Urban Wire AI Could Alter Mortgage Lending, but Government Leadership Is Needed
Michael Neal, Janneke Ratcliffe, Matthew Pruitt
Display Date

.
Body

On October 30, 2023, the Biden administration announced steps to ensure the safe deployment of artificial intelligence (AI) within key industries and broadly across the US and global economy. With AI capabilities accelerating rapidly, establishing government oversight protocols to protect consumers, especially communities of color, can help pave the way for equitable innovation and adoption.

Although the announcement does not address mortgage lending specifically, mounting evidence confirms AI has great potential in the industry, but many questions remain about the security and safety implications. For borrowers of color specifically, algorithmic processes could add to the barriers that have caused long-standing disparate outcomes across the mortgage cycle and further entrench the racial homeownership gap.

Body

In our new study, Harnessing Artificial Intelligence for Equity in Mortgage Finance, we interviewed more than four dozen experts and mortgage stakeholders, scanned current research, and drew on economic theory to examine the tension between equity and efficiency. We learned that AI has the potential to generate powerful efficiency gains while correcting the long history of racially biased outcomes in mortgage lending, but this dual promise of greater equity and greater efficiency is far from guaranteed.

The current landscape of AI in mortgage lending

For many mortgage stakeholders, the primary motivation behind investing in AI is increasing efficiency through lower costs, faster responses, and greater per unit revenue. But current AI adoption is uneven across the loan sourcing, production, and management process. For instance, AI adoption is gaining traction in marketing, property valuation, and document processing but lags in areas such as mortgage servicing. 

Across the ecosystem, large mortgage lenders and financial technology firms are leading early adoption and innovation, having more capacity to absorb the high up-front costs AI typically requires. But most actors, particularly smaller and mission-oriented lenders such as minority depository institutions and community development financial institutions, still lag in adopting AI. Several nonprofit organizations are working to close this technological gap between larger and smaller lenders. Particularly as more mission-oriented institutions deepen AI adoption, the potential risks for communities could grow. Although larger lenders are more likely to have a larger impact on market activity, mission-oriented lenders typically have more concentrated impact in communities of color and for historically marginalized borrowers with nontraditional financial profiles.

In the mortgage sector, the lack of federal action and clarity around AI has prevented greater AI penetration in the industry. The Federal Housing Authority (FHA), the US Department of Veterans Affairs (VA), and the government-sponsored enterprises (GSEs), Fannie Mae and Freddie Mac, take the lead in establishing lending and servicing rules for most mortgages and control the models used to approve or deny loans. Without clear guidance from those entities, individual providers may have less incentive to innovate unilaterally. The complex federal regulatory landscape may further curb experimentation, especially where there are unknown risks to consumers or questions about safety and soundness, as there are with current AI models.

This caution is warranted, as much of this regulatory framework was originally developed in response to systematic bias and discrimination, and our report finds that AI has the potential to amplify existing racial disparities or create new ones. These disparities can lie within the algorithms, stem from the data used to train the algorithms, or arise within the algorithm’s iterative learning process. AI also introduces data privacy concerns, which may compound risk of exploitation for the most vulnerable borrowers. Without the transparency of traditional rules-based approaches, there is a nontrivial risk that biases will be imbedded, magnified, and hidden in AI black boxes. But some of the most promising tools to deter bias in AI-based models explicitly incorporate race, so regulators will need to clarify how fair credit and fair lending rules apply.

Implementing AI will require intentionality and clear guidelines

From our analysis, we recommend a combination of guidance from regulators and leadership from federal and federally chartered mortgage institutions to ensure that equity is centered in the adoption of AI in the home lending arena. To do so, these stakeholders can prioritize the following:

  • Intentional design. AI model designers could be required to determine whether the inputs and the underlying data that models learn from reflect the housing market’s disparate racial impacts. Prioritizing equity must happen from the start, not as an afterthought.
  • Pilot programs and innovation. The FHA, VA, and GSEs, including the Federal Home Loan Banks, could run pilot programs to test how AI models affect industry and consumer outcomes.
  • Increased regulatory guidance. Federal regulators could improve trust and equity in AI in the mortgage finance industry through interagency coordination and by setting clear guidelines for its application and debiasing.

These actions will advance equity by clearing the path for innovation and will spur greater AI adoption, which could drive market efficiencies. As the industry struggles with profitability challenges—rooted in market conditions and origination costs—and grapples with a legacy of institutionalized racism, federal mortgage leaders should consider taking steps to ensure that market use of AI supports both equity and efficiency.

Research Areas Housing finance Artificial intelligence
Tags Fair housing and housing discrimination Federal housing programs and policies Homeownership Housing finance data and tools Housing finance reform Housing markets Racial barriers to housing Technology, trade, and automation
Policy Centers Housing Finance Policy Center
Related content