Wasteful commuting: A re-examination

In a provocative article, Hamilton [l] demonstrates that the simple monocentric model of urban location, modified to allow for decentralized employment, does a poor job of predicting actual commuting distances in major U.S. cities. According to the monocentric model, households locate to maximize the utility received from housing and all other goods (= income commuting costs housing costs). If households are otherwise indifferent among housing locations, and if there is one worker per household, utility-maximizing location choices minimize aggregate commuting distances, given the location of houses and jobs. The fact that average actual commutes are about 8 times the average minimum commute casts doubt on the validity of the monocentric model. Hamilton’s findings suggest that one should modify the monocentric model to incorporate other determinants of location choice. In the discrete housing-choice literature, for example, households receive utility from neighborhood amenities, as well as from the commuting distances of all workers in the household. When utility is defined in this way, utilitymaximizing location choices need not minimize aggregate commuting distances; thus, the more general location-choice model may explain the divergence between average-actual and average-minimum commutes. In this paper we examine by how much a broader definition of utility of residential location raises the average required commute. Specifically, we (1) estimate a utility function defined over housing and neighborhood attributes as well as the commuting distances of all workers in the