In this paper I propose a novel criterion for simultaneous GMM model and moment selection: the generalized focused information criterion (GFIC). Rather than attempting to identify the correct specification, the GFIC chooses from a set of potentially mis-specied overidentifying and parameter restrictions to minimize the mean-squared error of a target parameter specied by the user.

Both the focused moment selection criterion (FMSC) of DiTraglia (2012) and the focused information criterion (FIC) of Claeskens and Hjort (2003) can be viewed as special cases of the GFIC. In addition to the general theory, I specialize the GFIC to a simple dynamic panel model with correlated individual effects. In a simulation study based on this example, the GFIC performs well relative to alternatives from the literature.