Benchmarking in data envelopment analysis: balanced efforts to achieve realistic targets

Abstract

AbstractThe minimum distance models have undoubtedly represented a significant advance for the establishment of targets in Data Envelopment Analysis (DEA). These models may help in defining improvement plans that require the least overall effort from the inefficient Decision Making Units (DMUs). Despite the advantages that come with Closest Targets, in some cases unsatisfactory results may be given, since improvement plans, even in that context, differ considerably from the actual performances. This generally occurs because all the effort employed to reach the efficient DEA frontier is channeled into just a few variables. In certain contexts these exorbitant efforts in some inputs/outputs become unapproachable. In fact, proposals for sequential improvement plans can be found in the literature. It could happen that the sequential improvement plans continue to be so demanding in some variable that it would be difficult to achieve such targets. We propose an alternative approach where the improvement plans require similar efforts in the different variables that participate in the analysis. In the absence of information about the limitations of improvement in the different inputs/outputs, we consider that a plausible and conservative solution would be the one where an equitable redistribution of efforts would be possible. In this paper, we propose different approaches with the aim of reaching an impartial distribution of efforts to achieve optimal operating levels without neglecting the overall effort required. Therefore, we offer different alternatives for planning improvements directed towards DEA efficient targets, where the decision-maker can choose the one that best suits their circumstances. Moreover, and as something new in the benchmarking DEA context, we will study which properties satisfy the targets generated by the different models proposed. Finally, an empirical example used in the literature serves to illustrate the methodology proposed.