Most economic activity occurs in cities. This creates a tension between local increasing returns, implied by the existence of cities, and aggregate constant returns, implied by balanced growth. To address this tension, we develop a general equilibrium theory of economic growth in an urban environment. In our theory, variation in the urban structure through the growth, birth, and death of cities is the margin that eliminates local increasing returns to yield constant returns to scale in the aggregate. We show that, consistent with the data, the theory produces a city size distribution that is well approximated by Zipf's Law, but that also displays the observed systematic under-representation of both very small and very large cities. Using our model, we show that the dispersion of city sizes is consistent with the dispersion of productivity shocks found in the data.