Data Envelopment Analysis Models in Non-Homogeneous Environment

Abstract

Data envelopment analysis (DEA) is a non-parametric method that is widely used for relative efficiency and performance evaluation of the set of decision-making units (DMUs). It is based on maximization of a weighted sum of outputs produced by the unit under evaluation divided by the weighted sum of inputs of the same unit, and the assumption that this ratio for all other units has to be lower or equal to 1. An important assumption for applications of DEA models is the homogeneity of the units. Unfortunately, the homogeneity assumption is not fulfilled in many real applications. The paper deals with the analysis of efficiency using DEA models in the non-homogeneous environment. One of the problems lies in non-homogeneous outputs. In this case, the units under evaluation spend the same inputs but produce completely or at least partly different set of outputs. The paper formulates several models how to deal with this problem and compares the results on a numerical example. Other main sources of non-homogeneity are discussed as an excellent possible starting point for future research.