Inference in stochastic frontier analysis with dependent error terms

Stochastic frontier analysis (SFA) is often used to estimate technical eciency of entities such as rms, countries or municipalities. A potential dependence between the two components of the error term can be taken into account by a copula function. While estimation of the model is straightforward using the Corrected Ordinary Least Squares (COLS) and Maximum Likelihood (ML) methods, an open issue concerns the inference of the technical eciencies. We propose a parametric bootstrap algorithm which is an extension of an algorithm proposed by Simar and Wilson [18] to the dependence case. This allows us to estimate the eciency percentile condence intervals. We apply the model to the estimation of technical eciencies of moroccan municipalities.