DEA–DA for bankruptcy-based performance assessment: Misclassification analysis of Japanese construction industry

This study describes a practical use of Data Envelopment Analysis–Discriminant Analysis (DEA–DA) for bankruptcy-based performance assessment. DEA–DA is useful for classifying non-default and default firms based upon their financial performance. However, when we apply DEA–DA to a data set on corporate bankruptcy, we usually face three problems. First, there is a sample imbalance problem because the number of default firms is often limited. In contrast, we can easily obtain a large number of non-default firms. Second, there is a computational problem to deal with a large data set. We need to consider a computational strategy to reduce the dimension of a large data set. Finally, we need to consider data alignment because the location of default firms may exist within that of non-default firms. This study discusses a simultaneous occurrence of the three problems from the perspective of Japanese industrial policy on construction business. To handle the three problems, this study combines DEA–DA with principal component analysis to reduce the computational burden and then alters DEA–DA weights to address both the sample imbalance problem and the location problem. This study also discusses a combined use between DEA–DA and rank sum tests to examine statistically hypotheses related to bankruptcy assessment. As an important application, we apply the proposed approach to the Japanese construction industry and discuss why many Japanese construction firms are misclassified.