Location factors in real estate mass appraisal: comparing approaches

In practice, the structure of a location-based model for real estate appraisal (type of the spatial autocorrelation, neighbor count) is unobvious. We develop the simulation-based methodology to compare accuracy of computer-assisted mass appraisal (CAMA) routines. Then we compare routines differing in the form of location information accounting and build the best routine selection rule basing on spatial autocorrelation type in source data. We also provide routine robustness comparison.