The size distribution of cities in many countries follows some broadly regular patterns. Any good theory of city size distributions should (i) be able to account for this regularity, but also (ii) rely on a plausible economic mechanism and (iii) be consistent with other fundamental features of cities like the existence of agglomeration economies and crowding costs. Unlike the previous literature, the model proposed here satisfies these three requirements. It views small innovation-driven techno logical shocks as the main engine behind the growth and decline of cities. Cities grow or decline as they win or lose industries following new innovations. Formally, this is achieved by embedding the quality-ladder model of growth developed by Grossman and Helpman in an urban framework.