PITFALLS OF NORMAL-GAMMA STOCHASTIC FRONTIER MODELS

Although conceptually pleasing, normal-gamma frontier models lead to d ifficult estimation problems. It is shown here that unless the sample size reaches several thousands of observations the shape parameter of the gamma density is hard to estimate, and that this carries over to e stimates of the stochastic frontier, the individual inefficiencies, an d the allocation of the overall variance to the stochastic frontier an d to the inefficiencies.