Research Seminar: Laurent Gobillon (PSE)

Abstract

We use a combination of image processing and machine learning tools to extract data from historical maps of France. Granular information about the distribution of population allows us to delineate cities consistently (and relatively) over 260 years. Against a backdrop of rising population, the rate of urbanization increased but the number of cities declined and the size distribution of cities also became increasingly skewed. Within cities, population density rose and its gradient flattened. To bring these facts together, we propose a model of urban divergence for which we verify further empirical predictions.

Local Organizer: Niko Jaakkola