The second half of the twentieth century saw large-scale suburbanization in the United States, with the median share of residents who work in the same county where they live falling from 87 to 71 percent between 1970 and 2000. We introduce a new methodology for discriminating between the three leading explanations for this suburbanization (workplace attractiveness, residence attractiveness and bilateral commuting frictions). This methodology holds in the class of spatial models that are characterized by a structural gravity equation for commuting. We show that the increased openness of counties to commuting is mainly explained by reductions in bilateral commuting frictions, consistent with the expansion of the interstate highway network and the falling real cost of car ownership. We find that changes in workplace attractiveness and residence attractiveness are more important in explaining the observed shift in employment by workplace and employment by residence towards lower densities over time.