CONTEXTUAL MODELS OF URBAN HOUSE PRICES - A COMPARISON OF FIXED-COEFFICIENT AND RANDOM-COEFFICIENT MODELS DEVELOPED BY EXPANSION

Contextual quantitative models are receiving considerable attention in the geographic literature in the form of models developed by the ''expansion method.'' The majority of these models have involved the expansion of the fixed part of the model and are in many cases equivalent to standard ANOVA and ANCOVA procedures. Models from a ''multilevel perspective,'' in which the expansion takes place in the random part of the model, have also been developed. Although these two methods converge conceptually, major differences in the form of the models can be specified and estimated. This paper develops contextual models of house prices using both approaches. We argue that varying-parameter multilevel models, which recognize that houses are nested within districts, are preferable to the usual single-level, fixed-coefficient models. We illustrate both approaches with data on house prices in London.