Land Price Model Considering Spatial Factors

Abstract

Many studies have highlighted that Ordinary Least Square (OLS) regressions lack the ability to consider spatial dependency including spatial non-stationarity which then lead to bias and inefficient estimations. Land prices are usually depending on locations yielding the prices vary from place to place. Therefore, estimates obtained from the OLS ignoring spatial factors may be inappropriate. Geographically Weighted Regression (GWR) is an alternative model considering spatial non-stationarity. In addition to its appropriateness, GWR produces local specific parameter estimates which then are very useful for the policy makers to avoid a misleading judgment. Some geographic social characteristics and related infrastructures are often used as the determinants or the explanatory variables of applied models for the land prices. In this study, the residuals of both GWR and OLS models were contrasted to obtain the best model. The maps of coefficients of the determinants varied from place to place. Also, we found that the most influential factors were the distance to the high-way gate, the distance to the artery-road and the distance to public facilities.