Putting a price on your neighbour

• Published in: Journal of housing and the built environment Vol. 32; no. 1; pp. 157 - 175
• Main Authors: Nordvik, Viggo, Osland, Liv
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer 01-03-2017; Springer Netherlands; Springer Nature B.V
• Subjects: Composition effects; Economic conditions; Estimates; Geography; Houses; Housing; Housing prices; Human Geography; Hypotheses; Immigrants; Inhabitants; Landscape/Regional and Urban Planning; Mathematical models; Minority & ethnic groups; Multiculturalism & pluralism; Neighborhoods; Noncitizens; Population; Regression; Robustness; Social Sciences; Specifications; Studies; Urban areas; Willingness to pay; Home prices; Oslo; Urban sorting; Neighbourhood characteristics; Diversity
• ISSN: 1566-4910; 1573-7772
• DOI: 10.1007/s10901-016-9506-5

Neighbourhood population composition affects the willingness to pay for housing units. This paper utilises a large and rich data set and hedonic regression techniques to disentangle the effect that neighbourhood affluence and presence of inhabitants with an immigrant background have on home prices. Furthermore, we specify an empirical model in a way that also enable us to test for the effect of diversity, both in terms of income levels and of the composition of the immigrant population of a neighbourhood. The hedonic model can be viewed as a variety of an amenity interpretation of the population composition of a neighbourhood. Estimation of effects of population composition is not straightforward as there is good reason to believe that population composition is both endogenously determined together with house prices and that area level omitted variables could bias estimates. This is addressed by lagging the composition measures and by formulating two different models that address these difficulties in different ways. We estimated one random effects model that instruments within neighbourhood variation in population composition and one fixed effects model that control for omitted variables. We find that coefficient estimates are robust across these specifications.