Modification of Variable Returns to Scale Stochastic Data Envelopment Analysis (DEA) Models

Data Envelopment Analysis (DEA) was introduced under the name of a deterministic model assuming all the deviations from the estimated production frontier were one sided indicating technical inefficiency. Biased estimations of inefficiency and production are provided by the model when deviations do not originate only from inefficiency but also from measurement errors. In 1988, Banker developed Data Envelopment Analysis as a stochastic model to reflect inefficiency and statistical noise simultaneously. However, from deterministic to stochastic, the problem with weak efficient frontiers and related biased results stayed the same. This dissertation proposes a modification over Banker’s stochastic DEA (SDEA) model by applying a limitation on the coefficients of inputs in the original model in order to change weak efficient hyperplane(s) while keeps general assumptions behind production function unaffected. This can change the production possibility set (PPS) while the frontier has the potential to give a better representation of the true production frontier. Comparing the results from the stochastic model and suggested modified model shows that the achieved model is providing a new benchmark for relative efficiency evaluation and production frontier estimation. Keywords: Data Envelopment Analysis (DEA), Stochastic Data Envelopment Analysis (SDEA), Modified Model, Weak Efficient Frontier.