US mortgage approval algorithm racial discrimination

August 2021
Updated: February 2022

An investigation has found that mortgage lenders are much more likely to turn down Latino, Asian, Native American and Black applicants than White ones.

A Markup assessment of mortgage data, co-published by the AP, discovered that mortgage loan applicants of colour were 40%–80% more likely to be denied than their White counterparts across the US, and that the disparity was greater than 250% in some urban areas.

The algorithms used by lenders are mostly mandated by Freddie Mac and Fannie Mae, whose own underwriting algorithms are a closely held secret.

The Markup's investigation was cited as a reason for the introduction of the Algorithmic Accountability Act of 2022. AIAAIC's Legal Transparency Requirements Repository sets out the transparency requirements for the bill.

Operator: Freddie Mac, Fannie Mae
Developer:
Freddie Mac, Fannie Mae
Country: USA
Sector:
Banking/financial services
Purpose: Assess mortgage applications
Technology: Underwriting algorithms
Issue: Bias/discrimination - race, ethnicity
Opacity: Governance; Black box

Reference

Research, audits, investigations


News, commentary, analysis