Modelling the locational determinants of house prices: neural network and value tree approaches

Abstract

Tom Kauko's book comprises an analysis of the locational element in house prices. Locational features can increase or decrease the value of a house compared with a similar one elsewhere. So far, the problem of isolating this element has been well documented in the literatures on spatial housing market modelling and property value modelling. These lines of research usually use the economic equilibrium model as theoretical umbrella. Kauko's approach extends this conventional model towards involving problematic aspects such as multiple equilibria, institutions and diversified preferences. By doing so, Kauko argues that using one approach only is insufficient, and therefore he applies two different methods for the empirical part of the analysis. The first one is essentially a mass-appraisal approach based on neural network modelling that identifies segments, location, and omitted variables. The second one is a dis-aggregated approach based on multiattribute value tree modelling that encapsulates the behavioural element - perceptions, preferences, price/quality relationships, and agency effects. The results show the strengths and weaknesses of the new methods as tools for a variety of appraisal purposes