Good and bad outputs in bank performance: A relational network DEA approach

This paper applies the relational network data envelopment analysis model in order to examine the efficiency of Japanese banks. The network framework allows deposits to be treated as an intermediate variable and as a result it provides a solution to the deposits dilemma. Furthermore, we incorporate non-performing loans into to our analysis by modifying the traditional relational network model. In order to do so, we assume that the existence of good and bad loans are not necessary for each other. Although (total) loans are needed to be generated in order for bad loans to exist, bad loans are not necessary to be produced as a byproduct of good loans. Therefore, we assume that good and bad loans are not weakly disposable. The incorporation of both good and bad loans into the analysis provides a more accurate representation of the banking system and offers a richer level of information to the decision makers. There is a large dispersion of inefficiency levels in Japanese banking system.