Listen Closely : Using Vocal Cues to Predict Future Earnings

In this study, we aim to advance the prediction of firm earnings - an important task for many business applications. While existing earnings prediction models only rely on numerical financial data, we hypothesize and find that vocal cues from manager speech yield substantial predictive power: the area under the receiver operating characteristics curve ranges from 61.20% to 63.44%, significantly higher compared to models based on detailed financial data and textual inputs. We further analyze the models' economic value to investment practitioners. We find that investors can use the models' earnings forecasts to implement trading strategies that beat the market by 8.8% on average per year. Moreover, financial analysts can use vocal cues to improve their earnings forecast accuracy by more than 40%. Collectively, our results imply that managers' vocal cues are important information signals for future earnings that investment practitioners currently overlook