Machine learning in bank merger prediction: A text-based approach

This paper investigates the role of textual information in a U.S. bank merger prediction task. Our intuition behind this approach is that text could reduce bank opacity and allow us to understand better the strategic options of banking firms. We retrieve textual information from bank annual reports using a sample of 9,207 U.S. bank-year observations during the period 1994-2016. To predict bidders and targets, we use textual information along with financial variables as inputs to several machine learning models. We find that when we jointly use textual information and financial variables as inputs, the performance of our models is substantially improved compared to models using a single type of input. Furthermore, we find that the performance improvement due to the inclusion of text is more noticeable in predicting future bidders, a task which is less explored in the relevant literature. Therefore, our findings highlight the importance of textual information in a bank merger prediction task.