Previous work on hedonic price functions tends to have focused on one of a number of specification and estimation issues; namely, market segmentation, choice of functional form, multicollinearity or spatial autocorrelation. The purpose of this paper is to bring together these various strands to provide a comprehensive modelling approach. In particular we use a combination of factor analysis and cluster analysis to define market segments and reduce collinearity in the data. We adopt Robinson's semiparametric specification of the hedonic price function and account for spatial autocorrelation using Kelejian and Prucha's generalized moments estimator. The modelling approach is applied to a large and extremely detailed dataset for the City of Birmingham constructed from multiple data sources and compiled with the use of GIS. The focus of this application is the identification of implicit prices for noise pollution from road, rail and air traffic sources.
clustering factor analysis Hedonics market segmentation partially linear model spatial autocorrelation