The Role of Corporate Governance and Estimation Methods in Predicting Bankruptcy

In a sample covering bankruptcies in public US firms in the period 2000 to 2015, we find that the addition of governance variables significantly improves the classification power and prediction accuracy of various bankruptcy prediction models. We also find that while adding governance variables improves the performance of bankruptcy prediction models, the additional explanatory power provided by adding the governance measures improves the further we are from bankruptcy, which implies that governance variables tend to provide earlier and more accurate warnings of the firm’s bankruptcy potential. Our analysis of five commonly used statistical methods in the literature showed that regardless of the bankruptcy model used, hazard analysis provides the best classification and out-of-sample forecast accuracy among the parametric methods. Nevertheless, non-parametric methods such as neural networks or data envelopment analysis appear to provide better classification accuracy regardless of the model selected.