The Hedonic Modeling of Property Prices: Case Study on Cape Town, South Africa

Abstract

The real estate and valuation discipline has embraced the recent developments in spatial data analysis as a way of remediating obvious limitations of the ordinary least squares (OLS) approach in handling spatial effects. However, despite the development, the South African property market has yet to embrace spatial analysis when estimating property prices. The objective of this study is to understand the rationale behind the use of spatial hedonic modeling on the Cape Town property market by first testing the data against the existence of spatial effects and using the appropriate techniques to correct the glitches. A spatial error autocorrelation model and geographically weighted regression (GWR) were employed to correct spatial dependence (autocorrelation) and spatial heterogeneity on 3,232 observations. The relative performance of the two spatial modeling techniques as revealed by their goodness-of-fit are quite impressive but the spatial error model marginally outperforms the GWR. Thus, it is recommended that any of these techniques can be used in modeling property prices in the Cape Town market.