Using Parametric Bootstrapping for Estimation of the Incidence of Inefficiency: Improving Features of Non-Parametric Bootstrap Estimator

This study proposes parametric bootstrap estimation methods to improve features of the non-parametric bootstrap estimator of Incidence of Inefficiency proposed in the literature. This study is the first to propose parametric bootstrap estimation methods to improve features of the IOI estimators. In the simulation study, the Maximum Likelihood-based parametric bootstrap method yields the best results in small sample sizes and a limited number of input and output variable situations. However, in cases where the number of input and output variables increases, which reduces the discrimination power of classical Data Envelopment Analysis models, the Bayesian estimator with latent variable adjustment tends to yield better results than the proposed estimators for IOI in the literature. Additionally, it is experimentally demonstrated that the parametric bootstrap based on the Bayesian method applied to a specific posterior distribution converges to the same features as the estimators obtained with the Bayesian estimator based on that posterior distribution.