NCRC Comment on HUD’s Proposed Disparate Impact Rule


August 24, 2021

Office of General Counsel
Rules Docket Clerk
Department of Housing and Urban Development
451 Seventh Street SW, Room 10276
Washington, D.C. 20410-0001

Re: Docket No. HUD-2021-0033
FR-6251-P-01 Reinstatement of Discriminatory Effects Standard

Dear Assistant Secretary Worden:

The National Community Reinvestment Coalition (NCRC) and the undersigned organizations would like to applaud the U.S. Department of Housing and Urban Development (HUD) for this proposed rule, which would recodify the 2013 standard for disparate impact theory. NCRC is deeply appreciative of HUD’s recognition of disparate impact theory as a crucial tool for preventing housing discrimination. This recodification of the 2013 standard will affirm HUD’s commitment to a strong framework for the resolution of disparate impact claims.

As a national organization that promotes fair housing and fair lending practices, NCRC shares HUD’s belief in the importance of disparate impact theory. NCRC represents more than 600 member organizations that are all committed to fighting discrimination in their communities. The use of disparate impact theory has allowed NCRC and its members to challenge unnecessary practices that have a discriminatory effect on members of protected classes. NCRC and its members often collaborate with the real estate industry to pursue fair housing goals, which has allowed us to witness the ways that disparate impact theory stimulated innovative and inclusive business practices.

Proper enforcement of the Fair Housing Act is impossible without a strong disparate impact standard. Many important victories for housing equality have been possible only because of HUD’s recognition of disparate impact theory. Disparate impact theory has been used to protect women from being evicted just because they were the victims of domestic violence. It has also been used to prevent mortgage lenders from giving subprime loans to applicants of color while giving prime loans to White applicants, and from raising interest rates and broker fees for customers of color but not their White customers.

HUD’s 2013 Disparate Impact Rule contains the most suitable framework for robust fair housing enforcement

The Disparate Impact Rule promulgated by HUD in 2013 was greatly welcomed by fair housing advocates. The rule’s three-part burden-shifting framework is a straightforward format, which initially places a burden on plaintiffs to make a prima facie showing of disparate impact. If the plaintiff succeeds at this, the burden shifts to the defendant to prove the legitimate business purpose or necessity of the challenged practice. If this occurs, the burden shifts to the plaintiff to prove that another practice with a less discriminatory effect could meet the substantial, legitimate, nondiscriminatory interests that the practice supported.

The burden-shifting framework provides a reasonable opportunity for both plaintiffs and respondents to test policies and practices and incents financial institutions to implement robust compliance management systems to mitigate against the risk of discrimination.

The Supreme Court ruled on the applicability of disparate impact theory to the Fair Housing Act in Texas Department of Housing and Community Affairs v. Inclusive Communities Project, Inc.[1] The majority opinion acknowledges the burden-shifting framework that HUD codified in its 2013 Rule. It also recognizes that disparate impact theory prevents discriminatory ordinances, and plays a role in uncovering discriminatory intent. The ruling’s holding affirmed that disparate impact claims are cognizable under the Fair Housing Act, and took no issue with the burden-shifting framework.

Now that eight years have passed since the adoption of the 2013 rule, the housing industry has become very familiar with its terms. Many in the industry, including the National Association of Realtors, have endorsed the use of disparate impact theory, and the specific framework codified by the 2013 rule. The current disparate impact standard guides housing providers and lenders towards creative methods that are inclusive of qualified renters and lenders who belong to protected classes.

HUD is correct to set aside the 2020 Proposed Disparate Impact Rule, which was misguided, and deeply flawed.

In 2020, HUD proposed a new disparate impact rule that overhauled the 2013 framework. The rule featured a five-part framework. The first element required a plaintiff to show that the challenged policy or proposal is “arbitrary, artificial, and unnecessary.” The second element stated that the plaintiff must show a “robust causal link” between the policy or practice, and the disparate impact itself. The third element required a plaintiff to show that the challenged policy or practice has an adverse effect on members of a particular class. The fourth element required a plaintiff to show that the disparity is significant, and the fifth required a plaintiff to demonstrate that the complaining party’s alleged injury is directly caused by the policy or practice.

This framework puts an extraordinarily heavy burden on plaintiffs. Before plaintiffs even had a chance to get to the discovery stage, they had to meet all of the above requirements. The first element – showing that a policy is “arbitrary, artificial and unnecessary” – is itself an exceptionally difficult burden to satisfy before trial.

The “robust causal link” requirement also places an undue burden on a plaintiff. The 2020 rule stated that when statistical data is used to demonstrate a disparity, the plaintiff must show that the challenged policy is the “actual cause” of the disparity. Proving that a particular policy is the cause of any disparity, rather than just a significant cause, is an enormous hurdle, particularly at the pleading stage.

