In new economic geography models, geographic concentration can't arise because of workers' mobility or vertical linkages between firms. We examine a setup that combines those two approaches in conjunction with local congestion costs. We find that, as trade costs are lowered, the geographic concentration of total activity (agglomeration) follows an inverse u-shaped evolution, while the degree of specialization of regions increases. These results shed light on regional development within a country as integration proceeds: when trade costs are high, firms evenly spread between the regions to supply local demand at low costs, hence diversified regions
at intermediate trade costs, we have coexistence of a diversified core and a specialized periphery and at low trade costs, each industry clusters in one region to fully exploit returns to scale externalities. US city centers and non-metropolitan areas during the period 1850-1990 depict such specialization and agglomeration patterns. These results show that a country's effort to improve accessibility across its portfolio of places can favor a win-win regional allocation of firms based on each location's competitive advantage.