Sensitivity for data envelopment analysis with an application to Texas agricultural production.

摘要

For decades, management scientists have tried to develop a satisfactory procedure for assessing the relative technical efficiency of production units with multiple inputs and multiple outputs. A promising new methodology for classifying such production units as efficient or inefficient, Data Envelopment Analysis (DEA), is superior to existing methods which are usually too complicated to be of much practical use. Although there are many DEA models to choose from, the additive model is the one of primary interest for this dissertation. The objective of this dissertation is to determine the stability (or sensitivity) of each production unit's classification. A radius of stability is computed, providing the manager or economist with a measure of a classification's robustness to changes in the inputs and outputs. A small radius of stability indicates that small changes or perturbations can change the classification. Organizations whose production units exhibit this characteristic, whether currently efficient or inefficient, should be alerted. Conversely, efficient production units with large radii of stability might well be labelled "robustly" efficient. In each case, the exact perturbation which limits expansion of the "stability region" can be explicitly computed.