Timothée Demont : timothee.demont[at]univ-amu.fr
Roberta Ziparo : rziparo[at]gmail.com
How much of the distribution of economic activity today is determined by history rather than by geographic fundamentals? And if history matters, does it matter much? We develop an empirical framework that enables answers to these questions. Our model combines a workhorse model of trade subject to geographic frictions with features of local agglomeration externalities as well as an overlapping generations model of labor mobility also subject to spatial fractions. We derive parameter conditions, for arbitrary geographic scenarios, under which equilibrium transition paths are unique and yet steady states may nevertheless be non-unique — that is, where initial conditions (“history”) determine long-run steady-state outcomes (“path dependence”). We then estimate the model’s parameters (which govern the strength of agglomeration externalities and trade and migration frictions), by focusing on moment conditions that are robust to potential equilibrium multiplicity, using spatial variation across US counties from 1800 to the present. We then simulate a range of counterfactual scenarios that vary the initial conditions of US economic geography in order to shed light on the extent to which path dependence is costly — or equivalently, the extent to which the modern U.S. distribution of economic activity is inefficient because of the long arm of history.