Receive free High frequency trading updates
Sign up to receive a complimentary myFT Daily Digest email summarizing the latest news on High frequency trading every morning. While many CEOs may secretly enjoy earnings calls, they can also be stressful. The goal is not only to avoid saying anything foolish, but to also avoid saying anything of substance. This is where a recent paper by John Bai, Nicole Boyson, Yi Cao, Miao Liu, and Chi Wan comes in. They introduce a unique measure of information content called Human-AI Differences (HAID), which compares the responses given by corporate executives during earnings calls to those generated by Large Language Models (LLMs) like ChatGPT, Google Bard, and other open-source LLMs. The researchers found that HAID is a strong predictor of stock liquidity, abnormal returns, analyst forecast revisions, forecast accuracy, and manager guidance. These results emphasize the importance of using LLMs to uncover hidden insights within earnings calls.
According to finance columnist Matt Levine, some earnings calls closely resemble what ChatGPT would generate, with little new information revealed during the Q&A session. On the other hand, other earnings calls contain answers from executives that would not have been predicted by the chatbot. Interestingly, the non-robotic earnings calls were found to be more informative, leading to larger stock movements and more accurate future earnings forecasts. While this may seem paradoxical, it actually makes sense. CEOs don’t want to reveal too much information or set high expectations during earnings calls. They prefer to exceed expectations when the results are announced. CEOs are coached to be as generic as possible during these calls, using phrases like “at the end of the day” to avoid revealing too much. However, in today’s market, earnings calls are not only analyzed by analysts and investors, but also by trading algorithms. These algorithms can make trading decisions based on subtle cues in the CEOs’ language, creating volatility in stock prices. As a result, companies are now using language-AI systems to assess algorithmic responses to their prepared remarks and adjusting their language accordingly.
The language-AI systems have identified certain trigger words that can have a negative impact on stock prices, leading companies to reduce their usage. Some companies have even altered their tones to avoid triggering the algorithms. For instance, managers of firms with higher expected algorithmic readership tend to exhibit more positivity and excitement in their vocal tones, suggesting that managers are seeking professional coaching to improve their vocal performances in quantifiable ways.
In light of these findings, Levine proposed that ChatGPT or other LLMs should handle earnings calls, especially when the results are not favorable. This would eliminate the risk of revealing sensitive information or causing market disturbances. While this may be a radical idea, it highlights the potential role of LLMs in improving the efficiency and effectiveness of earnings calls.
Denial of responsibility! Vigour Times is an automatic aggregator of Global media. In each content, the hyperlink to the primary source is specified. All trademarks belong to their rightful owners, and all materials to their authors. For any complaint, please reach us at – [email protected]. We will take necessary action within 24 hours.
