The Sound of Doubt: A Computational Linguistics Approach to Examining Managerial Speech Uncertainty in Conference Calls

Abstract

We study whether the uncertainty expressed via the acoustic features of managerial speech in conference calls impacts analyst behavior using computational linguistic techniques. Our study builds on the literature in linguistics and psychology showing that the manner in which individuals speak can provide information above and beyond the actual spoken words. We predict that when managers express greater acoustic uncertainty in their responses to questions during conference calls, analysts rely less on the information provided by the manager, affecting their subsequent forecasts. Using over 30,000 hours of conference call audio, we develop a novel approach to capturing managerial uncertainty based several acoustic features of their speech, such as hesitation and stuttering. We find that when managers sound more uncertain in their responses to analyst questions, analyst forecast dispersion and errors increase, even after accounting for characteristics of the actual language being used by managers and analysts. We also find that analysts are less likely revise their recommendations in response to news from the call when responses are conveyed in an uncertain manner. Our findings suggest that the uncertainty expressed in the acoustic features of managerial speech can affect market participants. Furthermore, our study creates and validates a new approach to study the acoustic features of executive speech.