Selection of Inputs and Outputs in Data Envelopment Analysis

Abstract

Productive efficiency lies in producing the maximum output from a given bundle of inputs or using the minimal input for a target bundle of outputs. The method of Data Envelopment Analysis (DEA) introduced in the OR/MS literature in the late 1970s and subsequently refined and extended over the decades has become a popular analytical device for measurement of efficiency. However, in many empirical applications inadequate attention is paid to the selection of inputs and outputs. Production is the process of creating value through transformation of inputs into outputs. It is important to ensure that the resources defined as inputs in a specific context do in fact contribute to the outcomes treated as outputs. In this paper we start with the scope of decision making by the producer to define the ‘boundary’ of the firm. This enables us to distinguish inputs (resources that enter into the jurisdiction of the firm from outside) and outputs (that get out of the boundary and are not subject to further processing by the firm). We visualize a firm as a vertically integrated organization with sub-centers of decision making at different stages of production. This allows us to differentiate an intermediate output (or a throughput) from a pure output or input. We discuss the appropriate choice of inputs and outputs in different areas of empirical application including manufacturing, banking, education, and health care. Special attention is paid to the treat of undesirable outputs (like pollution and industrial waste) in DEA. Finally we consider contextual or environmental variables that affect production but are not subject to manipulation by the producer.