Testing the conventional wisdom about land use and traffic congestion: The more we sprawl, the less we move?

The paper explores relationships between seven dimensions of land use in 1990 and subsequent levels of three traffic congestion outcomes in 2000 for a sample of 50 large US urban areas. Multiple regression models are developed to address several methodological concerns, including reverse causation and time-lags. Controlling for prior levels of congestion and changes in an urban area's transport network and relevant demographics, it is found that: density/continuity is positively related to subsequent roadway ADT/lane and delay per capita; housing centrality is positively related to subsequent delay per capita; and housing-job proximity is inversely related to subsequent commute time. Only the last result corresponds to the conventional wisdom that more compact metropolitan land use patterns reduce traffic congestion. These results prove two points: that the choice of congestion measure may substantively affect the results; and that multivariate statistical analyses are necessary to control for potentially confounding influences, such as population growth and investment in the transport network.