The basic neoclassical model of migration suggests that migration is induced by real income differentials across locations and will, ceteris paribus, serve to reduce those differentials. And yet there is evidence on growing spatial inequality despite increased migration from poorer to richer areas. At a theoretical level, one route to addressing this potential inconsistency is to introduce agglomeration effects into the standard neoclassical setup. This paper explores an alternative route, based on a theoretical and empirical proposition of the migration literature, namely, that migration is a selective process. Focusing on skilled migration, the paper demonstrates the different forces in play that make selective migration a force for both divergence and convergence, and characterizes where each set of forces dominates. Finally, it explores the consequences for convergence of combining both migration selectivity and agglomeration effects arising from migrant networks.