Maryland, R.H. Smith School of Business. His work has been published in published in Contemporary Accounting Research. His work has also been featured in media outlets such as the Financial Times and Bloomberg.

A.I. and Earnings Calls

Ever wondered how big companies communicate their finances and future plans to the world? Earnings calls, which are basically a “behind-the-scenes” look at corporate financial updates, play a significant role in keeping investors in the know. Earnings calls are meetings held by companies where executives and managers discuss the company’s financial performance with shareholders and analysts. These calls provide insight into the company’s financial health and upcoming plans which serves as a key source of information for investors.

In earnings calls, a significant portion of information comes from what corporate managers say. However, trying to gauge the amount of information they contribute poses a challenge due to the unstructured nature of human language and the pre-existing public information of the capital market.

For a new working paper, my colleagues and I thought of a way to use ChatGPT and other AI models as a solution. After feeding the AI model more than 100,000 transcripts of earnings calls from publicly listed companies, we asked it to generate answers to the questions asked by the financial analysts during the calls. Then, we analyzed the degree of dis-similarity between what top executives actually said and the generic responses from the chatbot.

Interestingly, we found that the degree of difference between human and AI generated answers is informational – it predicts market reactions such as abnormal returns, abnormal trading volume, stock liquidity and analyst forecast errors. In other words, the more incremental information being provided by the managers, the better information environment for the firm around the conference call.

This makes a lot of sense, when you consider that boilerplate language can be used as a cover-up when the outlook of a company’s fundamentals isn’t clear. On the other hand, when a company is riding high with a clear path ahead, executives will want to crow about it in detail.

Therefore, we suggest that AI can be used as a tool to help investors and analysts read between the lines of executives’ public statements and discussions.

Read More:

Email: ycao25@gmu.edu

LinkedIn: https://www.linkedin.com/in/yi-cao-82678516/

Relevant publications:

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4150829

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3605303

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3794690

https://www.ft.com/content/8eebc172-1f7b-44fe-a12e-e890d123f7e9