1214 Spatial Hedonic Models
26.6 Empirical evidence: spatial heterogeneity 1239
26.6.1 Discrete heterogeneity: sub-markets 1239
26.6.2 Continuous spatial heterogeneity: spatially varying coefficients 1241
26.7 Concluding remarks 1243
26.1 Introduction
Hedonic pricing models have become a common tool in applied microeconomics,
going back to the classic contributions of Lancaster (1966) and Rosen (1974) (for a
recent review, see, e.g., Malpezzi, 2002). The ability of hedonic models to relate the
price of a product to the relative contributions of different characteristics has led to
a wide range of applied econometric work using these specifications. An important
class of applications pertains to house price models, in which characteristics of the
property, the neighborhood and other amenities are included in an econometric
specification for the sales price or assessed value of a housing unit. Such models
are now routinely used in mass appraisal exercises as well as in the valuation of
non-market amenities that contribute to the price of the house. The rationale for
the latter is that, in an efficient market, superior amenities (such as clean air, access
to parks or beaches, and views) should be capitalized into the value of the house. In
other words,ceteris paribus, houses with superior amenities should be more expen-
sive and the price differential constitutes a measure for the value of the amenity as
expressed through market transactions.
In this chapter, we focus on some econometric aspects related to a sub-set of
hedonic house price models, which we refer to asspatial hedonic models. In these,
the locational aspects of the observations are treated explicitly, as an application of
spatial econometrics. As defined in Anselin (2006), spatial econometrics “consists of
a sub-set of econometric methods that is concerned with spatial aspects present in
cross-sectional and space-time observations.” These methods focus, in particular,
on two forms of so-calledspatial effectsin econometric models, referred to asspatial
dependenceandspatial heterogeneity(Anselin, 1988).
As outlined in Anselin, spatial dependence or spatial autocorrelation is a spe-
cial case of cross-sectional dependence in which the structure of the covariation
between observations at different locations is subject to a spatial ordering. This
ordering is related to the relative positioning, distance or spatial arrangement of
the observations in geographic space, or, more generally, in (social) network space.
This type of dependence differs from time series dependence in that it is both
two-dimensional as well as multidirectional. This implies a simultaneous feedback
between observations (“I am my neighbor’s neighbor”), which requires the appli-
cation of specialized techniques that are not simply an extension of time series
methods to two dimensions.
Spatial heterogeneity is a special instance of structural instability, which can be
observed or unobserved. The spatial aspect of this issue is that spatial structure pro-
vides the basis for the specification of the heterogeneity. This may inform models
for spatial structural change (referred to as spatial regimes), heteroskedasticity, or
spatially varying and random coefficients.