Are Appraisers Statisticians?

Traditional hedonic pricing models, based upon an impressive corpus of statistical and economic theory, often exhibit prediction errors with a standard deviation in the range of 28–50%.1 In contrast, statistically challenged appraisers following ad hoc procedures often exhibit prediction errors with a standard deviation around 10%.2 The juxtaposition of these purported facts suggests the fruitfulness of examining elements of appraisal practice from a statistical perspective. In this vein Pace and Gilley (1998) showed the grid adjustment estimator employed by appraisers is a restricted version of the simultaneous autoregressive (SAR) estimator from spatial statistics. They suggest spatial statistics provides a unifying intellectual framework for reconciling appraisal practices with statistical theory.3