Ed. #106: Robust Causality & Fair Lending
What’s a mortgage compliance officer to do now?
On a recent mortgage industry podcast, I was asked what mortgage originators should be doing today to be compliant with fair lending. I must admit, even after sitting through multiple fair lending panels at the MBA’s Legal Issues and Regulatory Conference (LIRC)[1] earlier this month with outstanding insights from terrific speakers, I had trouble providing an off the cuff, cogent, and condensed answer to that question because, frankly, it’s complicated.
Notably, it is not particularly helpful to remind lenders to never intentionally discriminate based on prohibited characteristics. That is still illegal and, moreover, it is an own goal caliber mistake for business and reputation. No doubt, however, I can do better with my insights than that. While, as always, this Musing is not legal advice and people and companies should get their own legal counsel to obtain customized guidance, I will attempt to provide a better perspective on the fair lending state of play.
While it’s true that prudent mortgage originators should “stay the course” and just keep doing what they have always been doing (i.e., train and maintain policies to not discriminate and treat everyone fairly) what can or should metrics-driven mortgage compliance folks do differently to ensure fair lending compliance in light of: (i) the current prohibitions on enforcement’s use of disparate impact/effects testing posed by (a) Trump’s DEI Executive Orders (EOs), and (b) the April 2026 ECOA/Reg B regulatory amendments, (ii) the questionable durability of these Trump initiatives arising from future political winds shifting back, and (iii) the threat of expanded state and private fair lending initiatives? What kind of compliance program is needed for mortgage fair lending confidence in this new environment?
Recent history
First, however, a reminder of some recent fair lending happenings. On November, 1, 2024, Townstone Financial and the CFPB announced a settlement of the fair lending case originally filed at the tail end of Kathleen Kraninger’s CFPB in 2020. Two days after hearing news of that settlement (right before Trump’s re-election), in a Musing, I suggested that the Townstone settlement would be remembered as the “peak of fair lending enforcement”. At the time, by “peak” I was referring to the legal underpinnings of fair lending enforcement under the Fair Housing Act and ECOA. As I noted then, “the Townstone settlement is an unequivocal recognition by CFPB that its overzealous fair lending interpretations are unlikely to survive judicial challenge.” As it turned out, the Townstone settlement itself had to survive Trump’s CFPB’s judicial challenge to vacate it after Trump’s reelection. But those old fair lending interpretations themselves still have not yet faced judicial challenge.[2]
Fast forward past Trump’s reelection in November 2024 to the eve of last year’s LIRC in April 2025. I then wrote about disparate impact’s demise and Trump’s EOs and CFPB/Justice Department staffing changes dooming federal fair lending enforcement, at least for the rest of his term. Meanwhile, states, like New Jersey and New York, took those developments as an affront to their interpretation of fairness in lending and adopted rules and interpretations expressly recognizing disparate impact. Then, last month, the CFPB issued a final rule under Reg. B amending ECOA’s regulation to eliminate disparate impact liability and, among other things, limiting discouragement claims to affirmative actions, rather than omissions.[3]
Robust causality?
Despite all of the above, disparate impact, as I noted in Musing #92, is only “mostly dead”. It is still recognized under the Fair Housing Act (FHA)[4] pursuant to Justice Kennedy’s 5-4 majority opinion in the Supreme Court’s 2015 Inclusive Communities case.[5] Kennedy’s opinion sought to balance the tensions posed by disparate impact being used to shift the burden of proof about discriminatory intent by requiring courts find a “robust causality”[6] between effects and discriminatory conduct. [7]
But, the redlining enforcement actions of the Biden era demonstrated that fair lending enforcers could bypass judicial review entirely through the settlement process’s costs of defense and the reputational damage posed by mere allegations of discrimination based on disparate impact. Absent a court’s review at the onset of allegations, enforcers could look solely to HMDA data and treat the need for robust causality like Federale badges in the Treasure of Sierra Madre: “Robust causality? We don’t need no stinkin’ causality!”).
The lesson of the Biden enforcement era is that Kennedy’s robust causality requirement is, in practice against a government prosecution, a right without a remedy.[8] Until Inclusive Communities is expressly overturned or narrowed, settlement pressure will continue to foreclose access to court review of the causality test in any future fair lending enforcement action.
Nuances and burden shifting
Given all that, the legal and practical reality today is that fair lending compliance is significantly more nuanced than it was when you (thought?)[9] you could just objectively look at your lending patterns and determine whether you were in line with your peer groups to establish compliance.[10] Lending pattern analysis is bound up with arguments about structural racism and critical race theory assumptions about causes and effects which are endlessly debatable.[11] Shifting the burden of proof was the name of the game for proponents of disparate impact theory on the grounds that if the results are disparate, the burden is on the lender to prove that it was not due to illegal discrimination.[12]
So, fair lending compliance management has essentially been about the ability to disprove a negative: i.e., prove the absence of discrimination. That is, whether your lending is “fair” arrives in the form of an extremely loaded question.[13] But now, after the Reg. B amendment, even the use of remedial or well-intentioned programs such as special purpose credit programs (SPCPs) could actually get you in trouble for reverse discrimination. While you still have state law driven disparate impact questions to contend with in many “blue” states,[14] and the prospect of future federal enforcement pendulum swings, demonstrating fair (non-discriminatory) intent also remains elusive.
