Reena Agrawal of Veros: Using AVMs to Test for Appraisal Bias

Reena Agrawal is a Research Economist at Veros Real Estate Solutions, Santa Ana, Calif., a provider of enterprise risk management, disaster data and collateral valuation services. She is a subject matter expert for all affairs related to real estate economic research and valuation solution. For more information on industry research and AVMs, contact her at

Reena Agrawal

We all know appraisals are necessary for determining the fair market value of a home whenever a mortgage is required in buying or refinancing a property. Unfortunately, appraisals are not always fair. Whether implicit or explicit, bias can be present in some appraisals.

Appraisal bias occurs when a property is undervalued or inaccurately assessed based on any of the protected class variables. Any appraisal bias based on race of the homeowner or racial make-up of a neighborhood can seriously affect minority homeowners. Whether they are trying to sell their property or tap into its equity, receiving less than what their home is actually worth can drastically affect many financial outcomes for a minority household.

How does appraisal bias affect minority homeowners?

This goes beyond the transaction itself. Thinking long-term, an undervalued appraisal impedes a household’s ability to build wealth. Not only does this diminish what a household can afford in the present, but it also diminishes what it can pass down to future generations. According to Census data, home equity is the second largest component of households’ wealth in the U.S. after retirement accounts, but homeownership rates for minority communities are much lower than that for White families, according to Pew.

Any discrimination in the housing market that makes homeownership less accessible to minority communities can widen the wealth gap between them and White households. Especially during economic crises, households with lower wealth are worse-off because they have fewer financial resources to adapt and recover.

How can you identify appraisal bias?

Technology can help identify appraisal bias or discrimination in a few different ways. While an appraisal is a manual process that typically involves the appraiser inspecting the target property, technology tools like an Automated Valuation Model (AVM) can help eliminate the potential for a conscious or unconscious bias.

AVMs have the unique advantage that they do not use any demographic and socioeconomic characteristics information about the homeowner or neighborhoods, whereas these factors could potentially influence an appraiser in their valuation decision.  Hence, an AVM can offer an objective and unbiased value determination and can serve as a quality control check for appraisals by comparing an appraisal’s estimate of value with an AVM estimate. Any significant difference between the two estimates could flag the appraisal for further review.

Furthermore, bias in appraisals can be flagged by use of technology that can detect biased or subjective words in an appraisal.  Performing a bias word search can help identify potential red flags from the language used by the appraiser in their comments in the appraisal report. There are different words or phrases that are more subjective than factual and can denote potential bias. Phrases such as “pride of ownership,” “crime-ridden area,” “gentrified,” “undesirable/desirable neighborhood,” and “up-and-coming,” can raise a red flag.

By switching to correct and unbiased language, appraisers can ensure they are using objective data for valuations rather than subjective data or opinions. For example, instead of stating that a home is desirable and affordable, an appraiser can list the amenities of the home and state if the valuation is aligned with the neighborhood price range. Appraisals that have been flagged for review based on bias words can then be escalated for a further review.

How do you know there’s no bias built into an AVM?

 As a first step, an AVM modeler can ensure that none of the inputs that they use in their models or algorithms directly represent, or measure protected classes as defined by the Fair Housing Act. They can then test their model predictions or estimates using several methodologies. For example, they can test if their AVM is free of bias by investigating the correlation between large undervaluations by their AVM (difference between sale price and AVM value) against the racial compositions of neighborhoods. If the two are uncorrelated, or if undervaluations do not increase as minority populations increase, then it can be inferred that the AVM is unbiased. The modeler should ideally test their models in several different geographical regions of the country to ensure that their AVM is unbiased.

On the other hand, if an AVM is consistently bringing back a larger proportion of lower-than-expected valuations in minority areas, then there is likely some algorithmic bias that needs to be addressed. Given the rising concerns regarding bias in the housing industry, it has become imperative that industry stakeholders including AVM vendors as well all other providers of valuation services address this issue to provide a fair and equitable transaction to everyone.

What can the industry do to address potential appraisal bias?

 A key step taken by the industry to address discrimination in housing has been through the Property Appraisal and Valuation Equity (PAVE) Task Force, which was commissioned by President Biden to evaluate “the causes, extent, and consequences of appraisal bias and to establish a transformative set of recommendations to root out racial and ethnic bias in home valuations.” The task force includes 13 different government agencies including the FHFA, HUD, the CFPB, the Federal Reserve and the FDIC.

The housing industry is also responding to bias concerns by promoting and advocating for increased diversity in the appraiser community. The Appraisal Diversity Initiative (ADI) is a nationwide program led by Fannie Mae, Freddie Mac, and the National Urban League. The goal of ADI is to reach out to diverse, talented candidates and educate them about the appraisal profession, as well as provide resources for interested candidates to help them get started in the profession.

Tackling the issue of bias is no easy task but having both a proactive and reactive strategy can help lenders, appraisers, and technology providers act on their commitment to combat and mitigate bias. Addressing bias in a meaningful way will require a group effort through a combination of technological solutions, diversity initiatives, and policy changes.

(Views expressed in this article do not necessarily reflect policy of the Mortgage Bankers Association, nor do they connote an MBA endorsement of a specific company, product or service. MBA NewsLink welcomes your submissions. Inquiries can be sent to Mike Sorohan, editor, at; or Michael Tucker, editorial manager, at