Predicting the Direction of Movement of Abnormal Returns using Earnings Conference Calls and FinBERT

Abstract: Earning conference calls is an important communication tool for companies to provide relevant information about the latest quarter based on the reported earnings. However, the research on the impact of the earning conference calls has for long been a relatively unexplored subject. This study was born out of the ambition to try and expand the research done on earnings conference calls and the possibility to analyse them with FinBERT. This was conducted using a sample of 1118 quarterly earning conference calls, comprising 74 firms from the Information Technology Sector of the S&P 500 in an event study outline. The results suggest that the earning conference calls can be used for predicting the direction of abnormal returns with the help of FinBERT and machine learning. Moreover, the Support Vector Machines achieved the highest accuracy of the tested models on the classification problem.