Estimating Property Values by Replication: An Alternative to the Traditional Grid and Regression Methods

This paper proposes and develops a replication method for estimating property values, in which optimal weights of comparable property attributes that best duplicate the subject property are determined. In a setting where the number of comparables is large compared to the number of attributes, replication weakly outperforms traditional general least squares regression by making use of potential correlations in the error structure. A similar result obtains in comparison to the grid method, which may suffer from subjective price adjustment factors. The replication method suggests using a large sample regression analysis to obtain the functional form of the error variance-covariance, and then replicating the subject with a smaller, attribute-close set of comparable properties.