I'm wondering why 'restrict' isn't possible after GMM. Performing Wald-type tests should not pose extra problems once we have the estimator and the covariance matrix, right? Or am I missing something? thanks, sven