GSEs further expand AVMs, desktop and hybrid appraisals

GSEs further expand AVMs, desktop and hybrid appraisals

The Federal Housing Finance Agency released voluminous plans last week developed by Fannie Mae and Freddie Mac to make the housing market more equitable, in part through changes to the appraisal process.

Fannie Mae and Freddie Mac’s equitable housing finance plans further expand non-traditional property appraisals, which sometimes rely on property tax information, data collected by third parties, or algorithms to assess a property’s value. Both GSEs argue that these approaches advance equity.

“Using automated tools to establish home values helps remove human bias, although it limits the collection and evaluation of a property’s current condition,” wrote Freddie Mac.

Desktop appraisals and hybrid appraisals, where an independent third-party inspects the property, both “reduce costs to the borrower and reduce potential risk of bias by creating greater separation between the appraiser and borrower,” wrote Fannie Mae.

Non-traditional appraisals to the rescue

Freddie Mac research concluded in 2021 that appraisal gaps, which negatively impact Black and Latino borrowers and homeowners, exist on purchase appraisals. That was before a study spelled out the appraisal industry’s regulatory dysfunction, and before a federal task force promised to combat appraisal bias.

A follow-up study Freddie Mac conducted in May found that “even after controlling for important factors that affect house values and appraisal practices, properties in Black and Latino tracts are more likely to receive appraisal values that fall below contract prices, and this likelihood increases as the Black or Latino concentration in the neighborhood increases.”

The cure? According to Freddie Mac’s equitable housing finance plan, it could be the expansion of automated valuation models.

Using its automated valuation models “leads to relatively lower racial gaps,” Freddie Mac said.

Freddie Mac currently uses that technology to speed appraisals on some purchase transactions, but only those with loan to value ratios up to 80%, but that excludes most Black and Latino borrowers. Starting in 2023, Freddie Mac said it will look at expanding the use of its automated collateral evaluation for mortgages with loan to value ratios greater than 80% through a targeted lending program.

But researchers at the Urban Institute recently found that automated valuation models, while they “represent the promise of greater efficiency and lower costs for the mortgage industry,” perform differently in majority-Black neighborhoods. 

The researchers write that, “even with data improvement and artificial intelligence, we still find evidence that the percentage magnitude of AVM error is greater in majority-Black neighborhoods. This indicates that we cannot reject the role historic discrimination has played in the evaluation of home values.”

A Freddie Mac spokesperson said that the GSE and its regulator conduct routine fair lending analyses to ensure the automated system fully complies with fair lending laws, including a review to ensure that no factor is a proxy for protected classes.

“We are constantly refining our system, and we frequently bring in new technologies to improve our capabilities,” a Freddie Mac spokesperson said.

Automated valuation would largely benefit lenders via more efficient originations and borrowers potentially through reduced costs and a shorter wait time from application to approval, Freddie Mac said. Freddie Mac would also benefit, “via an understanding of the valuation method and the ability to deploy it consistently throughout the organization,” a company spokesperson said.

Asked about the implications of the Urban Institute findings, a Fannie Mae spokeswoman emphasized that its equity plan focuses on the expansion of desktop and hybrid appraisals, which was not the focus of that research.

Fannie Mae also included efforts to “modernize” appraisals in its equity plans. The GSE will modify its selling guide to allow for desktop as an appraisal option, after its large-scale experiment with that option as a result of the COVID-19 pandemic. In March, Fannie Mae said it would start offering desktop appraisals for some loans.

But its equity plan also looks to increase use of hybrid appraisals — those where the property inspection is done by an independent third party.

“Both options reduce costs to the borrower and reduce potential risk of bias by creating greater separation between the appraiser and borrower,” the Fannie Mae plan read.

Introducing a third party to the transaction to conduct the inspection in order to reduce costs and eliminate bias does not sit well with some industry stakeholders. Appraisers have, in the past, fretted over liability and data reliability. Other stakeholders wonder how cost reductions would impact them.

Peter Christensen, principal at Christensen Law firm, which advises on legal and regulatory matters concerning valuation, believes that third-party data collection will both reduce costs and the potential for bias.

Those performing property data collection do not command the hourly rates that certified appraisers do. Using a third party is “moving that labor to essentially the lowest common denominator,” Christensen said.

As for its mitigating effect on bias, separating the analysis from the data collection can counter biases. A property data collector may well have unconscious biases unleashed by a photo of a Black family on the wall. But the data collector would not include his analysis or a photo of the family portrait in the report for the appraiser, said Christensen, who has written contracts for property data collectors.

Still, there are some kinks to work out of third-party property data collection, which Christensen described as the “Wild West.”

“Appraisers until this point don’t get very good fair housing training, but while it’s not perfect, far from, USPAP does make a reference to fair housing law,” said Christensen. “But for all the weaknesses amongst appraisers, who is training the average property data collector on this stuff? Fair housing, are you kidding me?”

My data, my research

Both of the GSEs have plans to conduct research to better understand bias. Neither indicate they will give outside researchers the ability to replicate that research, however.

Fannie Mae said it would use its database of roughly 54 million appraisals to analyze undervaluation that could indicate bias. Fannie Mae said it would share those research results via an external industry memo, and then a research paper for industry stakeholders sometime in the first half of 2022.

Among the 21 proposed research projects proposed in Freddie Mac’s equitable housing finance plan is an appraisal gap analysis that Freddie Mac said would help it understand “if, how and why” automated valuation might be part of the solution to the appraisal gap.

That research, however, will be conducted by Freddie Mac researchers. No outside researchers will be checking Freddie Mac’s analysis to test the conclusions it reaches. It’s a limitation that has irked academics and researchers who have long sought access to appraisal data from both the government sponsored enterprises.

“Certainly the GSEs have smart researchers,” said Michael Neal, a former Fannie Mae official who is now a researcher at the Urban Institute. “But democratizing the data they have available would benefit policy development.”

Both Fannie Mae and Freddie Mac also have plans to ramp up their quality control systems, following a FHFA blog that, while light on specifics, found at least some instances of references to protected classes in appraisal reports.

Fannie Mae will implement a “new awareness flag data point” that will activate when internal data indicate a possible undervaluation, triggering a targeted quality control review by the second quarter of 2022.

Freddie Mac also wants to use technology to better detect undervaluations or the use of “biased words or phrases,” such as “pride of ownership,” or “crime ridden.” It hopes to deploy the capability to detect undervaluations this year, and detect the use of biased language by 2023.

Read More

Write a comment

Your email address will not be published. All fields are required