Neighborhood weight matrix in a spatial-quantile real estate modeling environment: Evidence from Brazil

This paper analysis two central influences on real estate urban market: (a) neighborhood as cognitively-perceived identity units, and (b) heterogeneity of preferences of households. In doing so, two methodological changes are applied. Firstly, a neighborhood spatial matrix is proposed and compared to regularly used matrices and, secondly, a spatial-quantile regression is tested. The results highlight the fact that spatial influence brought to regression models through weight matrix should be carefully used, and that the matrix's choice might carry unobserved and unwanted effects into the estimation. Results also demonstrated that using information of neighborhood identity can optimize the understanding of city's complex influence on real estate markets. Finally, the quantile estimation should always be tested against in real estate estimation, as preferences of households seem to differ significantly for different levels of prices.