The Hidden Bias in Mortgage Approval Algorithms
By EMMANUEL MARTINEZ and LAUREN KIRCHNER, The Markup
The new four-bedroom home in Charlotte, NC was Crystal Marie and Eskias McDaniels’ personal American dream, the reason they moved from beloved Los Angeles.
A long, lush lawn, 2,700 square feet of living space, a sparkling kitchen, a neighborhood pool and a play area for their son, Nazret. All for $ 375,000.
Pre-qualifying for the mortgage was a snap – they had high credit scores, earned around six figures each, and had saved more than they would need for the down payment.
But two days before they were supposed to sign, in August 2019, the loan officer called Crystal Marie with bad news: the deal was not going to end.
“It seemed like it was being rejected by an algorithm,” she said, “and then there was a person who could step in and decide whether or not to override it.”
She was told that she was not eligible because she was an entrepreneur, not a full-time employee, even though her colleagues were also entrepreneurs. And they had mortgages.
Crystal Marie’s colleagues are white. She and Eskias are black.
âI think it would be really naive for someone like me not to consider race played a role in the process,â she said.
A survey by The Markup found that lenders in 2019 were more likely to turn down home loans to people of color than to white people with similar financial characteristics – even when we controlled for new financial factors available to the mortgage industry. has in the past said would explain the racial disparities in loans.
Maintaining 17 different factors in a complex statistical analysis of the more than 2 million conventional home purchase mortgage applications reported to the government revealed that, compared to similar white applicants, the lenders were:
- 80% more likely to reject black applicants
- 70% more likely to refuse Native American applicants
- 50% more likely to refuse applicants from Asia / Pacific Islands
- 40% more likely to reject Latino applicants
These are national tariffs.
When The Markup looked at cities and towns individually, it found disparities across 90 subways covering all regions of the country. Lenders were 150% more likely to reject black applicants in Chicago than similar white applicants there. Lenders were over 200% more likely to reject Latino applicants than white applicants in Waco, Texas, and to reject Asian and Pacific Islander applicants than white applicants in Port St. Lucie, Florida. And they were 110% more likely to turn down Native American applicants in Minneapolis.
âThe lenders were like, ‘It’s because you don’t have the lender profiles; ethno-racial differences would disappear if you had them, âsaid JosÃ© Loya, an assistant professor of urban planning at UCLA who has extensively studied public mortgage data and examined The Markup’s methodology. “Your work shows that this is not true.”
The American Bankers Association, Mortgage Bankers Association, Community Home Lenders Association, and Credit Union National Association have all criticized the analysis.
In written statements, the ABA and MBA rejected our findings for not including credit scores or government loans, which are mortgages guaranteed by the Federal Housing Administration, the Department of Veterans Affairs and others. .
Government loans have different approval thresholds, which bring people into the market who would not otherwise qualify, but generally cost buyers more. Even the Federal Reserve and the Consumer Financial Protection Bureau, the agency that publishes data on mortgages, separate conventional and government loans in their research into credit disparities.
It was impossible to include credit scores in the analysis because the CFPB removes them from the public version of the data – in part due to lobbying from the mortgage industry, citing borrower confidentiality.
While mortgage lending decisions are formally made by the loan officers at each institution, they are largely driven by software, most of which is commissioned by a pair of quasi-government agencies.
Freddie Mac and Fannie Mae were founded by the federal government to boost homeownership and now buy about half of all mortgages in America. As a result, they basically set the rules early on in the mortgage approval process.
They require lenders to use a particular credit scoring algorithm, “Classic FICO”, to determine whether an applicant meets the minimum threshold to consider for a conventional mortgage in the first place, currently a score of 620.
Launched over 15 years ago based on the 1990s database, Classic FICO is widely viewed as detrimental to people of color because it rewards traditional credit, to which they have less access than white Americans. It does not account for, among other things, on-time payments for rent, utilities and cell phone bills, but will lower people’s scores if they fall behind on those bills and get sent to debt collectors. . Unlike newer models, it penalizes people for past medical debts once they have been paid.
Yet Fannie and Freddie have resisted a flood of plaintive demands since 2014 from lawyers, the mortgage and housing industries, and Congress to authorize a newer model. They did not answer questions as to why.
The approval process also requires a green light from Fannie or Freddie’s automated subscription software. Even their regulator, the FHFA, isn’t sure exactly how they decide, but some of the factors companies say their programs take into account can affect people differently based on their race or ethnicity, researchers have found.
For example, traditional banks are less likely than payday loan vendors to place branches in neighborhoods with predominantly black populations. Payday lenders don’t report payments on time, so they can only hurt credit.
Concert workers who are people of color are more likely to report these jobs as their primary source of income, rather than a side activity, than white concert workers. It can make their income riskier.
Considering a plaintiff’s assets beyond the down payment, which lenders refer to as âreserves,â can cause particular problems for people of color. Largely due to intergenerational wealth and the racist policies of the past, the typical white family in America today has eight times the wealth of a typical black family and five times the wealth of a Latino family. White families have larger savings accounts and equity portfolios than people of color.
The chairman of the business group representing real estate appraisers recently acknowledged that racial prejudice is prevalent in the industry, which sets property values ââand has launched new programs to combat prejudice.
“If the data you provide is based on historical discrimination,” said Aracely PanameÃ±o, director of Latin American affairs at the Center for Responsible Lending, “then you are essentially cementing the discrimination on the other end.”
In written statements, Fannie said its software scans applications “regardless of race,” and both Fannie and Freddie said their algorithms are regularly evaluated for compliance with fair lending laws, internally and by the Commission. FHFA and the Ministry of Housing and Urban Development. HUD said he asked the pair to make changes accordingly, but would not release details.
Many large lenders also direct applicants through their institution’s underwriting software. How these programs work is even more mysterious; they are also owners.
Some fair loan advocates have started to question whether the mortgage value system should be changed.
“As an industry, we need to think about the least discriminatory alternatives, even if they are a valid predictor of risk,” said David Sanchez, former policy analyst at the FHFA, who currently heads research and development at the national non-profit community. Stabilization confidence. “Because if we let risk rule all of our decisions, we’re going to end up in exactly the same place where we are now in terms of racial equity in this country.” “
The lender of Crystal Marie and Eskias McDaniels denied that race had anything to do with their refusal. In an email, LoanDepot vice president of communications Lori Wildrick said the company follows the law and expects “fair and equitable treatment” for every candidate.
The couple refused to give up after the loan officer told them the mortgage hadn’t worked and asked their real estate agent for help. Crystal Marie’s employer sent several vouching emails.
At around 8:00 pm the day before the initial closing date, Crystal Marie received an email from the lender: âYou are cleared to close. She still doesn’t understand how she came to yes, but she was relieved and delighted.
âIt means so much to me, as a black person,â said Crystal Marie, who said her family descended from slaves in neighboring South Carolina, âto own property in a place where there is not so long ago you were a property.
âIt means so much. “
This story was reported by The Markup, and the story and data was distributed by The Associated Press.