This article contrasts traditional modeling approaches and discrete-choice models as methods to analyze locational attainment—how individual and household characteristics (such as race, socioeconomic status, age) influence the characteristics of neighborhoods of residence (such as racial composition and median income). Traditional models analyze attributes of a neighborhood as a function of the characteristics of the households within them; discrete-choice methods, on the other hand, are based on dyadic analysis of neighborhood attributes and household characteristics. I outline two problems with traditional approaches to residential mobility analysis that may be addressed through discrete-choice analysis. I also discuss disadvantages of the discrete-choice approach. Finally, I use data from the Panel Study of Income Dynamics to estimate residential mobility using traditional locational attainment and discrete-choice models; I show that these produce similar estimates but that the discrete-choice approach allows for estimates that examine how multiple place characteristics simultaneously guide migration. Substantively, these models reveal that the disproportionate migration of black households into lower-income tracts amounts to sorting of black households into black tracts, which on average are lower income.
Linear IV Regression Estimators for Structural Dynamic Discrete Choice Models
