A modification of relational two-stage DEA with variable returns to scale

Traditional studies in two-stage data envelopment analysis (DEA) assumed that the outputs of the first stage completely input into the second stage. But we think that assumption is not realistic becauses of efficiency loss. This paper modifies the traditional relational two-stage DEA with Variable Returns to Scale (VRS) by abandon this assumption. By introduction of efficiency loss to measure the difference between Intermediate outputs and Intermediate inputs, the traditional model can be modified, and the new model is proved to be more powerful and usesful in calculating efficiency. In order to test the model, a case of 14 Chinese listed banks is usesd as an example to illustrate the whole idea.