The science of cities seeks to understand and explain regularities observed in the world's major urban systems. Modelling the population evolution of cities is at the core of this science and of all urban studies. Quantitatively, the most fundamental problem is to understand the hierarchical organization of city population and the statistical occurrence of megacities. This was first thought to be described by a universal principle known as Zipf's law(1,2); however, the validity of this model has been challenged by recent empirical studies(3,4). A theoretical model must also be able to explain the relatively frequent rises and falls of cities and civilizations(5), but despite many attempts(6-10) these fundamental questions have not yet been satisfactorily answered. Here we introduce a stochastic equation for modelling population growth in cities, constructed from an empirical analysis of recent datasets (for Canada, France, the UK and the USA). This model reveals how rare, but large, interurban migratory shocks dominate city growth. This equation predicts a complex shape for the distribution of city populations and shows that, owing to finite-time effects, Zipf's law does not hold in general, implying a more complex organization of cities. It also predicts the existence of multiple temporal variations in the city hierarchy, in agreement with observations(5). Our result underlines the importance of rare events in the evolution of complex systems(11) and, at a more practical level, in urban planning. A theoretical model in the form of a stochastic differential equation is proposed that describes, more accurately than previous models, the population evolution of cities, revealing that rare but very large interurban migration is a dominant factor.