Valuing the built environment : a GIS approach to the hedonic modelling of housing markets.
The valuation of the built environment has been a traditional concern of geographers. A
particular interest has been the way in which the value of locational externalities are
incorporated into house prices through housing market dynamics. However, much of the
previous research into this process has been of North American origin, despite the fact that
house prices, and property valuations in general, have become a major part of British life.
This research aims to begin to rectify this shortfall by studying the spatial dynamics of the
Cardiff Housing Market. Implicit in this research is an attempt to move towards a valuation
of locational externalities at the micro-scale.
The research employs two distinct method of analysis. Firstly, ARC / INFO GIS is used to
construct a context-sensitive GIS of the Cardiff housing market. An important aspect of this
GIS is the use of Ordnance Survey's ADDRESS-POINT product to geo-reference individual
properties to a resolution of 0.1 metre. Several large and complex socio-economic and
property related datasets were then attached to this coverage, including house price survey
data, local taxation data, and data from a Housing Condition Survey of one in five dwellings
in the central area of Cardiff. This GIS is one of the most comprehensive constructed for any
city, and is relatively unique in this kind of research.
The second method of analysis employs the hedonic pricing technique to impute monetary
values for the implicit attributes of housing. An important part of the research is an
investigation into the specification of the hedonic house price function. The traditional
specification is essentially aspatial, and does not take into account the spatial nature the data,
and thus the spatial dynamics of the housing market that generates it. To rectify this, three
different specifications of the hedonic house price function are investigated: the traditional
specification, the spatial parameter drift specification and the multi-level specification. The
research concludes that the multi-level specification is best at modelling the spatial
heterogeneity and spatial dependence inherent in housing market data. The results from this
modelling show that the valuation of locational externalities are intimately bound up with the
attributes of the housing stock and the characteristics of the resident households, resulting in
a complex juxtaposition of positive and negative valuations of location at the local level.