Dictionaries for Post-bankruptcy Success Prediction: A Machine Learning Approach

Abstract

This is the first study to analyze bankrupt firms’ reorganization plans. Using machine learning, we generate a dictionary for predicting post-bankruptcy success from these documents. Word counts based on our dictionary predict post-bankruptcy survival even after considering variables utilized in previous studies. Our text-based metrics are the strongest predictors of firm survival in our analysis and are also informative about the operating performance of surviving firms. Our results highlight the potential of reorganization plans for predicting post-bankruptcy success. We demonstrate that established dictionaries mostly evaluate reorganization plans incorrectly, which emphasizes the need for context-specific dictionaries in finance and accounting research. 

Keywords: post-bankruptcy performance; textual analysis; machine learning

JEL Classification: G33