We provide new insights on the city size distribution of countries around the world. Using more than 10,000 cities delineated via geospatial data and a globally consistent city identification scheme, we investigate distributional shapes in all countries. In terms of population, we find that Zipf's law holds for many, but not all, countries. Contrasting the distribution of population with the distribution of economic activity, measured by nighttime lights, across cities we shed light on the globally variant magnitude of agglomeration economies. Deviations from Zipf's law are to a large extent driven by an undue concentration in the largest cities. They benefit from agglomeration effects which seem to work through area rather than through density. Examining the crosscountry heterogeneity in the city size distribution, our model selection approach suggests that historical factors play an important role, in line with the time of development hypothesis.

No abstract

The emergence of cities in specific locations depends on both geographical features (such as elevation and proximity to rivers) and institutional factors (such as centrality within an administrative region). In this paper, we analyse the importance of these factors at different levels of the urban hierarchy. To do so, we exploit a unique data set on the location of cities of different status in imperial China from 221 BCE to 1911 CE, a geographically diverse empire with a long history of centralised rule. Developing a stylised theoretical model, we combine econometrics with machine learning techniques. Our results suggest that the higher a city is in the urban hierarchy, the less important are local geographical features compared to institutional factors. At the lower end of the scale, market towns without government responsibilities are most strongly shaped by geographical characteristics. We also find evidence that many cities of political importance in imperial times still enjoy a special status nowadays, underlining the modern relevance of these historical factors.