Flexible Estimation of Heteroskedastic Stochastic Frontier Models via Two-step Iterative Nonlinear Least Squares

This article illustrates a straightforward and useful method for incorporating exogenous inefficiency effects in the estimation of semiparametric stochastic frontier models. An iterative estimation algorithm based on two-step nonlinear least squares is developed allowing for any flexible and monotonic specification of the production technology. We investigate the behavior of the proposed procedure through a set of Monte Carlo experiments comparing its finite sample properties with those of available alternatives. The new algorithm provides very good performance, outperforming the competitors in small samples and in presence of small signal-to-noise ratios. Two applications to agricultural data illustrate the usefulness of the proposed algorithm, even when it is used as a tool for sensitivity analysis