Where are we now?
A recent (2024) mathematically-grounded University of Chicago paper looking at the relationship of underwriting criteria to racial outcomes concluded (on pg. 7), “innovations that improve validity, by better aligning underwriting decisions with borrowers’ ability to repay, can also reduce loan approvals for some applicants, increasing disparities in outcomes.” In other words, quite unsurprisingly, underwriting standards can cause racial disparities and easing those standards to improve racial balance only leads to more bad borrower outcomes.
This should not come as a shock to anyone familiar with the persistent racial homeownership gap. This is why 50+ years of fair lending enforcement against mortgage originators has not made a dent in that gap: discrimination by mortgage originators (intentional or effects-based) is not the problem. Regardless of the legal questions about whether lending patterns alone should be actionable, however, the debate over whether lending patterns prove or validate fairness in lending is not going end with the current environment. As I have said many times in these Musings, fairness is in the eye of the beholder and thus is not measurable through lending patterns. This leaves compliance officers with a practical rather than legal problem.
Know your narrative
Proving intent is incredibly challenging and compliance officers are not thought police who can run around finding guilty minds. Still, nondiscriminatory intent can be demonstrated though a consistent nondiscriminatory narrative. That is, if your business model or loan product design (underwriting) results in racially imbalanced results, be prepared to explain why that is the case (presumably it has nothing to do with intent to discriminate).[15] Also, have a firm understanding of your employment numbers and marketing outreach to support your business model specifically, not just to reach a general audience.
As with RESPA compliance, fair lending compliance can be driven by your fair lending narrative, and your data needs to support your narrative. That said, when the enforcement pendulum swings, until a court is tasked with assessing robust causality, knowing your narrative allows you to make the best case you can to people (who may not be listening) and hoping, as the compliance officer's version of faith, that being demonstrably one of the “good guys” still counts for something.
[1] The MBA LIRC is always a highlight of my year and never disappoints for information and connecting with peers. Special shout out and thank you to Waterstone’s Stephanie Ziebell whose heartfelt expression of gratitude on LinkedIn touched me deeply.
[2] It would be enormously helpful to fair lending jurisprudence and mortgage compliance management if the issue of disparate impact under the Fair Housing Act were to be revisited by SCOTUS in light of recent discrimination related cases involving voting rights and college admissions, but the right vehicle(s) and driver(s) remain elusive. In that regard, based on their respective CEO’s penchant for pushing back against government overreach, I double dog dare a would-be fair lending enforcer and/or plaintiff to make a disparate impact-premised case against Rocket Mortgage or Chase.
[3] I may return to the topic of discouragement in future post, but for now, I think most of the concerns about discouragement under ECOA I raised in earlier Musings have been resolved by the Reg B amendment (assuming courts agree with the amended rule in a post-Chevron legal interpretation landscape).
[4] Home mortgage lenders are subject to both ECOA and the Fair Housing Act.
[5] Texas Dept. of Housing v. Inclusive Communities Project, 576 U.S. 519 (2015). As I noted in footnote #21 to Musing Ed. 92, I could not locate the Inclusive Communities decision on the Supreme Court’s website, but I was able to find it on the Justice Department’s website. Note to the next Attorney General: You might want to ask someone why that is posted.
[6] Robust causality? Is that like “Grave danger”? “Is there another kind?” Or to use an old law school lesson, “Gross negligence is just negligence with a vituperative epithet.”
[7] Kennedy wanted to ensure disparate impact doesn’t become a strict liability regime based solely on outcome disparities and trusted the courts, but not enforcers, to make that determination.
[8] A due process argument about the absence of judicial gatekeeping might be structurally interesting, but faces the classic consent-order problem: banks that signed voluntarily arguably waived the hearing they never got.
[9] Perhaps lending pattern analysis provides a false sense of security for fair lending claims based on intentional discrimination. Robust causality also prevents defense of an intentional discrimination claim based on lack of discriminatory impact.
[10] Peer group lending pattern analysis is costly for sure, and, while less objective than might appear (due to how peer groups are established) is more data driven than divining the discriminatory “intent” of an organization.
[11] Footnote 11 of the U of C paper mentioned infra summarized the debate over use of disparate impact in fair lending as follows: “Disparate impact claims are debated over whether they primarily address discriminatory effects or serve as a proxy for intent. The intent-based view sees unjustified effects as evidence of hidden motive, while an effect-based approach focuses on remedying harm regardless of intent.” See also, the discussion of critical theory and “cognitive dissonance” used to quell debate in this Musing edition about the CFPB’s Covid-era servicing rule proposal.
[12] Trump’s EOs (and a recent Orrick webinar) referred to a “nearly insurmountable presumption of unlawful discrimination.”
[13] Like, “when did you stop beating your wife?”
[14] And look who is back running the consumer protection show in California now…,
[15] For example, if you have bilingual originators you may end up with a greater percentage of borrowers who do not speak English.