Detecting and validating residential housing submarkets

Abstract

PurposeThe purpose of this paper is to describe a segmentation technique based on geostatistical modeling methods utilizing geographically weighted regression (GWR) to identify submarkets which could be applied within the mass appraisal environment.Design/methodology/approachGiven the spatial dimension within which neighbourhoods/submarkets exist, this paper has sought to utilize the geostatistical technique of GWR to identify them.FindingsThe efficacy of the procedure is established by demonstrating improvements in predictive accuracy of the resultant segmented market models as compared to a baseline global unsegmented model for each of the study areas. Optimal number of segments is obtained by measures of predictive accuracy, spatial autocorrelation in the residual errors and the Akaike information criterion.Research limitations/implicationsThe three datasets used allowed for an evaluation of the robustness of the method. Nonetheless it would be beneficial to test it on other datasets, particularly from different regions of the world.Practical implicationsMany researchers and mass appraisal practitioners have established the benefit of segmenting a study area into two or more submarkets as a means of incorporating the effects of location within mass valuation models. This approach develops the existing knowledge.Social implicationsThe research ultimately is developing more accurate valuation models upon which the property tax is based. This should create an environment of fair and acceptable assessed values by the tax paying community.Originality/valueThe contribution of this work lies in the methodological approach adopted which incorporates a market basket approach developed through a process of GWR. The importance of the research findings illustrate that submarket segmentation need no longer be an arbitrary process.