Neighborhood Effects and Neighborhood Choice: Testing Necessary Conditions∗

This paper contributes to the social interactions literature by uti-lizing data on households ’ residential sorting. We employ a standard neighborhood effects model in which households ’ valuation of neigh-borhoods derives from neighborhood effects in the formation of human capital and in the process of enculturation. We use micro data from the PSID merged, using geocodes, with contextual information at the levels of census tracts from the 2000 US Census. We find support for a key implication of neighborhood effects models: households with chil-dren are more likely to move out of neighborhoods whose attributes are not favorable to the production of human capital and the transmis-sion of parents ’ cultural traits, and to move into neighborhoods with desirable such attributes. This effect is absent for households without children. ∗We are grateful to Yelena Ogneva-Himmelberger and the staff of the Tufts GIS Lab for priceless assistance with GIS work, and to Bob Haveman for help in acquiring access to the sensitive geocoded data of the Panel Study of Income Dynamics. Nate Baum-Snow