The main purpose of this paper is to estimate panel data models with endogenous regressors and nonadditive unobserved individual heterogeneity including, for example, linear and nonlinear models where all the parameters can vary across individuals. The quantities of interest are means,...