Accents, AI Captioning, and Information Transmission in Earnings Conference Calls

Abstract

We use AI-based tools to examine how the spoken accents of managers and analysts during earnings conference calls affect information transmission. We find that calls with a higher share of accented analysts exhibit weaker immediate price reactions to bad news, wider post-call bid–ask spreads, and larger forecast errors among peer analysts. Following the adoption of real-time AI captioning, accent-related penalties largely persist, though forecast accuracy improves on average. Analyst race introduces important heterogeneity: in accent-heavy calls with greater Asian analyst participation, market adjustment improves and peer forecast errors decline after captioning adoption. Overall, accents create comprehension frictions, but as AI captioning lowers processing costs, technically detailed questions raised by accented analysts improve the transparency of market information.