Mass appraisal at the census level: Israeli case

Abstract

Mass appraisal valuation is increasingly based on multiple regression analysis in order to fit valuation statistical model for calculating market value of all properties in a target population. In this paper we develop and assess the predictive performance of different models for the estimation of dwelling values for all types of dwellings at a nationwide level in Israel. In order to investigate the representativeness of transaction data, the regression coefficients derived from the hedonic models estimated on transaction prices and on self-reported dwelling valuations, are compared. The results of this analysis are used to build several prediction models. Regression coefficients' stability is tested by a quantile regression method that allows properties to be divided into sufficiently homogeneous estimation cells parsed by geographic and economic criteria. Our prediction model allowed obtaining acceptably precise dwelling value estimators at a census tract level. The proposed method applied to additional time points, gains stable estimators with rather small fluctuations in accuracy indices and standard deviations. Dwelling values estimated at the nationwide level enable to produce new statistical products at high geographic resolutions on a range of topics, e.g., the behavior of the housing market, the economic profile of residential areas, well-being and inequality, etc.