We use a large repeated cross-section of houses to estimate a selection model of the supply of owner-occupied and rental housing. We find that physical characteristics and unobserved heterogeneity and not location are important for selection. We interpret this as strong evidence in favor of...