A Hybrid Cost and Market-Based Estimator for Appraisal

Straightforward application of OLS to statistical mass appraisal models does not always produce defensible estimates due to debilitating levels of multicollinearity in the data as appraisers attempt to improve the predictive ability of their models through the inclusion of a long list of correlated property characteristics. The traditional cost approach does not suffer from this malady, but often does not predict market values well. An ideal assessment method would combine the predictive power of statistical appraisal models with the lack of cross-sectional error from the cost approach. This paper presents such a method and demonstrates its advantages both theoretically and empirically. The hybrid estimator, based upon a priori information and current market data is superior in prediction to standard OLS estimation under a variety of conditions. This superiority is demonstrated through cross-validation on random hold-out samples of various sizes.