Web Appendix for Task Specialization in U.S. Cities from 1880-2000: Not for Publication

In this section, we outline a theoretical model that we use to rationalize the empirical results reported below. First, the model explains a secular reallocation of employment over time towards more interactive occupations in terms of differences in productivity growth across occupations and inelastic demand. Second, the model predicts that this reallocation is stronger in metro areas than in non-metro areas because of falling trade costs and specialization according to comparative advantage. To the extent that densely-populated locations are relatively more productive in interactive tasks (because agglomeration forces are stronger for interactive tasks), reductions in task trade costs induce densely-populated locations to specialize in more-interactive occupations, while more sparsely-populated locations specialize in less-interactive occupations. Third, and for the same reason, the model predicts that the reallocation towards more interactive occupations is stronger in denser metro areas. Fourth, the model explains how this allocation of tasks across areas that differ in their population density could have taken place both between sectors and within sectors. In this respect our model is more general than existing models that focus on differential changes across sectors (e.g. manufacturing vs. services) or within sectors (e.g. relocation of headquarters). Fifth, the model accounts for our finding that in 1880, when trade in tasks was still costly, metro areas were not more interactive than non-metro areas. Finally, the model ∗CEP, LSE, Houghton Street, WC2A 2AE, London, UK. Email: [email protected]. Tel: +44(0)20-7852-3518. †Department of Economics, University of Oxford, Manor Road, Oxford, OX1 3UQ. Email: [email protected]. Tel: +44(0)77214-88037. ‡Fisher Hall, Princeton, NJ 08540. Email: [email protected]. Tel: +1(609) 258-4016.