Efficacy in Modelling Location Within the Mass Appraisal Process

Abstract

Although Geographic Information Systems (GIS) has long been recognised as a natural partner in the computer assisted mass appraisal (CAMA) process, it has not always been clear how CAMA management (practitioners) may be able to utilise this partnership to produce regular re-assessments at an acceptable level of accuracy and cost. The objective of this study is to help demonstrate how this may occur by examining the accuracy generated by various simple, transparent and cost effective approaches traditionally used to model location as part of the CAMA process and compare the accuracy of the predicted result to that generated by using an integrated GIS environment to model location. Models were constructed to account for ‘location’ in three ways. They are firstly, in an a priori fashion based on established suburb and post code administrative boundaries. Secondly, by utilising the GIS to generate location factors based on the residuals of location ‘blind’ global hedonic models and creating an interpolated location factor surface that can be applied to global hedonic models to give a predicted value. Finally, by using hedonic Geographically Weighted Regression (GWR) that allows the regression coefficients to vary across geographic space in response to local variation. These last two approaches take advantage of the parcel’s spatial coordinates to model location within a GIS environment. All three approaches are used to generate values using available secondary data normally collected as part of the determination of capital market value integrated within the spatial framework of a digital cadastre. The results indicate an acceptable degree of accuracy can be achieved when using basic hedonic GWR models that account for location in an intuitively simple way, thus providing transparency and efficiency to the mass appraisal process. GWR accounts for location giving comparatively more accurate results at little or no extra cost.