Bias hidden in mortgage approval algorithms

0
17

[ad_1]

EMMANUEL MARTINEZ & LAURENE KIRCHNER, The Markup

The new four-bedroom home in Charlotte, North Carolina was the personal American dream of Crystal Marie and the Eskias McDaniels, the reason they moved there from expensive Los Angeles.

A lush long lawn, 2,700 square feet of living space, a gleaming kitchen, and an adjoining pool and play area for their son Nazret. For only $ 375,000.

Pre-qualifying for a mortgage loan was easy: they had high credit ratings, earned roughly six figures, and accumulated more than they needed to make a down payment.

But two days before they were supposed to sign, in August 2019, a loan officer called Crystal Marie with the bad news: The deal was not going to close.

“It seemed like the algorithm was rejecting it,” she said, “and then there was a person who could step in and decide whether to undo it or not.”

She was told she was inadequate because she was a contractor and not a full-time employee, even though her colleagues were contractors too. And they had a mortgage.

Krystal Marie’s colleagues are white. She and Eskias are black.

“I think it would be naive for someone like me not to consider that race played a role in this process,” she said.

An investigation by The Markup found that lenders in 2019 were more likely to deny home loans to people of color than to whites with similar financial characteristics, even as we controlled the new available financial factors that the mortgage industry has claimed in the past to be. explain racial disparities in lending.

By keeping 17 different factors stable in a comprehensive statistical analysis of over 2 million common home mortgage applications submitted to the government, it was found that, compared to similar white applicants, the lenders were:

  • 80% more likely to reject black candidates
  • 70% more likely to reject a Native American nomination
  • 50% more likely to reject Asian / Pacific Islander candidates
  • 40% more likely to reject Latin American candidates

These are national rates.

When The Markup examined the cities individually, it found differences in 90 subways, covering all regions of the country. Lenders were 150% more likely to reject black candidates in Chicago than similar white candidates there. Lenders were more than 200% more likely to reject Hispanic applicants than whites in Waco, Texas, and rejected applicants from Asia and the Pacific than whites in Port St. Lucie, Florida. And they were 110% more likely to refuse Native American applicants in Minneapolis.

“Lenders told us, ‘It’s because you don’t have a credit profile; ethno-racial differences would disappear if you had them, ”said José Loya, assistant professor of urban planning at the University of California, Los Angeles, who has carefully studied government mortgage data and studied The Markup’s methodology. “Your work shows that this is not true.”

The American Bankers Association, the Mortgage Bankers Association, the Community Housing Lenders Association, and the National Credit Unions Association have criticized this analysis.

In written submissions, the ABA and MBA rejected our findings because they did not include credit ratings 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, resulting in people entering the market who would otherwise be ineligible but usually cost more to buyers. Even the Federal Reserve and the Bureau of Consumer Financial Protection, the agency that publishes mortgage data, separate conventional and government loans in their research on loan inequality.

It was impossible to include credit ratings in the analysis because the CFPB is removing them from the public release of the data – in part due to lobbying by the mortgage industry, citing borrower privacy.

While home loan decisions are formally made by loan officers at each institution, they are mostly made by software, much of which is sanctioned by a couple of quasi-government agencies.

Freddie Mac and Fannie Mae were founded by the federal government to stimulate home ownership and now buy about half of all mortgages in America. As a result, they essentially set the rules from the very beginning of the mortgage approval process.

They require lenders to use a special credit rating algorithm, “classic FICO”, to determine if an applicant meets the minimum threshold that must be considered for a regular mortgage, which is currently 620 points.

Launched over 15 years ago based on 1990s data, Classic FICO is widely considered harmful to people of color because it rewards traditional credit, which they have less access to than white Americans. It does not account for timely payments for rent, utility bills, and cell phone bills, among other things – but would lower people’s ratings if they delay paying those bills and send them to debt collectors. Unlike later models, it punishes people for past medical debt after it is paid off.

However, Fannie and Freddie have resisted a flood of complaints since 2014 from lawyers, representatives from the mortgage and housing sectors, and Congress to authorize the new model. They did not answer questions about why.

The approval process also requires a green light from Fannie or Freddie’s automatic underwriting software. The researchers found that even their regulator, the FHFA, doesn’t know exactly how they decide, but some of the factors that companies say their programs take into account can affect people differently depending on their race or ethnicity.

For example, traditional banks are less likely than payday loan sellers to open branches in areas populated mostly by people of color. Payday lenders do not report timely payments, so they can only damage the loan.

Giant workers who are of color are more likely to see this work as their main source of income, rather than a side occupation, than white convertibles. This can make their income more risky.

Considering an applicant’s assets other than the down payment, which lenders refer to as “reserves,” can cause special problems for people of color. Thanks in large part to the wealth of generations and past racist policies, the typical white family in America today has eight times more wealth than the typical black family and five times more than the Hispanic family. White families have larger savings accounts and stock portfolios than people of color.

The president of a trading group representing real estate appraisers recently acknowledged that racial bias is widespread in the property value industry and launched new programs to combat bias.

“If the data you enter is based on historical discrimination,” said Araceli Panameño, director of Latin American affairs at the Responsible Lending Center, “then you are essentially reinforcing discrimination on the other side.”

Fannie said in written statements that its software analyzes applications “without regard to race,” and Fannie and Freddie said their algorithms are regularly reviewed for compliance with fair credit laws both internally and at FHFA and the Housing and Urban Development Department. … HUD said it asked the couple to make changes as a result, but did not disclose details.

Many large lenders also launch candidates through their institutions’ own underwriting software. How these programs work remains a mystery; they are also proprietary.

Some proponents of fair lending have begun to question whether the value system in mortgage lending should be changed.

“As an industry, we need to think about which alternatives are less discriminatory, even if they are a valid predictor of risk,” said David Sanchez, a former FHFA policy analyst who currently leads research and development at the nonprofit National Community. Stable trust. “Because if we let risk drive all our decisions, we will be in the same place we are now when it comes to racial equality in this country.”

The lender to Crystal Marie and Eskias McDaniels denied that race had anything to do with their denial. In an email, Lori Wildrick, vice president of communications at CreditDepot, stated that the company follows the law and expects every applicant to be treated “fairly and equitably”.

The couple refused to give up after the loan officer told them the mortgage had failed and enlisted the backing of their real estate agent. Krystal Marie’s employer sent several letters for her.

At around 8:00 pm the night before the original closing date, Crystal Marie received an email from the lender: “You can close.” She still doesn’t understand how she got to yes, but she felt relief and glee.

“It means so much to me as a black man,” said Crystal Marie, who said her family was descended from slaves in neighboring South Carolina, “to have property in a place where you were property not many generations ago.

“It means so much.”

This was reported by The Markup, and the story and data were released by the Associated Press.

[ad_2]

Source link