Geospatial Modeling of House Price Using MGWRSAR: A Case Study in DKI Jakarta

The dynamics of house prices in Jakarta are influenced by both spatial dependence and spatial heterogeneity. Conventional models, including the Spatial Autoregressive (SAR) and Geographically Weighted Regression (GWR), are only capable of partial capture of these effects, which often result in model misspecification. The present study employs the Mixed Geographically Weighted Regression-Spatial Autoregressive (MGWR-SAR) model to concurrently address the spatial dependence and heterogeneity in house price determinants. We used a dataset of houses listed in 2024, incorporating house and location characteristics. The construction of spatial dependence weight matrix was based on semi-variogram analysis, and optimal kernel bandwidths were selected using cross-validation method. The model is estimated using The Spatial Two-Stage Least Square (S2SLS) approach to address endogeneity in spatial lag component. The findings of the study indicate that the land and building area are the most significant price determinants, with spatial variation in their influence across Jakarta's subregions. Jakarta Pusat exhibits stronger spatial spillovers, while Southern district demonstrates heightened sensitivity to building area. The MGWR-SAR model effectively eliminates residual spatial autocorrelation and captures location-specific variation, offering enhanced explanatory power for urban housing market. These findings underscore the importance of incorporating spatial structure in housing price modeling for more targeted policies and valuation strategies.