Céline ROZENBLAT, University of Lausanne, Switzerland
Urbanization is perhaps the most remarkable spatial process that affected human societies during the last two centuries. In the case of industrialized countries, urbanization is generally seen as the consequence of industrialization. Formal economic models show that industrialization results from technological evolution and consumer preferences (Desmet & Henderson, 2015), but generally overlook the role played by multilevel spatial factors in the processes of urbanization and industrialization.
Building on the substantial empirical evidence about the role played by the spatial factors that are agglomeration economies at the meso level of each city (Kim, 2006) and transportation costs at the macro level (Taylor, 1962; Herrendorf et al., 2009), we develop a spatial micro-to-meso-to-macro agent-based model of the American economy between 1800 and 1920. The model shows the necessity of the decrease of transportation costs of labor and of the presence of matching economies to the American urban transition. The model features heterogeneous boundedly rational (Simon, 1955) interacting economic agents: firms and workers-consumers. Space plays a structuring role in the model by limiting the firm-workers interactions to the local level and by imposing distance friction on migration and trade. To establish the dependence of the urban transition on spatial factors, we proceed in two steps. The model is first calibrated to empirical data: we use a spatial structure that roughly reproduces the evolution American transportation network and fit the model parameters to the observed urbanization rates and city size distributions.
We then explore what would have happened if the parameters controlling the decrease in transportation costs and the importance of labor matching economies had been different. The exploration shows that urbanization levels would significantly differ under higher transportation costs and a nonlocal labor matching process.
Keywords: urbanization|multi-level|agent-based model|United-States