In short, the framework would make it effectively impossible for plaintiffs to successfully bring a disparate impact lawsuit. The rule was presented by HUD as a necessary step to reflect the Supreme Court’s Inclusive Communities ruling, but nothing in the ruling necessitated a change to the disparate impact framework, or even hinted that a change of any kind was necessary. Subsequent decisions by lower courts recognized that HUD’s 2013 disparate framework was affirmed by the ruling, including Mhany Mgmt., Inc. v. City. of Nassau,[2] and Prop. Cas. Insurers Ass’n of Am. v. Carson.[3]

For these reasons, it is essential that HUD reinstitute the 2013 rule. Disparate impact theory is essential to remove policies that prevent Americans of color from owning homes and building wealth, or from living in the neighborhood of their choice. There is a massive racial wealth gap in America, with the average White American holding roughly ten times as much wealth as the average Black American, and the average Latino American. This gap cannot be eliminated without the availability of crucial fair housing enforcement tools such as disparate impact theory.

Additional guidance would offer clarity on the use of predictive underwriting techniques.

In certain ways, US regulators have not moved with the same speed or scope as have regulators in Europe, where active supervision and enforcement regimes have been explicitly expressed for the deployment of artificial intelligence (AI), machine learning (ML), and the new alternative data common to their usage. For example, the European Union’s proposed framework represents an effort to provide granular standards that define key updates for consumer protections.[4]

In addition to the framework laid out in HUD’s 2013 disparate impact rule, NCRC also urges HUD to consider providing separate guidance related to the use of AI, ML and alternative data to reduce the risk of discrimination.

This additional guidance should include:

  • Clear language stating that the use of predictive underwriting technologies is subject to disparate impact claims. These technologies include (but are not limited to) algorithms, artificial intelligence, machine learning, and/or alternative data for the underwriting of mortgages and other home loans.
  • A clarification/restatement that other uses of predictive underwriting technologies, including but not limited to marketing[5] or tenant screening,[6] are also subject to disparate impact claims.
  • Guidance on determining standards for practical significance (i.e., determining when there is a meaningful difference in how groups are affected by a policy or practice). This could include guidance on which factors should be considered when establishing standards for evaluating practical significance, what types of outcomes will be considered, and whether factors should be applied differently for different product types.
  • Input on how lenders should select, monitor and review data inputs used for algorithmic modeling. This guidance should address the level of scrutiny not only before input but also while in-process and post-process. The guidance should also require lenders to document what testing has been conducted and what alternatives were considered, and verify that data used for testing and training models are representative. These steps are important to minimize the risks related to sample bias, overfitting and model drift.
  • Language affirming that profit alone cannot justify a practice that causes a disparate impact. A creditor practice that causes a disparate impact can only be justified if it meets a “substantial, legitimate, and non-discriminatory interest” that cannot be reasonably achieved through any alternative and less-discriminatory method.
  • A requirement that models should be explainable. If lenders cannot determine whether inputs (or combinations of inputs) have created proxies for protected classes, then they cannot reasonably safeguard against disparate impacts. If lenders cannot explain decisions made by their models, then they similarly cannot safeguard against disparate impacts. HUD should convey to lenders that while advice from third-party subject matter experts can be helpful, the use of a third-party service to build or explain a model does not shield a lender from disparate impact liability.
  • Guardrails related to the self-testing privilege that will encourage lenders to enhance their proxy methodologies. HUD should clarify the extent to which lenders can use demographic data for self-testing, as long as that data is not included in marketing, underwriting, servicing, fraud protection or other credit-related decisions.
  • A commitment that HUD will apply all approaches to enforcement, including issuing a Matters Requiring Attention, non-public enforcement actions, referrals to the Department of Justice, or public enforcement actions. As a component of such approaches, HUD should make use of a testing data set, either under its own facility or through a partnership with the Federal Financial Institutions Examination Council, to implement enforcement. The data set should be updated regularly to protect against general shifts in population as well as for model drift in the systems utilized by industry.

These recommendations are consistent with comments NCRC submitted on July 1, 2021, in response to interagency request for information on the use of artificial intelligence[7] (“Request for Information and Comment on Financial Institutions’ Use of Artificial Intelligence, including Machine Learning”) and our December 2020 response[8] to a request from the CFPB to provide additional guidance on how it should supervise disparate impact concerns under the Equal Credit Opportunity Act (ECOA) related to the use of AI and ML. Our response to the CFPB, included a letter from NCRC’s Innovation Council which includes six prominent fintech lenders addressing the need for additional guidance.[9]


HUD’s 2013 disparate impact framework is far superior to the 2020 framework, which was deeply flawed. NCRC entirely supports HUD’s decision to reinstitute the 2013 framework. This step will affirm HUD’s commitment to a strong disparate impact standard, and send a clear message that the 2020 framework has been rejected. The additional guidance we suggest would enhance the disparate impact framework for the use of AI, ML, and alternative data.

