Discrete choice models of housing demand

In this study, discrete choice models of housing demand are estimated from the 1980 Public Use Microdata Samples (PUMS) for Chicago, Pittsburgh, Houston and San Diego. Specifically, the study focuses on the estimation of multinomial logit and nested logit models. The choices in question are choice of specific housing unit, choice of a housing submarket and choice of tenure.;Housing preferences are explained by the disposable income, the number of bedrooms and indices of physical housing quality based on a point system constructed and tested in this dissertation. An innovation in this study is a method for reducing the need for market segmentation--the grouping of households by socioeconomic characteristics into segments for which different utility functions are postulated. Instead of grouping households into segments, some coefficients of the utility function are made a function of the socioeconomic characteristics of the household. We call this the master utility function. This approach gives reasonable own-rice elasticities of housing demand and marginal rates of substitution (MRS). Nonwhite households have lower elasticities keeping income constant, but higher elasticities overall. The MRS between rooms and disposable income increases with income. Changes in model specification modify the level but not the pattern in these elasticities and MRS.;The validity of the master utility function for various market segments is tested by segmenting households by race-of-head, income, age-of-head and family type. When households are segmented by these characteristics, then segment-specific models are comparable to pooled models but some of segment-specific models require respecification.;The hypothesis that housing choice is nested within submarket choice is accepted. This two-level nested logit model gives much better fit and more reasonable MRS between disposable income and rooms. At the same time, the own-price elasticity is increased somewhat. We also test the hypothesis that submarket choice and the choice of dwelling are both nested within the tenure choice decision. Although such a nesting has a lot of intuitive appeal, it does not appear that these models are better than the two-level nested models.