This paper uses maximum decisional efficiency (MDE) principle to derive cross efficiency (CE) scores for input and output oriented frontier efficiency models. The MDE CE models are parametric models that derive their cross efficiencies by either maximizing log-likelihood (input-oriented models) or minimizing negative log-likelihood (output-oriented models). Using real-world and simulated datasets, we compare our MDE models with several competing CE models from the literature. Our results illustrate that the MDE models based CE scores have higher CE averages when inputs are independent or correlated with half-normal inefficiency distributions. We also find that MDE models provide model consensus scores that are highly consistent. When inputs are correlated and inefficiency distributions are exponential, the log-likelihood estimation procedures appear to suffer in performance when compared to data envelopment analysis models with secondary objectives.
All Science Journal Classification (ASJC) codes
- Computer Science(all)