Can « ChatGPT » mentions in regulatory filings move stock prices?
Marcin Pietrzak studies short-term market reactions to corporate ChatGPT-related announcements in his paper « A trillion dollars race: how ChatGPT affects stock prices ».
He analyses 6-K and 8-K current reports from January to May 2023 among Russell 3000 firms to identify whether ChatGPT-related corporate communication affects abnormal returns.
His conclusions include:
- 59 reports contain « ChatGPT », with 23 out of 59 generating statistically significant abnormal returns on day 0 or 1, which shows not every AI-related disclosure is market-moving.
- Negative announcements are dominated by one firm: Chegg’s stock price declined by 25% after their February 2023 report, and 50% after their May 2023 one warning about ChatGPT’s impact.
- Consumer staples and information technology tend to benefit from ChatGPT mentions, while communication, energy, financials, and materials are negatively affected.
- Out of the 12 positive corporate reports, health care, industrials, information technology, and real estate show positive abnormal returns, while consumer staples, financials, and utilities show negative ones.
- Information technology appears structurally advantaged: the sector displays positive abnormal returns both when announcements are framed as threats and when they are framed as opportunities.
- Firms with high beta loads get more positive effects, while high price-to-earnings ratios and younger firms are associated with lower abnormal returns.
This research suggests « ChatGPT » mentions are potential catalysts for abnormal returns, but highlights sparse significance and strong firm and sector dependence.
Investors can use the same kind of filing-based screens to separate threat- from opportunity-type disclosures and determine whether sector exposures amplify or hedge the shock.
The author highlights several limitations to his study, like its focus on short-term price reactions, abnormal returns attribution to the announcements, or the sample restriction to Russell 3000 firms.