Optimal Comparable Selection and Weighting in Real Property Valuation

This paper formalizes certain aspects of the sales comparison approach to valuation that heretofore have been quite ad hoc. Specifically, it applies statistical theory to decisions about how many comparables to select, what the criteria for comparable selection should be, and how the proper weights for each adjusted value estimate can be determined such that the final value estimate is both unbiased and of minimum variance. Several results are derived that run counter to conventional practice; for example, it may not always be optimal to consider first the "best" comparables because of a lack of independence among their adjusted value estimates. Furthermore, it is always desirable to consider more comparables (regardless of how "bad") so long as their adjusted value estimates are optimally weighted in the final value estimate. Finally, weights usually selected for "inferior" comparables are typically too small. A final exercise empirically applies the methodology to a sample of sales.