SMU SOE Seminar (Nov 2, 2016): Optimal Transport Networks in Spatial Equilibrium

    We develop a framework to study the globally optimal transport network in general-equilibrium spatial models. We embed an optimal transport problem into a general neoclassical environment with arbitrary many locations arranged on a graph. Goods must be shipped through linked locations and resources can be invested into lowering trade costs in any link. We define the globally optimal transport network as the solution to the utilitarian planner's problem of simultaneously choosing the allocation, the gross trade flows across the network, and the capacity in every link. We approach the dimensionality and non-convexity challenges intrinsic to optimizing over the space of all networks by introducing congestion in transport. If congestion in transport is strong relative to the returns to network-capacity investments, the planner's is a standard convex optimization problem, guaranteeing convergence of efficient gradient-descent based algorithms. The framework nests the neoclassical multi-sector trade models commonly used in international trade and allows for factor mobility. We illustrate the uses of the framework through basic applications, and we also discuss how to approach cases with scale economies in transport, in which the convexity of the problem is not ensured.

Keywords: Optimal Transport Networks; General Equilibrium; Factor Mobility