NCRC and our more than 600 member organizations appreciate the opportunity to share our views on this proposed rule. If you have any questions or need additional information regarding our comment, please do not hesitate to contact Brad Blower, General Counsel at bblower@ncrc.org.


National Community Reinvestment Coalition (NCRC)

Affordable Homeownership Foundation Inc.

African American Chamber of Commerce of Montgomery County

African Diaspora Directorate

Alpha Phi Alpha Homes, Inc.

California Reinvestment Coalition

CASA of Oregon


Chicago Community Loan Fund

Community Reinvestment Alliance of South Florida

Cornerstone West CDC

Delaware Community Reinvestment Action Council, Inc.

Fair Finance Watch

Georgia Advancing Communities Together, Inc.

Home Ownership Center if Greater Cincinnati

Housing Action Illinois

HousingWorks RI

Jewish Community Action

Justine Petersen

LINC UP Nonprofit Housing Corporation

Low Income Housing Institute

Making Change, Inc

Massachusetts Affordable Housing Alliance

Metropolitan St. Louis Equal Housing and Opportunity Council

Multi-Cultural Development Center

National Association of American Veterans, Inc.

National Housing Resource Center

NC Housing Coalition, Inc.

Neighborhood Improvement Association

Northwest Indiana Reinvestment Alliance

Real Estate Education and Community Housing Inc.

Sandhills Community Action Program


Southern Dallas Progress CDC


Working In Neighborhoods

  1. Texas Department of Housing and Community Affairs v. Inclusive Communities Project, Inc., 576 US 519 (2015).
  2. Mhany Mgmt., Inc. v. Cty. of Nassau, 819 F.3d 581, 618 (2nd Cir. 2016).
  3. Prop. Cas. Insurers Ass’n of Am. v. Carson, 2017 WL 2653069 (N.D. Ill. June 20, 2017) at 8.
  4. European Union. Proposal for a Regulation Laying Down Harmonised Rules on Artificial Intelligence (Artificial Intelligence Act) and Amending Certain Union Legislative Acts (2021). https://digital-strategy.ec.europa.eu/en/library/proposal-regulation-laying-down-harmonised-rules-artificial-intelligence.
  5. Cty. of Cook v. HSBC N. Am. Holdings Inc., 314 F. Supp. 3d 950, 955 (Northern District of Illinois, Eastern Division May 30, 2018).
  6. Connecticut Fair Housing Ctr et al v. CoreLogic Rental Property Solutions, LLC, 478 F. Supp. 3d 259 (D. Conn. 2020).
  7. National Community Reinvestment Coalition. “Request For Information and Comment on Financial Institutions’ Use Of Artificial Intelligence, Including Machine Learning.” Comment Letter, July 1, 2021. https://ncrc.org/request-for-information-and-comment-on-financial-institutions-use-of-artificial-intelligence-including-machine-learning/.
  8. National Community Reinvestment Coalition, “NCRC Comment on The CFPBs Request For Information On The Equal Credit Opportunity Act.” Comment Letter, December 1, 2020. https://ncrc.org/ncrc-comment-on-the-cfpbs-rfi-on-the-equal-credit-opportunity-act/.
  9. National Community Reinvestment Coalition, Lending Club, Affirm, Varo Bank, Oportun, PayPal, and Square. “NCRC, Fintechs Call on CFPB To Clarify Applying Fair Lending Rules To Artificial Intelligence.” National Community Reinvestment Coalition, June 29, 2021. https://www.ncrc.org/ncrc-fintechs-call-on-cfpb-to-clarify-applying-fair-lending-rules-to-artificial-intelligence/.
Print Friendly, PDF & Email

Redlining and Neighborhood Health

Before the pandemic devastated minority communities, banks and government officials starved them of capital.

Lower-income and minority neighborhoods that were intentionally cut off from lending and investment decades ago today suffer not only from reduced wealth and greater poverty, but from lower life expectancy and higher prevalence of chronic diseases that are risk factors for poor outcomes from COVID-19, a new study shows.

The new study, from the National Community Reinvestment Coalition (NCRC) with researchers from the University of Wisconsin–Milwaukee Joseph J. Zilber School of Public Health and the University of Richmond’s Digital Scholarship Lab, compared 1930’s maps of government-sanctioned lending discrimination zones with current census and public health data.

Table of Content

  • Executive Summary
  • Introduction
  • Redlining, the HOLC Maps and Segregation
  • Segregation, Public Health and COVID-19
  • Methods
  • Results
  • Discussion
  • Conclusion and Policy Recommendations
  • Citations
  • Appendix

Complete the form to download the full report: