The ``Wrong Skewness'' Problem : Moment Constrained Maximum Likelihood Estimation of the Stochastic Frontier Model

It is well known that when the empirical skewness of the OLS residuals display the opposite sign of what is expected from the stochastic frontier model, maximum likelihood estimation will be equivalent to OLS and no inefficiency will be recovered. A variety of approaches to operate in this environment have been proposed, typically involving some type of respecification. Here we propose imposition of theoretically consistent moment conditions as constraints while engaging in maximum likelihood estimation. When the empirical skewness is incorrect, these moment conditions are unlikely to hold. Monte Carlo simulations show that our constrained MLE approach indeed alleviates the wrong skewness problem