On variable reductions in data envelopment analysis with an illustrative application to a gas company

Abstract

Data envelopment analysis (DEA) is a non-parametric data oriented method for evaluating relative efficiency of the number of decision making units (DMUs) based on pre-selected inputs and outputs. In some real DEA applications, the large number of inputs and outputs, in comparison with the number of DMUs, is a pitfall that could have major influence on the efficiency scores. Recently, an approach was introduced which aggregates collected inputs and outputs in order to reduce the number of inputs and outputs iteratively. The purpose of this paper is to show that there are three drawbacks in this approach: instability due to existence of an infinitesimal epsilon, iteratively which can be improved to just one iteration, and providing non-radial inputs and outputs and then capturing them. In order to illustrate the applicability of the improved approach, a real data set involving 14 large branches of National Iranian Gas Company (NIGC) is utilized.