Use of Spatial Regression Models in the Mass Appraisal of Real Estate in the Coastal Neighborhoods of Maceió, Alagoas, Brazil

This paper presents the results obtained in the research that has been carried out to study the price of apartments in the coastal neighborhoods of the city of Maceió, State of Alagoas, Brazil: Ponta Verde, Cruz das Almas, Jatiúca, Pajuçara. The research studies the price of real estate in these seaside neighborhoods as support for understanding the value of the square meter practiced. This research uses a comparative method of market data using spatial econometric models supported by Geographic Information Systems (GIS). The data on the properties on offer was collected from real estate agents and from property listings on the Internet. It was then systematized in electronic spreadsheets, with the plane coordinates georeferenced to the UTM system and the information imported into QGis. The computer programs GeoDa and GeoDaSpace were used to carry out Exploratory Spatial Data Analysis and classical and spatial regressions for the estimation process. The results obtained show that the Spatial Lag (SAR) model using two-stage spatially generalized Instrumental Variables (IV) estimators (S2SLS + MGM/HAC) was able to eliminate the effects of spatial autocorrelation of the residuals present in the classic linear regression model. The exogenous and endogenous variables used to compose the model were statistically significant. These variables characterize the surrounding information, the characteristics of the properties and the socio-economic aspects of the neighbourhood. The results show that apartment prices are more influenced by the variables garage, sea view, floor area, apartment age, average per capita household income and land occupation rate.