Monday's edition of the Federal Register featured a proposed rule that, depending on your viewpoint, would amend the Department of Housing and Urban Development's (HUD's) interpretation of the Fair Housing Act's disparate impact standard, (HUD) or "make a major change to a well-settled standard on how the agency and the courts review claims of discrimination under the Fair Housing Act of 1968" (Urban Institute.)

The announcement from HUD says its proposed rule would provide more appropriate guidance on what constitutes unlawful disparate impact to better reflect the Supreme Court's 2015 ruling in Texas Department of Housing and Community Affairs v. Inclusive Communities Project, Inc.

The court ruled that "disparate impact" is one arising out of an action that is not necessarily discriminatory on its face, but none-the-less has an outsized impact on one or more protected classes under the Fair Housing Act (FHA). In the Texas case, the plaintiffs held that the Texas agency responsible for administering housing tax credits disproportionately allocated too many of them into predominantly black inner-city areas and too few to predominately white suburban neighborhoods. The ultimate impact was to restrict the ability of black Texans to live in predominately white areas.

The majority opinion, written by Justice Anthony Kennedy, held that Congress specifically intended for disparate impact claims to be included in the FHA but that the burden is on the plaintiff to prove it is the defendant's policies that cause the disparity.

Solomon Greene and Martha M. Galvez, writing in the Urban Institute (UI) blog said, "Over several decades, courts have worked out a straightforward process for evaluating claims of disparate impact, which HUD standardized and clarified in a 2013 rule.  Under the existing rule, courts engage in a three-step process of "burden-shifting" that recognizes legitimate business interests while guarding against discriminatory outcomes.

The new rule will replace that with a framework for establishing legal liability for apparently neutral practices that may have such unintended discriminatory effects and has no impact on acts of intentional discrimination. That framework shifts the burden of proof to the plaintiff by requiring five steps.

  • The plaintiff must plead that the practice or policy being challenged is arbitrary, artificial, and unnecessary to achieve a valid interest or legitimate objective.
  • The plaintiff must then allege a "robust causal link" between the challenged policy or practice and the claimed impact on members of a protected class.
  • The plaintiff must allege that the policy has an adverse effect on protected members
  • That the disparity is significant.
  • And that the alleged injury is directly caused by the challenged policy.

The rule states that "Plaintiffs will likely not meet the standard, and HUD will not bring a disparate impact claim, alleging that a single event-such as a local government's zoning decision or a developer's decision to construct a new building in one location instead of another-is the cause of a disparate impact, unless the plaintiff can show that the single decision is the equivalent of a policy or practice."

The proposed rule is raising alarm bells in several quarters. The Urban Institute says, "The proposal threatens to halt 50 years of progress toward fair and equal access to housing and credit, with sweeping implications for people of color, women, families with children, and people with disabilities. The protections against discrimination enshrined in the Fair Housing Act have curbed some of the worst abuses over the past 50 years. But discrimination persists in housing and lending markets."  UI also notes that in today's environment, an implicit or unintentional act of discrimination may be more common and harder to detect.

However, we found at least three sources who are concerned much of the negative effects of the rule may be borne by an unlikely group, software vendors. UI points out that algorithms are increasingly used in nearly every facet of housing and lending, including, tenant screening, appraisals, code enforcement, automatic underwriting. And they are increasingly incorporating a broader array or what would appear to be neutral inputs such as data from social networks.  

The new rule provides that where a plaintiff identifies an offending policy or practice that relies on an algorithmic model, the defending party, i.e. the landlord, or mortgage lender may defeat the claim by identifying the inputs used in the model, showing those inputs are not substitutes for a protected characteristic and the model is predictive of a risk or other valid objective and that a recognized third party, not the defendant, is responsible for creating or maintaining the model.

In other words, it appears that the landlord or mortgage lender can escape any disparate impact claims by proving he or she bought software off the rack and it is the algorithms that discriminated.

Andrew D. Selbst, writes in Slate that the new rule "would radically reinterpret anti-discrimination law to give landlords and mortgage lenders a recipe to discriminate by algorithm." Algorithms are widely used in housing, he says, but those predictions "are built on incomplete data and past decisions by humans, reproducing their past biases." He adds, "Malicious actors can easily devise algorithms that functionally redline."

Publication of the Rule has opened a two-month period in which HUD is inviting comments.  That period expires on October 19.