On Discrete Choice Models of Housing Demand

Abstract This paper explores several specifications of discrete choice models in estimating housing consumption for purposes of projection and policy simulation. We view housing consumption comprehensively as it includes the choice between three types of homeownership, five types of rented housing, and, as a ninth alternative, not heading a household. The familiar multinomial logit specification is rejected by several specification tests. The paper therefore studies a variety of nested multinomial logit models of hierarchical choice that allow a richer substitution pattern among the very disparate housing choices enumerated above. Estimations of different nesting structures reveal strong and plausible differences among nesting structures.