Mortgage lenders are more likely to turn down homebuyers of color than white applicants with similar debt-to-income and loan-to-value ratios, and algorithms are likely to blame, according to in-depth analysis of 2.4 million purchase loan applications by The Markup, a nonprofit newsroom that’s “watching big tech.”
Nationwide, Black applicants were 80 percent more likely to be denied conventional mortgages eligible for backing by Fannie Mae and Freddie Mac compared to similarly qualified white applicants, the analysis found.
The Markup’s analysis of 2019 loan data concluded that Latino homebuyers were 40 percent more likely to be rejected, and Asian/Pacific Islanders were turned down 50 percent more often than whites. The chances that Native American homebuyers would be turned down were 70 percent higher than for whites.
“In every case, the prospective borrowers of color looked almost exactly the same on paper as the White applicants, except for their race,” the study’s authors said.
In addition to the national level findings, The Markup reported racial disparities in 89 metropolitan areas including New York, Los Angeles, Chicago, Houston, Atlanta, Dallas, Phoenix, Minneapolis and Washington, D.C.
The Markup also singled out seven lenders with the widest disparities for applicants of color: DHI Mortgage Company, Lennar Mortgage, Pulte Mortgage, Freedom Mortgage Corporation, Movement Mortgage Corporation, Fairway Independent Mortgage Corporation and Navy Federal Credit Union.
All of those lenders told The Markup that they comply with fair lending laws, and several disputed the publisher’s analysis.
Details of The Markup’s analysis of local markets and individual lenders, along with the methodology employed, were published separately (How We Investigated Racial Disparities in Federal Mortgage Data).
Lending industry groups question findings
Lending industry groups were critical of the The Markup’s methodology, saying its analysis did not take into account borrowers’ credit scores or analyze applications for government-backed FHA, VA and USDA mortgages.
“An individual’s credit history can help explain why seemingly comparable applicants may not always end up with the same lending outcome,” the American Bankers Association told The Markup in a written statement. “The Markup’s analysis not only fails to consider credit history, it also fails to include the millions of mortgages that resulted from Federal Housing Administration and other government loan programs. These programs are specifically designed to serve the low-to-moderate-income families most at risk of being denied a mortgage. This glaring omission paints an incomplete picture of the mortgage market.”
The Mortgage Bankers Association provided similar feedback to The Markup before publication. After the analysis was published Wednesday, the MBA denounced it as “not only deeply flawed but clearly biased in its premise,” and misrepresenting “the problems and solutions needed to solve the very serious issues that result in unequal outcomes related to Black homeownership and wealth-building.”
Although the Home Mortgage Disclosure Act (HMDA) requires lenders to provide loan-level credit score data to regulators, the Consumer Financial Protection Bureau strips that information from public data, “in part due to the mortgage industry’s lobbying to remove them, citing borrower privacy,” the authors of The Markup’s analysis said.
Authors Emmanuel Martinez and Lauren Kirchner — who both have extensive experience in data-based investigative reporting — said they were also unable to analyze decisions made by Fannie Mae and Freddie Mac’s underwriting algorithms. Like credit scores, lenders report those decisions in their HMDA filings, but they aren’t available in public files.
Credit scores and algorithms to blame?
Nevertheless, Martinez and Kirchner concluded that credit scores, along with algorithms used by individual lenders and Fannie and Freddie, are the likely culprits for the disparities they say their analysis uncovered.
The Classic FICO score that Fannie and Freddie require lenders to use, they said, is “widely considered detrimental to people of color because it rewards traditional credit, to which White Americans have more access. It doesn’t consider, among other things, on-time payments for rent, utilities, and cellphone bills — but will lower people’s scores if they get behind on them and are sent to debt collectors. Unlike more recent models, it penalizes people for past medical debt even if it’s since been paid.”
Some of the inputs considered by Fannie and Freddie’s underwriting algorithms — including assets, employment status and debts — can also put people of color at a disadvantage, Martinez and Kirchner noted, citing other research.
“This is a relatively new world of automated underwriting engines that by intent may not discriminate but by effect likely do,” former Mortgage Bankers Association president and CEO David Stevens told The Markup.
Stevens and others told The Markup that outsiders have little visibility into the inner workings of the algorithms employed not only by Fannie Mae’s Desktop Underwriter and Freddie Mac’s Loan Prospector systems, but also similar software used by the FHA, USDA and individual banks and lenders.
Fannie Mae — which recently announced it plans to start taking on-time rental payments into account when evaluating borrowers with thin credit files — provided a written statement to The Markup saying Desktop Underwriter analyzes loan applications “without regard to race.”
Both Fannie and Freddie “said their algorithms are routinely evaluated for compliance with fair lending laws, internally and by the FHFA and the Department of Housing and Urban Development,” The Markup reported. “HUD said in an email to The Markup that it has asked the pair to make changes in underwriting criteria as a result of those reviews but would not disclose the details.”
Potential for new regulations
In a written statement provided to The Markup, the MBA said regulators have access to a richer set of data and haven’t been troubled by HMDA data in the past.
“The Federal Reserve, CFPB, and other regulators have been clear in saying that denial disparities in the HMDA data are not in themselves determinative with respect to assessing fair lending,” the MBA said. “Fair lending examinations include a much richer set of information regarding loans and borrowers.”
In a new analysis of 2020 HMDA data, however, the CFPB concluded that Black and Hispanic borrowers “continued to have fewer loans, be more likely to be denied than non-Hispanic White and Asian borrowers, and pay higher median interest rates and total loan costs. It is clear from that data that our economic recovery from the COVID-19 pandemic won’t be robust if it remains uneven for mortgage borrowers of color.”
And The Markup noted that the Biden administration’s new secretary of housing, Marcia Fudge, recently told Axios that she believes one reason Black homeownership rates have declined is that “we have never totally enforced the Fair Housing Act.”
In June, HUD said it’s planning to reinstate rules put in place by the Obama administration in 2013 for addressing “disparate impact” — discriminatory practices that are unintentional, but nevertheless unjustified. That could create legal liability for lenders who employ artificial intelligence and algorithms to evaluate borrowers.
Fannie and Freddie’s federal regulator, the Federal Housing Finance Agency (FHFA), wants the mortgage giants to meet new goals that support lending within minority census tracts. The FHFA, which wants at least 35 percent of the purchase mortgages backed by Fannie and Freddie to be taken out by low- and very-low income borrowers, is seeking comments on the proposed goals for 2022 through 2024.
