IOSPress

Various location specific attributes contribute to the spatial dynamics of housing markets. This effect may partly be of a qualitative and discontinuous nature, which causes market segmentation into submarkets. The question however is, whether the most relevant partitioning criteria is directly related to the transaction price of to other, socioeconomic, demographic and physical features of the location. Two neural network techniques are used for analysing statistical house price data from Amsterdam and Helsinki. The analytic hierarchy process is used as a supporting technique. With these techniques it is possible to analyse various dimensions of housing submarket formation. The findings show that, while the price and demand factors have increased in importance, supply factors still prevail as key criteria in both cases. The outcome also indicates that the housing market structure of Amsterdam is more fragmented than that of Helsinki, and that the main discriminating housing market features, and the ways they have changed in time, are somewhat different.