AbstractShould China build mega-cities or a network of linked middle-sized metropolises? Can Europes mid-sized cities compete with global agglomeration by forging stronger inter-urban links? This paper examines these questions within a model of recombinant growth and endogenous local amenities. Three primary factors determine the trade-off between networks and big cities: local returns to scale in innovation, the elasticity of housing supply, and the importance of local amenities. Even if there are global increasing returns, the returns to local scale in innovation may be decreasing, and that makes networks more appealing than mega-cities. Inelastic housing supply makes it harder to supply more space in dense con
fines, which perhaps explains why networks are more popular in regulated Europe than in the American Sunbelt. Larger cities can dominate networks because of amenities, as long as the benefits of scale overwhelm the downsides of density. In our framework, the skilled are more likely to prefer mega-cities than the less skilled, and the long-run benefits of either mega-cities or networks may be quite different from the short-run benefits.