With President Biden's November 2020 election, it appears that the four-year long "regulatory winter" of federal fair lending enforcement activity has come to an end. Indeed, with the appointments of Mr. Rohit Chopra as Director of the Consumer Financial Protection Bureau ("CFPB"), Ms. Lina Khan as the head of the Federal Trade Commission ("FTC"), Ms. Kristen Clarke as Assistant Attorney General for Civil Rights at the U.S. Department of Justice ("DOJ"), and Mr. Eric Halperin as the CFPB's Assistant Director for the Office of Enforcement, the Administration has put in place a senior regulatory leadership team that will actively work to fulfill its commitments to increased consumer protection, racial equity, and social justice through the types of aggressive examination and enforcement activities previously experienced in the Obama years (see charts below).
With the first year of the Biden Administration focused on the installation of a new, more aggressive regulatory leadership class, 2022 should be the start of a multi-year regulatory renewal in which these new leaders ramp up their supervision and enforcement activities to achieve Administration fair lending and consumer protection goals.
So, what are the likely fair lending focal points of these federal agencies? Based on public pronouncements, recent agency activities, a scan of the current fair lending landscape, and a bit of speculation, the following represent - in no particular order - the six areas where we may see increased fair lending scrutiny in the months ahead.
1. Redlining
Redlining has long been a top focus area of federal financial regulators - and expect 2022 to see supervision and enforcement activity rise to a new level. In fact, the DOJ announced in October 2021 the launch of a new Combatting Redlining Initiative that seeks to expand identification and prosecution of alleged redlining activity through:
• Expanding the DOJ's coordination with U.S. Attorney's offices nationwide to provide a surge of enforcement resources and local-market expertise.
• Expanding the focus of enforcement activity to include non-bank mortgage lenders.
• Strengthening relationships with federal financial regulatory agencies to ensure that identified redlining violations are appropriately referred to the DOJ.
• Increasing coordination of fair lending enforcement activity with State Attorneys General.
Non-bank mortgage lenders should particularly assess the adequacy of their Compliance Management Systems ("CMS") in this risk area due to the increased focus of the federal and state financial regulators on non-banks. Additionally, all lenders may wish to consider the following recommendations when evaluating the adequacy of their risk and control framework:
• Include all relevant geographic markets into your redlining monitoring process - not just your "primary" market, or only market(s) where you have physical locations.
• When considering strategic changes to your institution's geographic footprint - such as branch openings / closings, bank acquisitions / divestitures, and entering new markets, analyze the impact of such potential changes on your geographic lending patterns to assess potential impacts to redlining risk.
• Be sure to include an assessment of your institution's sales, advertising, and marketing activities into your redlining risk assessments. Targeted advertising campaigns, non-diverse visual representations of customers, non-diverse sales personnel, and non-diverse media channels may all contribute to redlining risk through potential "discouragement" of prohibited basis customers.
• Annual HMDA data plays a critical role in benchmarking your institution's geographic lending patterns against "peer" institutions. As the 2021 HMDA data should be released in June 2021, it would be prudent to update redlining analyses with such data as quickly as possible to identify emerging risks.
• Statistical analysis of geographic lending patterns is a critical component of redlining risk assessment. Beyond the standard majority-minority vs. majority-majority census tract lending comparisons, lenders should dive deeper into the majority-minority census tracts to see whether lending activity is actually focused on minority borrowers.
2. Algorithmic Bias / Robo-Discrimination
The rapid deployment of artificial intelligence ("AI") and machine learning ("ML") technologies - along with alternative data - to the consumer lending area is based on the promise of significant benefits such as speedier decisions, improved credit risk management, more consistent customer treatment, expanded financial access, and improved racial equity. However, these benefits also come with potential risks and costs to consumers that are still not yet well understood, and whose potential mitigations expose lenders to high levels of regulatory and legal risk.
Federal financial regulatory agencies clearly have their sights set on potential algorithmic bias as an emerging fair lending risk and are actively exploring the various consumer protection risks associated therewith. While testing lending-related algorithms for potential disparate treatment and disparate impact is considered an important component of an effective CMS today, there are - unfortunately - still a number of important open questions that need to be addressed by the financial regulators to truly advance consistent and effective compliance frameworks across the industry. Indeed, the following represents just a sample of these open issues:
• If the algorithm's scores or decisions are shown to be free from disparate impact, is it necessary to assess disparate impact at the individual variable / input level? This question is particularly relevant for AI/ML algorithms that contain hundreds (or even thousands) or variables.
• What is the appropriate fairness metric to evaluate (there are a multitude of possible metrics)? Should fairness be evaluated at the overall group-level or on an individual-level basis?
• What risks or limitations do common race/ethnicity proxies create for this type of disparate impact testing, and how should these be considered when interpreting test results?
• How specifically does the "less discriminatory alternative" construct apply to an algorithm's potential disparate impact?
• If disparate impact is detected, can the use of demographic data to "de-bias" the algorithm inadvertently create an illegal disparate treatment issue? How would safety-and-soundness examiners view banks' trade-offs of model accuracy for model fairness?
To address these CMS challenges, lenders should ensure they bring together the perspectives and expertise of outside regulatory counsel, fair lending specialists, model developers, and model risk management as these risks are multidimensional and interdependent in nature - thereby requiring a holistic approach from a CMS design perspective.
3. Digital Redlining and Marketing Bias
In general, digital redlining refers to targeted on-line marketing or advertising activities that result in a disproportionate lack of exposure to one or more prohibited basis consumer groups - thereby effectively discouraging such groups from applying to the institution . For example, if a lender targets on-line ads for mortgage refinances using criteria that either directly or indirectly disfavors Hispanic individuals, then such an ad campaign could be considered a form of digital redlining or marketing discrimination due to reduced Hispanic response rates to such ads (relative to White response rates).
