A two-stage multinomial logit selection model is used to model the relationship between demographic characteristics and housing density across Tennessee's six metropolitan statistical areas. The study finds that there is both spatial correlation and heterogeneity in the spatial distribution of housing both within and across the six areas. For example, Memphis, the most densely populated area, has the least amount of spatial correlation among housing density at the neighborhood level, while Johnson City, which has the lowest overall housing density, has the highest degree of spatial correlation.