No abstract

Purpose -This study examines and documents spatial heterogeneity in Istanbul housing market using Geographically Weighted Model (GWR). Methodology -A GWR model with a Gaussian kernel and an adaptive bandwidth based on cross-validation is employed on a cross-sectional housing listing data set. Additional analysis is provided using geographically weighted Spearman's rank correlation measure between prices and variables. Findings-GWR model substantially boosts goodness of fit in our pricing model compared to a standard hedonic regression model. The variation within GWR coefficients is high and of micro nature. Median GWR coefficients often differ from standard hedonic regression coefficients. The variability of coefficients is plotted on map. Conclusion-Findings suggest the existence of spatial non-stationarity in standard hedonic regressions and favor the use of models appropriate for spatial heterogeneity. Findings encourage further research in hedonic models applications such as in quality adjustments to price indices.

Real estate properties are naturally location-fixed, therefore, a spatial dependence is expected. When location-specific factors persist over time, spatial autocorrelation is likely to exist in a hedonic pricing regression. Spatial autocorrelation causes problems in the interpretation of the regression results due to inefficient estimators or complex models. The focus of this study is to identify and test the determinants of spatial dependence in the real estate market. Using a novel data set, our study contributes to the literature as we identify the specific spatial dependence factors through the analysis of different types of housing markets and extend the literature to an emerging market.

The effect of space on real estate prices is critical due to the inherent nature of real estate properties. There exist numerous methodologies in order to incorporate the effect of space into the analysis including the addition of district variables or the use spatial econometric models. Spatial models help with the identification and inference problems compared to the simple hedonic models, yet are harder to work with due to their complexity and resource requirements. In this study, using geospatial data, the authors explain the spatial econometric models and tools available for the analysis of cross sectional real estate data.