The spatial interactions among the neighborhoods of Dijon Métropole,France: A comparative analysis

Abstract

This paper makes a methodological contribution to the field of hierarchical-spatial models applied to housing prices by addressing the nature of spatial interaction at higher levels of analysis. While spatial interactions at the dwelling level are often straightforward to model, spatial processes at the neighborhood level are more abstract, serving as proxies for latent constructs such as neighborhood quality. The present study builds on the distinction between endogenous and exogenous interaction effects, as well as global and local spillovers, to compare three upper-level interaction structures: endogenous (HSAR), exogenous (HSLX), and mixed (HSDM). A Bayesian estimation strategy is employed in conjunction with the Deviance Information Criterion (DIC) to evaluate these specifications using 6566 housing transactions in Dijon Métropole, France, between 2013 and 2017. The findings of this study suggest that models incorporating endogenous upper-level dependence provide a superior fit to the data in comparison to the purely exogenous alternative. However, the distinction between global and mixed spillovers necessitates a meticulous evaluation. While the complete DIC provides slight support for the endogenous model, the conditional DIC and the lack of statistical significance for the autoregressive coefficient in the mixed specification provide stronger evidence for the HSAR model as the more parsimonious and appropriate representation of spatial dependence.