Veros Real Estate Solutions recently released a study of its automated valuation model (AVM)— VeroVALUE AVM—that said there was no evidence of any racial bias. “This validation provides an important quality control check for lenders seeking tools to identify potential racial bias in home valuations,” the company claims.
“The study utilized neighborhood racial compositions to determine if a correlation existed between the percentage of Black, Hispanic, Asian and White populations and the number of properties in the area that were undervalued or overvalued by more than 15%,” it added.
“With appraisals coming under increasing scrutiny due to allegations of racial bias, AVMs have a unique advantage, as they are blind to the demographic characteristics of the parties involved in real estate transactions, including the surrounding properties and neighborhoods that are used in the determination of the AVM results,” the report itself explained. “AVMs can estimate the value of a home without any human assessment in an objective and cost and time-effective manner. Further, many AVMs utilize numerous valuation methodologies that are independently analyzed and aggregated to produce a final estimate of value, which helps ensure more accurate results. Therefore, AVMs may be used as an objective and more efficient tool for home valuations in certain situations.”
The study was undertaken in Chicago, Atlanta, Houston, Los Angeles, and Philadelphia because of their diverse demographics. The company says there was no correlation with the proportion of black of Hispanic populations in a neighborhood and the proportions of incorrectly valued properties.
The question arose because of multiple recent analyses, including some from Fannie Mae and Freddie Mac, which showed bias in home appraisals. But then other studies did not find systemic racial bias by appraisers.
“Given the concerns regarding the appraisal process, industry stakeholders have begun to raise questions as to whether any similar real or perceived bias exists in other property valuation products and services, including appraisal alternatives, hybrids, AVMs or BPOs,” the report said.
One question that jumps out is the choice of a 15% difference in valuation, which seems like a large amount.
The report did mention that concern. “Based on feedback from our first report (Agrawal & Fox, 2022), we also analyzed the performance of P10L (the proportion of properties the AVM undervalues by more than 10%) and P10H (the proportion of properties the AVM overvalues by more than 10%) as it was suggested that undervaluations or overvaluations of 15% or more were too large,” it said. “On the other hand, valuation differences of less than 10% can often be attributed to nuances of a property’s condition, quality, characteristics, and features.”