Superlinear growth in cities has been explained as an emergent consequence of increased social interactions in dense urban environments. Using geocoded microdata from Swedish population registers, we remove population composition effects from the scaling relation of wage income to test how much of the previously reported superlinear scaling is truly attributable to increased social interconnectivity in cities. The Swedish data confirm the previously reported scaling relations on the aggregate level, but they provide better information on the micromechanisms responsible for them. We find that the standard interpretation of urban scaling is incomplete as social interactions only explain about half of the scaling parameter of wage income and that scaling relations substantively reflect differences in cities' sociodemographic composition. Those differences are generated by selective migration of highly productive individuals into larger cities. Big cities grow through their attraction of talent from their hinterlands and the already-privileged benefit disproportionally from urban agglomeration.