We propose reshaping the residential real estate appraisal process to start with automated valuation models (AVMs) as the baseline, and then adjust that value either based on physical characteristics not used by the AVM or based on the fact that the property data the AVM used are inaccurate.

AVMs can leverage the price data of various comparable properties, decide which properties are indeed comparable, and apply any trends in the data at the neighborhood, regional, and national levels. AVMs do not use information about race or ethnicity. Meanwhile, a human appraiser can ascertain whether the data are accurate and can adjust an AVM’s valuation to reflect other physical conditions that are not recorded or that an algorithm cannot adequately detect.

Our proposal capitalizes on the comparative advantages of AVMs and human appraisers and advances recent public discourse around appraisal inequities. Moreover, it offers a way to accelerate the goals of the federal Interagency Task Force on Property Appraisal and Valuation Equity and improve its long-term results without congressional approval or major rulemaking.