This paper explores Rosen’s (1974) suggestion that within the hedonic framework there are natural tendencies toward market segmentation. We show that market segmentation can be estimated on the basis of an augmented hedonic model in which marginal prices are separated by household characteristics into different classes. The classes can either be exogenously defined or endogenously determined based on an unsupervised machine learning algorithm or a latent class formulation. We illustrate the usefulness of these methods using American Housing Survey data for Louisville and show that there are distinct housing market segments within the Louisville metropolitan area based on income and family structure.
Droes, Martijn, Martin Hoesli, and Steven C. Bourassa. "Heterogeneous Households and Market Segmentation in a Hedonic Framework." In 26th Annual European Real Estate Society Conference. ERES: Conference. Cergy-Pontoise, France, 2019.
Keywords: Market Segmentation, Machine Learning, latent class, heterogeneous households and Hedonic Model
Section: Refereed Section: Workshop RE Economics