Commuting Costs and Urban Sprawl: Which Proxy Measures Up?

Empirical studies investigating urban sprawl and the determinants of city size with the Mills-Muth framework have struggled to find a reliable and ubiquitous proxy for the theoretical commuting costs variable. This study is the first to apply the Davidson-McKinnon non-nested specification test to address the long-standing issue in the literature of determining the best proxy measure for commuting costs. We employ this specification test to evaluate the three most widely available commuting costs measures from the literature: vehicle availability, public transit usage, and commuting speed. For a sample of all urbanized areas in 2000 and 2010, our results provide a degree of resolution. While we find for a pooled sample of all urbanized areas that commuting speed is the preferred proxy, subsample analysis reveals the prior result may be driven by larger urbanized areas spanning more than one county; commuting speed dominates the other proxies for these larger cities. Conversely, the sizes of single-county urbanized areas are explained by vehicle availability and transit usage, though neither of those proxy measures emerge as dominant, suggesting some unspecified measure may be better for these smaller cities.