Agglomerations in a Multi-region Economy: Polycentric versus monocentric patterns

Agglomeration externalities have been recognized as major sources of lumpy spatial distributions of industries and population. While the abstraction of interregional space has been a common exercise, the recent increasing availability of disaggregated geographical data and more sophisticated computational techniques have promoted counterfactual analyses based on many-region models of agglomeration externalities with explicit interregional space (e.g., Redding and Sturm, 2008; Allen and Arkolakis, 2014). A caveat is that incorporating interregional space to a many-region model with agglomeration externalities by itself does not warrant the formation of polycentric agglomerations in stable equilibria—a crucial property in order to replicate the observed geography of agglomerations. We elaborate this point by comparing a pair of new economic geography models: Forslid and Ottaviano (2003) and Helpman (1998). In a two-region economy, these models exhibit both "agglomeration" (i.e., a relative concentration of mobile agents in one of the regions) and "dispersion" (i.e., a uniform distribution of mobile agents across the two regions). But, if the location space were more disaggregated, only the former admits polycentric agglomerations in stable equilibria, while in the latter, only a monocentric agglomeration can occur if any.