The significant shift of traditional advertising and marketing activities to on-line channels over the last several years, along with the increasingly sophisticated analytical tools and ever-growing alternative data used by on-line advertisers to offer more powerful consumer ad targeting, has made this fair lending risk a growing regulatory concern. Lenders may wish to consider the following recommendations when evaluating the adequacy of their current CMS risk and control framework:
• Evaluate the sufficiency of current policies and procedures to address adequately this fair lending risk, and revise accordingly.
• Ensure that targeted on-line advertising / marketing campaigns are receiving appropriate and timely fair lending reviews and analyses - from both a disparate treatment and disparate impact perspective.
• In general, consumer targeting criteria should be transparent and well-documented. Potential fair lending risks should be evaluated using methods similar to those employed to identify fair lending risks for consumer lending scorecards / algorithms (i.e., through data analytics).
• Pay particular attention to targeting criteria that rely on geography (e.g., zip codes), distance to physical branch / store locations, or facially-neutral attributes that may be correlated with prohibited basis status.
4. COVID-Relief Activities
As part of the expansion fair lending examination activities around small-business lending, expect the federal bank regulatory agencies to examine closely the activities surrounding lenders' Paycheck Protection Program ("PPP") loan originations - including application sourcing, underwriting / processing, and servicing - with a particular focus on whether certain prohibited basis groups (e.g., racial/ethnic minorities or women) were disproportionately under-represented, or received a relatively lower quality of assistance, in all stages of the program's life-cycle. This focus on PPP lending is consistent with the federal financial regulators' overall increased fair lending focus on small-business lending.
From a consumer perspective, given the significant volume of financial relief actions taken by lenders during the pandemic - such as loan forbearance, loan modification, fee waivers, etc., expect these consumer COVID-relief activities to also receive fair lending scrutiny in the coming months.
5. Appraisal Bias
As described in their June 1, 2021 Fact Sheet, the Biden Administration announced its intent to "address racial discrimination in the housing market, including by launching a first-of-its-kind interagency effort to address inequity in home appraisals." This announcement was followed by a July 2, 2021 statement by the CFPB indicating "[t]he Bureau is pleased to participate in this important interagency initiative, and .... we have dedicated additional resources to evaluate tools and approaches to address inequities in home appraisals."
In further support of these claims, Freddie Mac released a Research Note in September 2021 that analyzed data on more than 12 million appraisals for purchase transactions submitted to Freddie Mac from January 1, 2015, to December 31, 2020. Its primary findings were:
• "Appraisers' opinions of value are more likely to fall below the contract price in Black and Latino census tracts, and the extent of the gap increases as the percentage of Black or Latino people in the tract increases. ... For example, 12.5% of the properties in Black tracts receive "appraisal value lower than contract price," compared to 7.4% for those in White tracts, leading to a gap of 5.2%."
• "As the concentration of Black or Latino in a census tract increases, the appraisal valuation gap increases. For example, the gap for properties in Latino tracts increases from 7.7% in the [50%-80%] bucket to 9.4% in the [80%-100%] bucket."
• "An analysis of the group of appraisers with enough observations in both minority tracts and White tracts to yield valid t-statistics reveals that a large portion of appraisers are generating statistically significant gaps."
• "Our preliminary modeling results suggest that even when taking structural and neighborhood characteristics into consideration, a property is more likely to receive an appraisal lower than the contract price if it is in a minority tract."
Although this is still a nascent, emerging fair lending risk area with much work still to be done to build out an appropriate and effective CMS, lenders should begin to work now with compliance, legal, and analytical partners to tackle the key foundational elements (i.e., policies, procedures, risk assessment, training, monitoring / testing, corrective actions, and reporting) that will comprise the developmental roadmap to their future-state CMS.
6. Pricing Exceptions
Pricing discrimination has a long history in federal and state fair lending enforcement activity and, accordingly, most lenders have a well-developed CMS to manage this fair lending risk. However, in the CFPB's December 2021 Supervisory Highlights publication, it was noted that certain weaknesses had been observed in these CMSs that warranted attention. Specifically,
"The examination team identified lenders with statistically significant disparities for the incidence of pricing exceptions for African American and female applications compared to similarly situated non-Hispanic white and male borrowers. Examiners did not identify evidence that explained the disparities observed in the statistical analysis. Instead, examiners identified instances where lenders provided pricing exceptions for a competitive offer to non-Hispanic white and male borrowers with no evidence of customer initiation. Furthermore, examiners noted that lenders failed to retain documentation to support pricing exceptions. Also, lenders’ fair lending monitoring reports and business line personnel raised fair lending concerns regarding the lack of documentation to support pricing exception decisions. Despite such concerns, lenders did not improve the processes or document customer requests to match competitor pricing during the review period."
Based on these observations, (1) lenders should expect a greater focus on pricing exceptions during fair lending examinations if such exceptions are a permissible (and non de minimis) activity of its loan originators, and (2) compliance personnel should re-assess the design and operating effectiveness of their current CMS (e.g., policies, procedures, and controls around price exceptions) - particularly with respect to the following:
• Appropriate price exception limits and approval authorities.
• Contemporaneous documentation of customer price exception requests and reasons for such requests (i.e., other competitive offer).
• For lender-initiated price exceptions, contemporaneous documentation of the exception reason and required approvals. Expectations would be that lender-initiated price exceptions would be limited to service quality adjustments (e.g., missed closing date), compliance requirements (e.g., remediation of potential Truth-in-Lending violation), or pricing error correction.
• Robust fair lending analytical testing of both the comparative frequency, and comparative amount, of price exceptions between prohibited basis and control groups - with appropriate corrective actions.
© Pace Analytics Consulting LLC, 2023.