Spatial Regression Modeling for Mass Value Estimation from Cadastral Cartography

Abstract

This research proposed the construction and performance evaluation of multiple linear regression models (conventional and spatial) for the city of Itororó (BA) and thus enable the elaboration of a Generic Values Plant (PVG) and estimation of the Urban Property Tax (IPTU). To this end, the elaboration process of these models included aerophotogrammetric mapping, and real state register, spatial analyzes, multicollinearity, normality and homoscedasticity of the residues, as well as spatial dependence tests according to NBR 14.653-2/2011. The results indicated that incorporating the effects of spatial autocorrelation through the reduced spatial lag model provided better performance than the conventional one. However the construction of the geographically weighted regression model also reduced, able to model spatial heterogeneity, it was even more adequate, providing almost all the explanation of the predicted values variation, as well as a sharp reduction of the prediction errors and the dispersion coefficient. The extrapolation of this model provided the elaboration of PVG with total values and simulation of IPTU. Thus, the increase in the 1% tax rate would provide a considerable share of internal revenues for municipal revenue.