Combining corporate governance indicators with stacking ensembles for financial distress prediction

Abstract In this paper, we use a stacking ensemble to construct a bankruptcy prediction model. We collect a comprehensive list of 40 financial ratios (FRs) and 21 corporate governance indicators (CGIs) for US companies, and conduct two experiments. In the first, we utilize all FRs and CGIs to build our model. Our results show that this model does not perform significantly better than the baseline models. In the second experiment, we use 6 specific FRs and 6 specific CGIs selected by a stepwise discriminant analysis to construct another model. We find that this model performs better than the baseline models, and exhibits strong performance when the costs of misclassifying bankruptcy companies are high.