Testing the Impacts on Inefficiency in a Semiparametric Stochastic Frontier Model

This paper is concerned with significance testing of the effects of exogenous determinants upon the one-sided deviation term of a semiparametric stochastic frontier model. Two nonparametric significance tests for all or a subset of the determinants of inefficiency are proposed. The proposed tests are based on conditional moment restrictions and stochastic processes, with critical values being simulated by means of a multiplier bootstrap procedure. Our testing methodology addresses the omitted variable bias that arises naturally in stochastic frontier models when accommodating the determinants of inefficiency and accounts for the estimation effects that appear by using the estimated composite error when constructing the test statistics. We investigate the theoretical properties of the proposed tests and the resampling approximations. The proposed tests are illustrated through simulation experiments and two empirical examples in which the hypotheses of no impacts on inefficiency need to be tested