Speech Emotion Recognition from Earnings Conference Calls in Predicting Corporate Financial Distress

Abstract

Sentiment and emotion analysis is attracting considerable interest from researchers in the field of finance due to its capacity to provide additional insight into opinions and intentions of investors and managers. A remarkable improvement in predicting corporate financial performance has been achieved by considering textual sentiments. However, little is known about whether managerial affective states influence changes in overall corporate financial performance. To overcome this problem, we propose a deep learning architecture that uses vocal cues extracted from earnings conference calls to detect managerial emotional states and exploits these states to identify firms that could be financially distressed. Our findings provide evidence on the role of managerial emotional states in the early detection of corporate financial distress. We also show that the proposed deep learning-based prediction model outperforms state-of-the-art financial distress prediction models based solely on financial indicators.