According to a new study released today by Veros Real Estate Solutions [1] (VEROS), Automated Valuation Models, or AVMs, may be the solution necessary to identify potential racial bias in home appraisals.
Veros’ study suggests that professional-grade AVMs can be leveraged to provide an independent valuation that could be used to help flag any potential instances of bias.
In the past several years, discussions around potential bias or discrimination within the housing market have been reignited with a focus on the property valuation process –beginning specifically with appraisals, and extending to questions about potential algorithmic bias in AVMs.
Several studies have been conducted on the subject, often coming to contradictory conclusions. Through an in-depth analysis of 50 Chicago ZIP codes, Veros sought to determine whether there is any evidence of bias in its AVM in minority communities.
Particularly, the study assessed whether the percentage of undervalued properties was related to the ZIP code’s racial composition. Chicago was selected because its ethnically diverse population spans the full spectrum of racial composition. The results revealed no evidence of racial basis in the VeroVALUESM AVM.
“We found that the proportion of properties that are undervalued by 15% or more is not correlated with the proportion of Black, Hispanic, Asian, or White populations,” said Eric Fox, Veros Real Estate Solutions Chief Economist. “This is crucial because it addresses concerns about undervaluations of minority-owned properties.”
Veros’ study is significant as the gap in minority homeownership has been pulled into the spotlight. In June of 2021, President Biden announced the creation of a task force on Property Appraisal and Valuation Equity, known as PAVE, to focus on eliminating racial bias in home appraisals. Since then, various discussions have emerged on whether AVMs could be part of this solution.
“The appraisal industry can use an effective, professional-grade AVM like VeroVALUE to address this problem,” said Fox. “An objective, cost-effective solution would be to use such an AVM to check if appraised values are in agreement and if so, would be deemed low risk for bias. Those appraisals in significant disagreement with a professional-grade AVM could be appropriately flagged or escalated for a more detailed review.”
Veros Research Economist Reena Agrawal also stressed the need for more analysis across all property valuation tools and services to determine the absence of any human or algorithmic bias.
“We also plan to continue the analysis of the VeroVALUE AVM by conducting studies in other cities across the country to obtain a more granular idea of its performance across regions,” said Agrawal.
The VeroVALUE AVM can be found in detail, here [2].