The Choice of Comparable DMUs and Environmental Variables

Abstract

Data envelopment analysis was introduced as a nonparametric linear programming estimator to estimate the production frontier and technical efficiency of observed production units. Allowing multiple inputs and multiple outputs, the approach can be considered a nonparametric quantile regression that approximates the production frontier using piecewise linear approximations to the frontier surface. Purportedly, the method provides a benchmarking approach by comparing a given inefficient unit to a convex combination of relatively efficient units. In many applications, relevant production variables are fixed or uncontrollable leading to frontiers that are conditional on these fixed factors. In public sector applications, the amount of output that can be achieved for a given level of discretionary inputs might depend on the socioeconomic environment. In education, for example, student achievement has been shown to depend not only on discretionary inputs but also on poverty and other socioeconomic variables. In this chapter, DEA models that control for environmental variables and the referent frontier decision making units (DMUs) will be presented.