Abstract
Data Envelopment Analysis (DEA) has been used to supply decision makers and analysts with new insights into the efficiency of peer entities called decision making units (DMUs). The advantage of DEA is that it provides an objective data-driven assessment of performance, free of user bias. However, because factor weights are determined by an algorithm and not a priori, many researchers and practitioners have difficulty understanding DEA models and the scores they produce. This may explain why DEA is seldom covered in university courses in the decision sciences. The result of this lack of awareness and understanding is that DEA is underutilized as a performance measurement tool in commercial, government, and military operations. This chapter aims to address this issue by providing a lucid overview of DEA, replete with examples and suggestions to make DEA more accessible for researchers and practitioners alike. Additionally, our didactic approach includes step-by-step instructions for preparing data, choosing DEA models, and avoiding pitfalls.
TopBackground
The first DEA model was developed by Charnes, Cooper, and Rhodes (1978), known as the CCR model, and used the ratio of weighted outputs to weighted inputs to measure the relative efficiency of DMUs, where the weights were determined via a constrained optimization model. Banker, Charnes, and Cooper’s (1984) BCC model extended the CCR model by introducing a convexity constraint which allowed for variable returns to scale. This allowed the separate measurement of technical and scale efficiencies. The CCR and BCC models are considered radial models because they measure radial distances between DMUs and the efficient frontier; a frontier comprised of a set of DMUs not dominated by any other DMU. Stemming from these early works, several models have since been developed to address specific questions in a variety of operational settings.
Key Terms in this Chapter
Data Envelopment Analysis (DEA): A series of models that measure relative efficiency.
Effectiveness: The degree to which goals are met, or results are achieved.
Principal Component Analysis (PCA): A data reduction technique used to reduce a data set down to its uncorrelated, orthogonal components.
Undesirable Outputs: Externalities, side effects, and adverse impacts produced or resulting from the course of operations.
Efficiency: The degree to which outputs are produced from inputs.
Manhattan Distance: The distance between two points which is calculated by summing the absolute differences of their coordinates.
Decision Making Unit (DMU): Represents a unit (process, product, polity, etc) under evaluation.