City Structure and Congestion Costs

This paper presents a model and an automated methodology for decomposing congestion costs in cities. We model a city as a directed graph and define the planner’s problem of finding the subgraph minimizing the congestion costs (latency) of endogenously-routing traffic flowing through the subgraph. We show that the minimal total latency subnetwork displays congestion directly proportional to city population. By applying an automated search algorithm on a widely available Internet mapping application, and hence sampling from the empirically implemented subgraphs, the paper estimates the congestion-minimizing distortions of city transit networks in a large sample of United States and Italian cities.