How does an increase in the size of the market due to fertility, immigration, or trade integration, affect welfare and GDP? We study this question using a model with heterogeneous firms, fixed costs, and monopolistic competition. We show how the shape of residual demand curves endogenously determines the variation of markups with firm size and the inefficiency of the decentralized equilibrium. We decompose changes in welfare from increased scale into changes in technical efficiency and changes in allocative efficiency due to reallocation. We non-parametrically identify residual demand curves with firm-level data and, using these estimates, quantify our theoretical results. We find that somewhere between 70% to 90% of the aggregate returns to scale are due to changes in allocative efficiency. In bigger markets, competition endogenously toughens and triggers Darwinian reallocations: big firms expand, small firms shrink and exit, and new firms enter. However, important as they are, the improvements in allocative efficiency are not primarily driven by oft-emphasized reductions in markups or deaths of unproductive firms. Instead, they are mostly caused by a composition effect that reallocates resources from low-markup to high-markup firms. Our analysis implies that the aggregate return to scale is an endogenous outcome shaped by frictions and market structure and likely varies with time, place, and policy. Furthermore, even mild increasing returns to scale at the micro level can give rise to large increasing returns to scale at the macro level.