A composition model for two‐stage networks in data envelopment analysis

Abstract

The efficiency evaluation focusing on network systems in data envelopment analysis (DEA), utilizing models such as the SBM‐NDEA (network slacks‐based measure model), frequently encounters challenges in defining stage efficiencies, frontier projection, determination of slacks and targets, and the treatment of potential conflicts between stages. In order to overcome the limitations of traditional additive and multiplicative decomposition approaches, this paper proposes the UNSBM (unique stage efficiencies network SBM model). The UNSBM adopts the composition paradigm and a bi‐objective approach based on the SBM‐NDEA multiplier. The main contribution of the model lies in its ability to provide unique and neutral stage efficiency scores, concurrently managing potential conflicts. Furthermore, the study presents a target‐setting model designed to guide inefficient units toward the efficiency frontier. When applied to a real dataset of Taiwanese non‐life insurance companies, the UNSBM demonstrated advantages in the accurate identification of inefficiency sources and the definition of improvement plans, thereby validating its effectiveness, compared to existing methods.