There is no single silver bullet that will rectify the pernicious impact of bias in residential real estate valuations — it is a complex problem that requires a multifaceted solution. But there is the promise of a better future on the horizon. The housing industry and Biden administration have begun a full throttled effort to solve the issues contributing to inequity.
By combining the power of emerging technologies like artificial intelligence (AI) and machine learning (ML) with a true commitment to diversity at every stage of the valuation chain, we can build a consciously unbiased appraisal system that delivers more equitable and accurate conclusions.
Yet, this will be a time-consuming and challenging effort that will require all parties to be aligned in their goals and their approach. First, we must acknowledge the impact that appraisal bias continues to have on minority homeowners and the broader housing system.
Despite the protections enacted by the Fair Housing Act of 1968, inequality and discrimination in the housing finance system continue to shape the contemporary valuation of homes. Recent studies from Fannie Mae, Freddie Mac, and academic researchers have found appraisal disparities for communities and borrowers of color.
As a result, homes located in minority neighborhoods have been chronically undervalued, exacerbating the racial wealth gap. Fannie Mae’s “Appraising the Appraisal” study comparing appraisals to automated valuation model (AVM) data for refinance transactions found that Black borrowers received slightly lower appraisal values relative to AVMs, while white borrowers received slightly higher appraisals.
Likewise, a 2021 Freddie Mac study of more than 12 million appraisals dating back to 2015 found that appraisers’ opinions of value were more likely to fall below the contract price in Black and Latino census tracts. Some researchers point to appraisal methodology as a culprit.
According to a study published by Oxford Academic Journals in 2020, modern appraisal techniques using sales comparisons and neighborhood comparisons actually perpetuate racial inequality, and in some cases exacerbate it, citing that the sales comparison approach preserves historical racial bias in today’s home values.
Similarly, the study found that neighborhood comparisons used in appraisals may be influenced by racialized assumptions of a neighborhood. Thankfully, this topic has been getting the attention it deserves from industry leaders, regulators and the current administration.
In March, the Biden administration announced a multi-step plan to advance equity in the appraisal process. The administration’s interagency taskforce on Property Appraisal and Valuation Equity (PAVE) issued an action plan including regulatory reforms and oversight to make the appraisal industry more accountable, provide consumers with assistance and awareness, prevent algorithmic bias, drive more diversity in the appraisal industry, and leverage federal data to benefit research and policymaking.
If passed, the Fair Appraisal and Inequity Reform Act of 2022 would establish a Federal Valuation Agency responsible for promoting a fair, unbiased, transparent and repeatable valuation process. The industry is ready and willing to tackle this challenge.
We all share a goal of creating a system that produces more accurate valuations free of racial and historical bias to open a more equitable path to wealth creation for all Americans. The difficulty is agreeing on exactly what a consciously unbiased appraisal system looks like, and how to achieve it.
Ideally, the solution leverages more advanced technology, better data, and a more diverse workforce across every sector of the valuation spectrum. With advances in data engineering and modeling, we now have the technology and tools available to begin correcting some of the bias and issues in the modern valuation process.
By working from objective data rather than information processed and curated by humans, AI and machine learning technologies can reduce subjectivity and unconscious bias from appraisals. However, every valuation technology provider has their own secret sauce when it comes to their algorithm for AVMs and other computer-based valuation tools.
A slew of agencies including the CFPB, FDIC, NCUA, and FHFA have collaborated to address this with newly proposed quality control standards for AVMs. Their proposed amendment to FIRREA aims to increase confidence scores, prevent manipulation of data, avoid conflicts of interest, and enforce random sample testing and reviews.
More research is needed to determine which algorithms should be used to ensure AVMs do not introduce their own model bias. This would require extensive testing using historical and current data to determine if the estimates generated by the technology accurately reflects reality.
Additionally, the standardization of data is another crucial variable that must be addressed for this approach to succeed. Just as each valuation technology provider relies on different algorithmic models, they also rely on different data sources. It will be impossible to standardize models unless the data all models are running on comes from a single source of truth.
One proposed solution is for the GSEs to provide open access to their data sets, which constitute the most comprehensive collection of comparative real estate data nationwide. That way, all participants are comparing apples to apples without the possibility of oranges sneaking into the data set.
There has been broad support for this approach across the public and private sectors, from fair housing advocates to valuation industry leaders, but it remains to be seen whether the FHFA will authorize the release of GSE data.
While technology and standardization are important tools to create a more equitable valuation process, a diverse workforce is another critical check on subjectivity and unconscious bias. There is a severe underrepresentation of diverse talent in the housing industry.
According to the Department of Labor’s Bureau of Labor Statistics, the appraiser profession is 97.7% white, and women comprise only 30.4% of the workforce. Looking at the broader scale, less than 13% of the housing industry workforce is Black and Hispanic. As an industry, we need a more diverse workforce and leadership that better reflects the population we serve. In response to this, the Appraisal Institute has launched an Appraiser Diversity Initiative with Fannie Mae, Freddie Mac, and the National Urban League.
Other initiatives like Fannie Mae’s Future Housing Leaders program are focused on sourcing a more diverse talent pipeline and matching them with employment opportunities in the housing industry. It is important to also focus inclusive recruiting efforts within the valuation technology and data science field, including those building and maintaining computer-based models.
These initiatives will take time, but with a consistent, united effort across the industry we can ensure there is an emphasis on promoting diversity when hiring new entrants and promoting to leadership positions. There may not be a single action or reform that can instantly solve the persistent issue of biased home appraisals, but there are ways to improve and, perhaps over time, remedy the problem using a combination of technology and diverse data.
Through a merger of expert knowledge, diversity of thought, standardized data, and advanced technology, we can develop more equitable valuation processes that are consistent, repeatable, and transparent. The scope of the challenge should not discourage us. Rather, the reward of achieving a more fair and equitable system that serves all Americans is well worth the effort.
This article was first featured in the May HousingWire Magazine issue. To read the full issue, go here.
Kade Clark is the senior vice president of Red Bell Real Estate, a homegenius company. He is responsible for the daily management of digital valuations, hybrid appraisal valuations and sales within the homegenius division of Radian.
This column does not necessarily reflect the opinion of HousingWire’s editorial department and its owners.
To contact the editor responsible for this story:
Brena Nath at [email protected]
