How Digitalization Changes How People Search for, Finance Housing

According to a new analysis from Harvard's Joint Center for Housing Studies (JCHS), research shows big data and data analytics have significantly changed how people search for and finance housing.

The changes have the potential to reduce or eliminate the longstanding biases that have made it harder for people of color to obtain affordable, high-quality housing in desirable neighborhoods.

While that hasn’t happened, two new working papers, which were first presented at Harvard's housing and digitalization symposium last year, contend that carefully designed regulations and public policies could help advance fair housing goals.

In “Digitalization of the Housing Search: Homeseekers, Gatekeepers, and Market Legibility,” Geoff Boeing (University of Southern California), Julia Harten (University of British Columbia), and Rocio Sanchez-Moyano (Federal Reserve Bank of San Francisco) report that while online housing searches seem likely to offer more options for those who have historically been shut out of certain markets, the evidence is mixed.

Online rental listings in neighborhoods with mainly Black or Latino residents have less information about unit and neighborhood amenities than listings in otherwise equivalent neighborhoods where most residents are non-Hispanic whites. They also note that the search for housing is just one step in a complex process that still involves agents, brokers, and other gatekeepers. As a result, the benefits of digitalization concentrate in already-advantaged communities, which indicates that it’s simply reinforcing longstanding patterns of residential sorting and segregation.

Policymakers, regulators, and practitioners could take several steps to address this, the authors note. They could use the data that is generated to better understand market conditions and design responses to challenges, including affordability. They could also prevent platforms from using personal information to create targeted marketing campaigns or screening criteria that might be at odds with fair housing goals. In addition, they could require that platforms used in housing searches include information about both units and neighborhoods and ensure that disadvantaged communities can easily access the information.

In “Algorithms for All: Has Digitalization in the Mortgage Market Expanded Access to Homeownership?” Vanessa Perry (GW School of Business) and Kirsten Martin (University of Notre Dame) note that access to mortgages could be supported by changes in four areas: the growing use of digital and mobile technologies in banking; the widespread use of digital advertising by lenders and others selling services to homeowners; the growing interest in using big data to facilitate alternative approaches to credit scoring; and the potential use of Artificial Intelligence (AI) in the appraisal process and mortgage underwriting.

However, the evidence that this is happening is lacking and inconclusive. In several areas, there is evidence that digitalization could even worsen discriminatory practices. For example, while online platforms can expand access to information about financing, several cases brought against Facebook showed that these platforms have sometimes allowed lenders to target their audiences in ways that are at odds with fair housing goals.

Similarly, while some research suggests that expanding the types of data used in underwriting has reduced racial disparities, other work indicates that the new forms of data could be proxies for demographic characteristics that will continue to bias decisions.

Perry and Martin argue that opportunities exist for proactive, responsible digital transformation to remove systemic barriers to mortgage credit access. Doing so requires that policymakers and regulators incorporate societal, ethical, legal, and practical criteria into efforts to monitor and evaluate the use of new digitalized tools.

The situations laid out in both papers underscore the need for public policies to govern digitalization, notes Laurie Goodman (The Urban Institute) in her commentary on the papers. New rules should ensure that the algorithms produce outcomes that comply with fair housing and anti-discrimination laws.

They also should set limits on whether and how personal data can be used in marketing or pricing. Finally, policymakers should require regular public release of information about how algorithms affect protected classes of people. This approach, she notes, already is being used in Neighborhood Watch, an online monitoring system developed by HUD and FHA to track mortgage delinquencies by location and lender.

Carrying out these efforts requires more clarity about key values and goals, writes Lauren Rhue (University of Maryland, Smith School of Business) in her commentary on the papers. Illustratively, she observes, fairness in credit is relatively easy to understand: all borrowers with the same profile should receive the same amount of credit. In contrast, “societal disagreement about what fairness means for previously redlined neighborhoods and their residents” will hamper efforts to use digitalization to address discrimination in those neighborhoods.

To read the full report, including more data and methodology, click here.