Report finds racial bias in mortgage AI choices, MBA fires again

Even decades after the Fair Housing and Equal Credit Opportunity Act was in place, loan denial rates lean heavily towards non-white borrowers, according to research by The Markup.

Compared to white mortgage applicants with similar financial characteristics, black consumers were on average 80% more likely to be denied, according to the non-profit analysis of data from the Home Mortgage Disclosure Act of 2019. Native Americans are 70% more likely to be rejected, 50% more Asian or Pacific Islanders, and 40% more likely to Latinos.

The report analyzed 2 million home purchase applications for conventional mortgages across the country. The analysis did not include loans secured by the Federal Housing Administration or the Department of Veterans Affairs, which have lower qualification standards and lower borrower profiles compared to traditional loans.

The Mortgage Bankers Association denied the suggestion that the industry perpetuated racial discrimination. The report was said to misrepresent the problem of inequality and its findings did not give a complete picture.

FICO scores were not used in the report because they are not published. However, traditional credit scores can be inherently disadvantageous for people of color as they have historically not had equal access to the banks or assets that build credit. Credit ratings have not taken rental payments into account, although Fannie Mae will begin using a positive rental payment history as an insurance factor from September 18 (those who have an inconsistent record of on-time rental payments will not get those records) as such Payments can be made in cash that would not have electronic evidence).

"Analysis of HMDA (from The Markup) data and its predetermined mortgage lending conclusions fail to take into account several key components that form the backbone of lending decisions, including a borrower's creditworthiness and credit history," the MBA said in a press release. "As we told the authors, the Federal Reserve, CFPB, and other regulators have made it clear that denial differences in HMDA data alone cannot be used to assess fair lending."

Today, much of the credit decision making and underwriting process is automated. The algorithms that power the technology are said to be "color blind" and not consider race or possible human biases. However, the problem might have less to do with the fairness of the algorithms themselves and more to do with the basic data on which they are based.

"This is how structural racism works," said Chi Chi Wu, a staff member with the National Consumer Law Center, in the report. "This is how racism is embedded in institutions, guidelines and practices without any animus."

According to HMDA data, credit history was the reason for mortgage rejection for 36.9% of Native American applicants, 32.5% of blacks, 21.3% of Latinos, 20.9% of whites, and 12.1% of Asians .

Some providers have tried to offer a more comprehensive rating for those who do not have a comprehensive credit history. VantageScore, a third-party credit score provider, estimated that under its model, 37 million borrowers would become creditworthy without a FICO score, and a third of that number would be either Black or Latino.

"They've been testing alternative ratings for years and I don't know why the process is taking so long," said Lisa Rice, president and CEO of the National Fair Housing Alliance, a consortium of hundreds of fair housing organizations. "Deserved consumers are left behind."

Related Articles