Knowledge flows in Data Envelopment Analysis. The role of peer effects

Although the benchmarking literature has amassed both theoretical developments and empirical evidence on the performance of DMUs, the void related to the learning perspective of benchmarking remains to be filled. Further required methodological and empirical justifications aiming to fill this gap, are presented in this paper. Specifically, the mechanisms of knowledge transfer among the examined entities and the identification of the most influential source of learning require further investigation. We introduce a novel heuristic algorithm based on the Peer Removal to Improve Mean Efficiency in a Data Envelopment Analysis context to explore knowledge transfer within a benchmarking set. Based on sequential re-modifications of the technology following the removal of knowledge transmitters, a taxonomy arises including the role models, the knowledge receivers and the minimum efficiency DMU. Knowledge transfer is quantified by the calculation of the learning trace, following the removal of knowledge transmitters. We employ the most productive scale size to identify the most influential unit in terms of knowledge contribution. Findings from an illustrative example and a case study on European regions indicate that knowledge flows are not equally strong across benchmarking rounds.