According to hedonic price theory (ROSEN 1974), real estate is valued for its utility-bearing characteristics. Because a property is fixed in space, a household implicitly chooses many different goods and services by selecting a specific object. From the methodological point of view, this can be explained with the hedonic price function f, describing the functional relationship between the real estate price P and object characteristics X_(O1),...,X_(On) as well as neighbourhood characteristics X_(N1),...,X_(Nm). Traditional approaches use a log-linear model structure (with the price and some of the continuous covariates logarithmically transformed), which reduces heteroscedasticity and nonlinearity. Nevertheless, locally varying equilibria or "submarkets" can be expected. If not accounted for, this leads to biased results and falsely induced spatial autocorrelation. Therefore, the literature provides a variety of local and global models (e.g. ANSELIN 1988, LESAGE & PACE 2009). One cutting-edge methodology that explicitly models heterogeneity is the geographically weighted regression (GWR, FOTHERINGHAM et al. 2002). Numerous applications (e.g. Yu et al. 2007) show the usefulness of this technique, as discussed later on. The main purpose of this research is therefore to define a hedonic pricing model that explains transaction prices for family dwellings in Austria accurately, taking into account structural and locational differences as well as spatial heterogeneity in intercept and slope parameters.
展开▼
