Navy Federal Credit Union, the largest credit union in the US, was found to have the widest disparity in mortgage approval rates between White and Black borrowers of any major lender, with automated underwriting systems approving 75% of White applicants but less than 50% of Black applicants in 2022.
Navy Federal Credit Union, which serves military servicemembers and veterans with 13 million members and $165 billion in assets, demonstrated significant racial disparities in mortgage lending through its automated underwriting systems. In 2022, the credit union approved 77.1% of White applicants for conventional home purchase mortgages while approving only 48.5% of Black applicants, creating a 28.6-percentage-point gap that was the largest among the 50 biggest mortgage lenders. This disparity persisted even when controlling for income, debt-to-income ratios, property values, and neighborhood characteristics. CNN's statistical analysis found Black applicants were more than twice as likely to be denied as White applicants with identical financial profiles. The credit union rejected approximately 3,700 Black applicants in 2022, and similar disparities existed for VA loans where 84.2% of White applicants were approved compared to 71.6% of Black applicants. Individual cases like Bob Otondi and Ted Spencer showed Black borrowers being denied by Navy Federal but quickly approved by other lenders for the same properties. The racial gap in Navy Federal's approval rates has grown significantly from 11.5 percentage points in 2018 to 28.6 percentage points in 2022.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
Unequal treatment of individuals or groups by AI, often based on race, gender, or other sensitive characteristics, resulting in unfair outcomes and unfair representation of those groups.
AI system
Due to a decision or action made by an AI system
Unintentional
Due to an unexpected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed