Natural Language Processing Sentiment Analysis of S&P500 Earnings Calls and Abnormal Stock Returns
Open Access
Author:
Sipe, Warren
Area of Honors:
Finance
Degree:
Bachelor of Science
Document Type:
Thesis
Thesis Supervisors:
J. Randall Woolridge, Thesis Supervisor Brian Spangler Davis, Thesis Honors Advisor
Keywords:
Natural language processing stock returns earnings calls
Abstract:
Corporate earnings calls offer investors in the public markets the opportunity to hear from executives directly about performance, risks, and forward-looking strategy. This direct interaction with management reveals information about corporate operations which would not otherwise be publicly available. A potential untapped source of this information is the tone those executives employ in the call and the sentiment about past and future performance which it represents. This thesis will consist of a natural language processing (NLP) sentiment analysis of randomly selected S&P500 companies since 2016. Sentiment scores for these earnings calls will be analyzed to check for correlation with equity performance, both forward and backward looking, over a variety of timeframes. While this analysis does not indicate any utility of such analysis in predicting excess returns, it does reveal factors influencing call sentiment and provides avenues for further research.