Abstract
<jats:p>In 1950, highly specialized U.S. regions had higher per capita incomes than those with greater industrial diversity. Since then, however, the more specialized regions have grown persistently slower. I rationalize this novel finding in a dynamic multi-industry model featuring two opposing forces. On the one hand, specialization raises productivity via agglomeration economies. On the other hand, it increases exposure to sectoral shocks. Real factor adjustment costs and financial frictions make reallocation in response to shocks costly and long-lasting. Disciplined by U.S. Census data, the model explains half of the observed relationship between initial specialization and subsequent growth, with financial frictions accounting for more than half of this adverse effect. A constrained-efficient planner allocation reveals that less specialization can raise welfare by reducing a region’s exposure to industry-specific downturns.</jats:p>