Skilled workers’ incomes have pulled away from those of unskilled workers in recent decades, reflecting increasing skill bias in production. How has this changed the spatial distribution of skill? We show nonhomothetic housing demand connects aggregate income inequality to spatial sorting. A household’s skill level determines its income, and therefore its housing expenditure share, sensitivity to housing costs, and location preferences. The result is spatial sorting by skill. Moreover, diverging incomes cause diverging location choices. Using consumption microdata, we estimate that housing is a necessity. Increasing total expenditure by 10% reduces housing expenditure shares by 2.5%. Skilled workers therefore sort into expensive cities, and by raising their relative incomes, increases in aggregate skill bias intensify sorting. Embedding our estimated preferences in a quantitative spatial model, we find that without rising aggregate skill bias, spatial sorting would have grown one quarter less since 1980.