Choices or Constraints? Disentangling Preferences from Restrictions in Housing Markets

Abstract

How much of the observed house price differentials and household sorting is due to restrictions in choice sets, arising from implicit and explicit forms of discrimination, versus household preferences? To answer this question, we build a two-sided housing matching model that combines admission rules with traditional discrete choice models, allowing for heterogeneous choice sets across demographic groups. Additionally, we construct a novel dataset that links households and real-estate developments to the historic street grid of the 1940 Minneapolis metro area. Using an instrumental variable approach, we find that explicit discriminatory restrictions in neighborhoods reduce the likelihood of non-White and White Eastern and Southern European immigrant households by 10 and 2.8 percentage points, respectively. We also show that traditional discrete-choice models of neighborhood demand with misspecified choice sets can yield biased estimates of preferences for neighborhood characteristics, especially when there is selection in the characteristics of neighborhoods restricted to a subset of residents. Our two-sided housing matching model jointly estimates heterogeneous choice sets and preference parameters, revealing that different demographic groups faced substantially different choice sets. On average, minority households are 8 percentage points less likely to have a neighborhood in their choice set than non-minority households, even after controlling for explicit forms of discrimination and rents. Simulating a counterfactual scenario without these restrictions and preferences for co-patrons, we find that household preferences account for three-quarters of the observed segregation in 1940.