Abstract
It has been more than 50 years since the passage of the Fair Housing Act, yet the homeownership rates of Blacks continue to be some 30 percentage points below those achieved by non-Latinx Whites. The relatively large Black-White homeownership gap reflects a variety of complex forces, ranging from continuing discrimination in the housing and mortgage markets to ongoing disparities in household incomes and debt burdens to policies that serve to disadvantage minority households. The onset of the Covid-19 pandemic is likely to exacerbate these issues, given the disparate impact the virus is having on both the physical and economic health of communities of color.
This paper focuses on the extent to which alternative ways of evaluating credit risk in the mortgage market could foster—or impede—the achievement of fair housing goals and increase homeownership rates, particularly for Black Americans. The mortgage market primarily relies on credit-bureau data to assess an applicant’s creditworthiness—a critical element of the underwriting process. Yet for a variety of reasons, Blacks are more likely to have weak or missing credit scores than other segments of the population. This has led many housing advocates to call for the use of alternative data that could either enhance or replace the data contained in a typical credit report.
Our analysis focuses on three major types of alternative data that could be used in the evaluation of a consumer’s creditworthiness:
● credit proxies, which capture the consumer’s payment history on on-going bills such as rent, utilities, cable, and telecommunications;
● banking data, which capture activity in the consumer’s banking and checking accounts; and
● non-financial personal data, which could include anything derived from a consumer’s digital footprint, ranging from their “likes” and “dislikes” to the places they visit or shop to the characteristics of their Facebook friends.
We then evaluate each type of data using a five-factor “SCALE” framework that incorporates several important considerations in addition to the data’s predictive power. These include:
● Societal values, i.e., Is the use of the data consistent with general social and ethical norms?
● Contextual Integrity, i.e., Does the data make sense in the context of mortgage lending?
● Accuracy, i.e., Does the data accurately reflect the household’s situation?
● Legality, i.e., Would the use of the data have a disparate impact on protected classes?
● Expanded opportunity, i.e., Would the use of the data expand access to mortgage credit by increasing the number of qualified borrowers or by reducing borrowing costs while maintaining or improving the accuracy of credit decisions?
Failure to meet any of these criteria does not necessarily mean that the data should not be used, nor does it imply that policymakers must intervene to prohibit or discourage their use. However, viewing alternative data sets through this multifaceted lens serves to highlight their potential strengths and weaknesses from a fair housing perspective, and will hopefully lead to better decisions on the part of credit providers, regulators, and Congress.
In general, application of the SCALE framework supports the use of “credit proxies” such as the timely payment of rent and utilities, as well as certain types of aggregate banking data, for example, net monthly inflows or outflows and total savings. However, because of fair lending, ethical, and contextual concerns, we believe that the use of certain types of granular banking data (e.g., where and how the household spends its money), as well as the plethora of data that can be harvested from social media (e.g., Facebook friends, shopping patterns, internet searches, “likes”, etc.), should generally be discouraged.
Based on these findings, we offer four broad recommendations to policymakers:
• First, existing legislative efforts to encourage the reporting of credit proxies such as rent, utilities, and telecom payments should be supported and strengthened by explicitly pre-empting existing state or local laws that prevent the sharing of such data. Not only does the use of such data make “sense” in the context of mortgage lending; its use has also been shown to increase the number of qualified minority borrowers.
• Second, bank regulators should continue to explore ways to encourage the use of certain banking data in the assessment of credit risk. While much of these data are already being used in mortgage underwriting, its digitalization would undoubtedly make the process more efficient and lead to better lending decisions. Banking data might also be used to capture credit proxies such as rent and utility payments. At the same time, however, certain types of banking data—for example, detailed checking data on how consumers spend their money—should be discouraged since they could easily serve as proxies for the consumer’s race, ethnicity, or sex.
• Third, Congress should revisit the ECOA, the Fair Housing Act, and other applicable laws to explicitly preclude the use of most, if not all kinds of non-financial personal data in lending decisions. Putting clear restrictions on how various types of consumer data can be used in credit decisions—and ensuring the transparency of the data to consumers—will help to address the most egregious misuses of consumer data and promote fair lending goals.
• Finally, policymakers should explore ways to mitigate the impact of the Covid-19 pandemic on homeownership opportunities for Blacks and other historically disadvantaged groups. However, this should not include prohibiting credit bureaus from collecting and reporting delinquency data during the pandemic or removing the forbearance flags that are currently included in credit reports. Such data could ultimately be key to understanding how consumers have responded to the challenges raised by the pandemic, which in turn could provide more accurate assessments of credit risk.