Household Sorting and Neighborhood Formation

We develop a new model of household sorting in a system of residential neighborhoods. We show that this model is partially identified without imposing parametric restrictions on the distribution of unobserved tastes for neighborhood quality and the shape of the indirect utility function. The proof of identification is constructive and can be used to derive a new semiparameteric estimator. Our empirical application focuses on residential choices and housing demand in a system of neighborhoods in the Pittsburgh metropolitan area. We find that there are significant differences in the observed sorting of households with and without children. In particular, households with children exhibit more stratification by income than households without children.