Hedonic price model: defining neighbourhoods with Thiessen polygons

Purpose - The purpose of the paper is to verify whether the version of neighbourhoods created from the lowest geographical level improve a predictive accuracy of hedonic model in comparison with those based on upper geographical levels. Design/methodology/approach - The paper proposes a method for defining neighbourhoods using Thiessen polygons. The clustering technique is based on fuzzy equality. Clustering is started at different geographical levels: municipalities, traffic analysis zones, and apartment blocks' Thiessen polygons. Delineated neighbourhoods are incorporated into hedonic model of apartment prices, the applied methodologies are ordinary least squares and spatial error. Findings - With ordinary least squares regression, the slight superiority of Thiessen polygons is found in both in‐sample analysis and ex‐sample prediction. With spatial error technique, the clusters of Thiessen polygons do not always provide the best outcome, and their superiority is contested by the highest geographical level of municipalities. Research limitations/implications - This paper is the first attempt to apply the proposed method, which not always demonstrates clear superiority. In future study, the method of neighbourhood delineation could be used in combination with market segmentation. Practical implications - The proposal to use Thiessen polygons as a transition from points to continuous space can outline a base for the use of different clustering techniques, which are applicable to delineate neighbourhoods in housing market studies, in particular for the assessment purpose. The fuzzy equality clustering algorithm itself can be applied to polygonal data. Originality/value - The originality of the proposed method is that it defines neighbourhoods starting from individual observations applying fuzzy equality. Its advantages are an increased independence from existing boundaries, self‐determination of a number of clusters, and total coverage of an area.