Sorting and the Skill Premium: The Role of Nonhomothetic Housing Demand

Since 1980 skilled and unskilled workers have increasingly sorted into different cities. In this paper we propose and quantify a novel driver of increased spatial sorting — diverging preferences over location attributes caused by diverging incomes across skill groups. The root cause of these diverging preferences is income inelastic housing demand, which we show to be a robust feature of consumption microdata. As the incomes of skilled workers rise relative to those of unskilled workers, their housing expenditure shares fall and they become more willing to cluster together in expensive, skill-intensive cities. By adding nonhomothetic preferences to an otherwise standard quantitative spatial model we show that this mechanism explains 22% of the increase in sorting since 1980.