WASHINGTON – The Financial Services Innovation Coalition (FSIC) and Creative Investment Research released a report that finds technology and math can be used to reduce racial disparities in mortgage lending.
The report, Artificial Intelligence and Algorithmic Lending Has the Potential to Reduce Discrimination in Mortgage Lending, explores the rise of non-bank lenders and their adoption of AI and algorithmic systems, the obstacles lenders face in creating non-biased underwriting, and the path forward for continuing to reduce discrimination in the home buying process.
“For too long, Blacks and other minorities in the United States have been victimized by bias – both conscious and unconscious – in the lending sector,” says Kevin Kimble, founder and CEO of FSIC. “Color blind application of appropriate data can help reverse this historical inequity.”
AI and algorithmic systems remove much of the human element from underwriting. That lowers the potential for bias, the group claims, piggybacking on an earlier study that found AI-based systems reduced racial bias by 40% and showed no discrimination in rejection rates.
The report also argues non-bank lenders have helped facilitate the automatic-loan-approval process. In 2013, non-bank lenders accounted for fewer than 40% of all loans. By May 2019, these lenders wrote nearly two-thirds of all new mortgages.
“Non-bank lenders are certainly part of the solution,” Kimble says. “The growth of these lenders, which have fewer qualifying factors and lower down-payment requirements than traditional banks, has opened the door to homeownership for millions of previously underserved families, especially in low-income and minority communities.”
However, the report states that loan underwriting based on any past credit analysis will always have some inherent bias. To address this, the authors call on the federal government to reform qualifying factors, as government-sponsored enterprises (GSEs) such as Fannie Mae and Freddie Mac guarantee over 90% of mortgages in the U.S.
“The full potential of AI and algorithmic lending systems cannot be realized until the GSEs modify the qualifying factors for the conventional home loans they’re willing to buy or insure,” Kimble says.