Using the Spatial Configuration of the Data to Improve Estimation

Using the well-known Harrison and Rubinfeld (1978) hedonicpricing data, this manuscript demonstrates the substantialbenefits obtained by modeling the spatial dependence of theerrors. Specifically, the estimated errors on the spatialautogregression fell by 44% relative to OLS. The spatialautoregression corrects predicted values by a nonparametricestimate of the error on nearby observations and thusmimics the behavior of appraisers. The spatialautoregression, by formally incorporating the arealconfiguration of the data to increase predictive accuracyand estimation efficiency, has great potential in realestate empirical work