UC Berkeley researchers find that lending algorithms discriminate against ethnic minorities

coloredited_discrimination_ariel-1
Ariel Lung/Staff

Related Posts

A new study conducted by researchers from the UC Berkeley Haas School of Business and School of Law shows that the money-lending financial technology, or fintech, algorithms may discriminate against African Americans and Latinx.

Interest rate discrimination from algorithms results in interest rates 5.6 basis points higher for African Americans and Latinx, the study found. In addition, each extra basis point costs minority mortgage holders about $100 million annually, according to the study.

Fintech describes any online algorithm that helps people apply for loans, according to study coauthor Adair Morse. She added that the researchers were initially interested in whether or not algorithms eliminated discrimination.

“Although algorithms remove biases where lenders can see faces, their use of strategic pricing data induces unintentional discrimination,” Morse said.

As of 2017, 17.3 percent, or $2.25 trillion of the $13 trillion, of total U.S. household debt comes from minority households, according to the study.

White ethnicity homeownership in 2017 was found to lie at 72.4 percent, compared to 48.4 percent for Latinx and 43 percent for African Americans, according to the study. The study added that if discrimination in lending practices was identified, it would help explain this disparity.

There were three main findings from the study, according to Morse: the use of strategic pricing creates discrimination, fintech allows borrowers to obtain better loans and digital lending eliminates the loan rejection discrimination associated with face-to-face lending.

The increased interest rates faced by ethnic minorities as a result of discrimination are numerically similar regardless of whether algorithms or face-to-face lending is used, according to the study. Morse added that African Americans and Latinx are unintentionally hurt by strategic pricing and the practice of charging more in areas lenders know they can make more profits.

“We took estimates from our paper and aggregated it to the amount of mortgages existing in the economy, and our estimates suggest African American and Latinx mortgage borrowers are paying $500 million more in interest in the U.S.,” Morse said.

According to the study, the other two main findings are silver linings. Innovations, including fintech lending, allow people to get better loans, said Morse. She added this is due to competition between big banks, mortgage brokers and online platforms for customers.

Fintech lending has also eliminated minority discrimination in loan rejections, according to the study, which also found that face-to-face lenders reject ethnic minorities about four percent more often than fintechs. Morse said this may be a result of personal bias from face-to-face lenders.

“I think it’s provocative in where the future is headed in dominance of big data,” Morse said. “It’s the tip of the iceberg in how we need to think about human touch being involved in algorithm decision-making.”

Yao Huang is the lead research and ideas reporter. Contact him at [email protected] and follow him on Twitter at @Yhoneplus.