Housing Technology Series: Equitable Homeownership Innovations

Innovative AI models can help lenders mitigate bias in lending decisions. Photo credit: iStock.com/fstop123

Studies show that, on average, Black Americans face more difficulties in qualifying for mortgages than their White counterparts. The nearly 30-point homeownership gap between Black and White households has remained virtually unchanged since the 1960s. On April 9, 2024, HUD and the University of California Berkeley’s Terner Labs hosted a symposium highlighting these persistent inequities in the housing market and investigated strategies to make homeownership more widely accessible. The third event in the Housing Technology series, “Equitable Homeownership Innovations,” featured two panels moderated by Terner Center staff as well as multiple speakers. The first panel, which featured Kasey Matthews, senior data analyst at Zest AI; Sarah Edelman, deputy assistant secretary for single family housing at HUD; and Syeed Mansur, chief executive officer of GreenLyne, discussed the extent to which technological innovations can reduce racial biases in underwriting. The second panel, which included Alex Cabral, senior principal of innovative finance at Grounded Solutions Network; Sonya Mays, president and chief executive officer of Develop Detroit; and Michael Neal, senior fellow at the Urban Institute’s Housing Finance Policy Center, investigated alternative paths to homeownership for those without ready access to conventional mortgages. Additional speakers included Thomas Voutsos, cofounder and chief executive officer of LadderUP Housing; George Scott, principal at Blackstar Stability; and Solomon Greene, principal deputy assistant secretary of HUD’s Office of Policy Development and Research.

Fairness in Underwriting

From discrimination in the mortgage market to targeted subprime products, households of color generally have been unable to obtain home financing at the same rate as their White counterparts. Even credit scores, which were created to serve as an unbiased assessment of credit risk, often result in less equitable outcomes for women and communities of color. Mansur believes that some models used to determine credit risk are outdated and lead to discriminatory outcomes. He used the example of loan-to-value ratio, which compares the mortgage amount to the property's appraised value. "We do not find in the work that we do that this variable has a meaningful impact on improving the probability of default or the probability of prepayment, but we do see it being used as a ratio against which lenders will make a decision on whether to lend or not to lend," Mansur said. The panelists pointed out that inequities in the homeownership market help widen the Black-White wealth gap.

Innovations in Underwriting

Some companies are designing software to reduce bias in underwriting. Matthews explained how Zest AI is developing artificial intelligence (AI) models to help lenders weed out bias. For example, Zest’s Race Predictor can estimate the race of applicants more accurately than the existing standard model, making it easier to ensure that, when applicants do not self-report their race, lenders do not deny loans to Black applicants at disproportionately higher rates than White applicants. Models by Zest and GreenLyne also incorporate adversarial debiasing, a method to mitigate bias in lending decisions. Mansur said that GreenLyne is developing models that yield predictions for default and prepayment risk that are less biased and more accurate than standard models, thus reducing the cost of the loan. “Instead of raising prices to counter the risk of the loan, there's an opportunity here to use the predictive models to find a sweet spot where the loan size is optimum so that we can subdue some of that default risk,” Mansur explained. Meanwhile, the Federal Housing Administration, which serves homeowners at the lower end of the market, has updated its algorithms to expand credit access for first-time buyers. Edelman said that the agency’s 2022 decision to allow borrowers to use a positive rental history to support their mortgage qualifications has increased the number of first-time homebuyers who can qualify for a mortgage.

Alternative Homeownership Models

Some arrangements allow homeowners to purchase houses without conventional loans. For example, several speakers highlighted shared equity models, such as community land trusts and cooperatives, which allow low-income families to purchase subsidized homes subject to long-term affordability requirements such as limited resale values. The subsidy, typically provided by a government entity or nonprofit, acts as a first downpayment and reduces the monthly mortgage. Households who purchase the home sign a ground lease or deed restriction with resale provisions. Owners who later choose to sell the house can recapture their initial equity plus a certain percentage of the appreciated value, and the remainder returns to the shared equity organization and is passed on to the next buyer. Although this arrangement reduces the gains to the initial homebuyer from the home’s appreciation, Mays emphasized that reducing renters’ housing cost burden must remain the organization’s main priority. In fact, approximately 60 percent of shared equity buyers eventually purchase homes in the more traditional housing market, suggesting that shared equity purchases can serve as a stepping stone for first-time homebuyers. Cabral also pointed out that shared equity arrangements still leave homeowners with approximately $14,000 in equity, significantly more than they spent for their initial purchase.

In other instances, organizations can help individuals and families advance toward homeownership by supporting them while they are renters. Voutsos discussed how LadderUp Housing purchases and renovates homes that it leases to eligible working-class households, helping those with subpar credit scores eventually become homeowners. LadderUp uses alternative methods to assess financial security. Reiterating the point earlier speakers made, Voutsos explained that “credit score doesn't tell the whole story, so we're looking at other factors, such as on-time utility payments, rent payments, internet payments, car insurance, [and] anything that's not captured in that credit score that can determine if somebody's financially stable.” During the renters’ tenure, a nonprofit financial coaching organization helps the tenants increase their credit scores, including by promptly paying their rent, which is capped at one-third of their income. Once the tenant’s credit score improves to the point at which they can qualify for a mortgage, LadderUp offers to sell the rental home to the tenant. Voutsos said that it takes tenants an average of 3 years to reach the required credit score of approximately 640.

Some entities also help homebuyers obtain more favorable lending terms. Blackstar Stability purchases contracts for deeds, which are seller-financed agreements in which the homebuyers do not officially own their property until all the expenses have been paid, at which point the seller converts the agreements to conventional mortgages. Scott explained that, although contracts for deeds are not inherently bad for consumers, some sellers have used them in predatory ways. Scott said that contracts for deeds often have interest rates of between 10 and 20 percent and can have strict requirements. In some cases, homeowners can be removed for missing a single payment.

Moving Forward

Despite the potential of new technologies and methodologies, the speakers warned that some of these reforms, if not designed for all populations in mind, could simply reinforce existing biases. "Through the use of appraisal waivers and automated evaluation models … it is important to make sure that communities of color experience the same type of outcomes and accurate outcomes as others," Neal said. Mansur reiterated that new models to determine mortgage eligibility must be both fair and accurate, and he underscored that fairness can be quantified. Matthews added that having more guidelines from regulators will be key to ensuring that these new technologies and methods are fairly implemented.

The speakers also made clear that these advancements must be well tailored to and understood by their target communities to be successful, particularly among residents who generally have not benefited from new products and methods. Mays emphasized that some households in the Black community might be understandably skeptical of these innovations, including alternative homeownership models such as shared equity arrangements. "There are definitely some aspects in our society where technology has democratized things for Black and Brown people and created much more access, and so intuitively housing could be one of those places, but it's just all about how tightly tailored it is," Mays said. The speakers generally feel that recent advances in housing finance can make homeownership available to more families as long as the new methods and technologies are fairly used and applied and well understood by potential homebuyers.