Spatial issues on a hedonic estimation of rents in Brussels

Using Belgian microdata, we assess the impact, on a hedonic regression, of the distortions arising from the choice of either a specific zoning system or the delineation of the study area. We also evaluate the biases that arise when spatial effects are not accounted for. Given that the dependent variable is interval-coded, controlling for spatial dependence in this context is challenging. We address this problem with two alternative strategies. Firstly, we use the Gibbs Sampling algorithm to estimate spatial econometric models which extends the interval regression model. A major drawback of this approach is that the implied estimation is proned to the endogeneity biases inherent to our hedonic regression model. To circumvent the endogeneity issues triggered by the first estimation strategy, we also use a two-stage estimation procedure with locational fixed effects. In all specifications, results are sensitive to the Modifiable Areal Unit Problem (MAUP) and to the choice of the delineation of the study area. Moreover, they confirm the existence of substantive spatial dependence. Conversely to the previous results with a negative elasticity for the percentage of the area covered by agriculture and a positive elasticity for the potential accessibility to jobs, the second approach implies opposite effects for those two variables. This indicates that dwellings close to agricultural areas and with a lower accessibility to the main employment centers are highly demanded and that endogeneity biases are not negligible.(This abstract was borrowed from another version of this item.)