We demonstrate how a recently developed model selection procedure can be used to choose among competing stochastic frontier models where inefficiency depends on firm characteristics. We provide evidence on the power of this procedure. Moreover, we examine the effects of model choice on estimation results. We find that different models can lead to rather different magnitudes of the partial effects of the exogenous factors. However, because the model selection procedure gives an unambiguous choice of best model, and because bootstrapping indicates that the procedure is reliable, we conclude that it does not matter whether other models give different results.