A Conceptual K-means Framework for Interpretable Multidimensional Housing-Market Segmentation

Prices, rents and affordability ratios are the usual starting point for any housing market analysis, but they seldom tell the whole story. Two cities may be equally unaffordable but may differ greatly in construction activity, vacancy, tenure structure, provision of public housing, access to mortgages and spatial connectivity to employment. In this paper, we develop a conceptual and methodological framework for the use of K-means clustering as a theory-guided tool for housing-market segmentation. The framework treats each housing market as: X_i=[D_i,S_i,A_i,P_i,E_i], where D_idenotes demand conditions, S_isupply capacity, A_iaffordability and market outcomes, P_ispatial accessibility, and E_isocioeconomic and tenure-system characteristics. In this paper K-means is not presented as the most sophisticated clustering algorithm, but it is valuable for the practical visibility of centroid profiles. Where the input variables are continuous, standardised and substantively meaningful in Euclidean space, centroids can assist researchers and planners to inspect whether a cluster corresponds to a recognisable housing market condition. Where categorical or ordinal variables are introduced in the analysis in the form of tenure regimes, rent-regulation categories or planning-system types, k-means should be compared with mixed-data alternatives such as k-prototypes and PAM with Gower distance. The proposed typology identifies constrained high-pressure markets, expensive but supply-active markets, broadly balanced markets, stagnant oversupplied areas, declining low-demand markets, and low-cost areas where affordability is undermined by poor access to jobs and services. The paper does not present empirical results, but explains how a European pilot could be implemented, using harmonised data at the city or functional urban area level. Its contribution is thus pre-empirical: it provides the conceptual and methodological architecture that a subsequent empirical study would have to test.