The normal-gamma stochastic forntier model was proposed in Greene and Beckers and Hammond as an extension of the normal-exponential proposed in the original derivations of the stochastic frontier by Aigner, Lovell, and Schmidt. The normal-gamma model has the virtue of providing a richer and more...

This paper considers GMM estimation of autoregressive processes. It is shown that, contrary to the case where the noise is independent, using high-order moments can provide subtantial efficiency gains for estimating the AR model when the noise is only uncorrelated.