We consider identification in a "generalized regression model" (Han, 1987) for panel settings in which each observation … unobservables, and arbitrary heteroskedasticity. We provide sufficient conditions for full identification of the model, as well as … weaker conditions sufficient for identification of the latent group effects and the distribution of outcomes conditional on …

In parametric models a sufficient condition for local identification is that the vector of moment conditions is … the true value that are sufficient for local identification. We apply these results to obtain new, primitive … identification conditions in several important models, including nonseparable quantile instrumental variable (IV) models, single …

In parametric models a sufficient condition for local identification is that the vector of moment conditions is … imply local identification. It turns out these conditions are slightly stronger than needed and are hard to check, so we …