It does seem it is skill-based technical change, which rewards high-talent urban clusters, and in the broader equilibrium, also lowers geographic mobility. That is the theme of the job market paper of Elisa Giannone from the University of Chicago, here is the abstract:
“Skilled-Biased Technical Change and Regional Convergence,” (Job Market Paper)
Poorer US cities were catching up with richer ones at an annual rate of roughly 1.4% between 1940 and 1980. However, wage convergence across US cities went from 1.4% a year between 1940 and 1980 to 0% a year between 1980 and 2010. This paper quantifies the contributions of skill-biased technical change (SBTC) and agglomeration economies to the end of cross-cities wage convergence within the US between 1980 and 2010. I develop and estimate a dynamic spatial equilibrium model that looks at the causes of the decline in spatial wage convergence. The model choice is motivated by novel empirical regularities regarding the evolution of the skill premium and migration patterns over time and across space. The model successfully matches the quantitative features of the decline in US regional wage convergence, as well as other stylized facts on US economic growth. Moreover, the model also reproduces the convergence and the divergence in the skill ratio across US cities and other features on quantities, such as the secular decline in within US migration after 1980. Finally, the counterfactual analysis suggests that SBTC explains the approximately the 80% of the decline of regional convergence between 1980 and 2010 among high skill workers.
Here are her other papers.