EFFICACY OF THE GEOGRAPHICALLY WEIGHTED MODEL ON THE MASS APPRAISAL PROCESS

Single transparent models are increasingly being sought to account for site and capital value as well as location in the geographically smaller outer metropolitan urban suburbs in which there was often assumed little or no spatial variation. This paper compares the use of a Geographically Weighted Regression (GWR) hedonic model with the more traditional hedonic models where there is market evidence of both vacant land and improved residential values. This study found that where there is evidence of spatial variation and a presence of both improved and at least some vacant land sales the GWR model exhibited specification limitations that the more traditional models such as hybrid models did not and that the latter were able to more accurately predict vacant land prices while also accounting for location.