A modelling approach with geographically weighted regression methods for determining geographic variation and influencing factors in housing price: A case in Istanbul

Abstract

Determining real estate market dynamics has become an important issue in the city economy for achieving sustainable urban land management and investment planning. This study aims to determine the potential influencing factors of housing prices through applying global regression models including Ordinary Least Squares (OLS), Spatial Lag Model (SLM), and Spatial Error Model (SEM) and to examine their geographic variation by local regression approaches such as Geographically Weighted Regression (GWR) and Multiscale Geographically Weighted Regression (MGWR). Pendik district of Istanbul (Turkey) was selected as study area. For these purposes, a case geodatabase was created with twenty-eight attributes: structural, geographic, and neighbourhood variables. Initially, nine significant variables used as input in the other models were selected with a stepwise OLS approach. The modelling results suggested that local models achieve better performance in comparison to global models. Furthermore, local model factor coefficients and local R2 values were mapped for nine variables. Geographic variations of these variables with regards to housing price were analysed for GWR and MGWR models. This study provides a comprehensive methodology for the real estate market and city authorities to understand the geographic variations of factors affecting the real estate market in urban geography.