By Ed Kramer and Luke Wimer
While this shift is understandable, it carries with it some significant risks. Indeed, with regulators increasingly focusing their attention on fair lending, AI decisioning, and mortgage loan servicing, mortgage originators and servicers now more than ever must take steps to maintain, back check, and update their fair lending policies, procedures, and record keeping.
Fair lending concerns, old and new, trended up at the end of 2019. Redlining concerns, supposedly obliterated decades ago, reemerged as a hot topic. Lawmakers, in February 2020, formally expressed concern that the use of artificial intelligence (AI) tools in the lending process could result in disparate impact and disparate treatment, even if unintended. CFPB actions and reports in April and May (2020) reflected not just this focus on AI and redlining, but increasing complaints about servicers and a formal action against a West coast-based servicer which may presage increasing attention to an area where we can expect significant loan modifications or workouts in the coming months.
First, do no harm to existing compliance
Lenders and compliance staff have adopted new practices to accomplish social-distancing and layered on a new wave of forbearance. In this new operating environment, there is a key question that warrants attention: Do stay-at-home orders, potentially uneven use of online connectivity, or a rapidly shifting market put your existing controls in jeopardy? Savvy managers will document ongoing compliance, as well as any operational adjustments. Regulators are issuing COVID-19 letters and advisories regarding lending and loan servicing that require concentrated efforts. Lenders have to adapt quickly to these changes and expand their thinking about fair lending to encompass new technology and servicing activities under the new regulatory landscape.
Redlining, outlawed 40 years ago, unfortunately remains a reality to this day. In addition to enforcement and legal actions initiated against lenders, redlining – prior to the onset of the COVID-19 crisis – was emerging as a renewed area of focus. One needs only to conduct a quick internet search for recent examples:
- Last year, the Department of Justice settled a suit against an Indiana bank to resolve allegations that it engaged in redlining-type lending discrimination.
- In late 2019, an office of the Federal Reserve sponsored a “Fair Lending Interagency Webinar,” with two of its six topics specifically calling out the practice of redlining.
- In late 2018, the Pennsylvania Attorney General issued a broad request to mortgage borrowers to file redlining complaints with his office, prompted by an independent study of discriminatory mortgage lending in the region.
Furthermore, at least five states have established “mini-Consumer Financial Protection Bureaus” within their attorney general’s offices, ready, willing, and able to tackle redlining issues.
The Supervision, Enforcement and Fair Lending division of the CFPB continues to emphasize fair lending and redlining among their top priorities. Look-back periods in some recent examinations have run up to seven years, making it important to examine historical as well as current trends in an organization.
Opaque algorithms, subconscious bias, “dirty data”
Many and varied observers have expressed urgent concern that AI has the potential for unintentionally negative consequences. This is because well-intended business innovation can quickly outpace the ability to monitor compliance.
Just as one example, the Vatican, of all entities, in a doctrine titled “The Rome Call for AI Ethics,” put it this way: “AI must be developed with a focus on protecting and serving humanity, and that all algorithms should be designed by the principles of transparency, inclusion, responsibility, impartiality, reliability, security, and privacy.” It adds that the decisions made by algorithms need to be explainable, transparent, and fair.
All of which remarkably echoes a summary of particular interest to lenders, by the House Committee on Financial Services, dated Feb. 7, 2020, aptly titled “Equitable Algorithms: Examining Ways to Reduce AI Bias in Financial Services.”
It neatly sums up the potential areas of fair lending concern associated with AI: the inability to explain how a particular algorithm makes decisions; subconscious bias that may be imparted to an algorithm by its author; and biased decisions made by the algorithm based on faulty data.
Just a few excerpts from this document illustrate these points:
“Generally, the complexity of decision-making processes utilized in these technologies makes it difficult for human programmers to predict what the program will do and explain why it did what it did.”
“[H]uman programmers may unknowingly write historical biases into their programs.”
“Data sets can contain errors…Substandard data or ʻdirty data’ is problematic for AI programs because, all following inferences are susceptible to inaccuracies, incompleteness, or non-representative information.”
The federal government still is grappling with this. Witness an October 16, 2019 letter from the head of the Federal Trade Commission (FTC) to the Secretary of Housing and Urban Development (HUD), objecting to HUD’s proposed rule to amend the Fair Housing Act’s discriminatory effects standard. It says in part:
“First, algorithms are not neutral, and even valid inputs can produce discriminatory results. Second, it is inappropriate to create safe harbors around technologies that are proprietary, opaque, and rapidly evolving. Finally, outsourcing liability for algorithmic discrimination to third parties distorts incentives and could lead to a race to the bottom among vendors.”
As of this date, HUD has yet to make its rule final. The FTC issued, in early April, a lengthy statement regarding the use of AI, summarized as follows: “The use of AI tools should be transparent, explainable, fair, and empirically sound, while fostering accountability.”
So it’s clear that any resolution of this issue is still to come.
COVID-19 changes open new battle-fronts
Even as redlining and AI fair lending issues were starting to come to a boil, the pandemic hit hard. Customers lost jobs, couldn’t pay bills, and fell behind on loan payments.
Federal and state banking regulators responded with a series of joint statements, starting March 9, 2020. That one urged financial institutions to “work constructively with borrowers and other customers.” On March 19, 2020, another joint release clarified that financial institutions will receive Community Reinvestment Act (CRA) consideration for qualifying community development activities. That in turn led to a third statement “encouraging responsible small-dollar lending in response to COVID-19.” Other similar communications also were issued.
Then, the Congress passed and the President signed the CARES Act into law on March 27, 2020. The CARES Act enables borrowers with residential mortgage loans backed by the federal government − FHA, VA, USDA, Fannie Mae, and Freddie Mac loans − to request from their lenders forbearance for up to one hundred eighty (180) days and provides for one additional extension of up to another one hundred eighty (180) days (provided such is requested during the “covered period”). Before or at the end of the forbearance period, loan terms must be modified to allow for the options worked out with borrowers, such as a lump sum payment at the end of the forbearance period, an extension of the loan period where the missed payments are tacked onto the end of the mortgage, or adjusting the monthly payments to make up the deferred payments over time. Foreclosures are forbidden on these loans until June 30, 2020, but at that time some borrowers may be unable to resume payments and will need to enter into some form of workout or foreclosure proceedings.
At any one of these stages, servicers will want to ensure fair and equal treatment of their customers. Fair servicing has gotten relatively little examination focus historically, but is increasingly discussed by regulators. Servicers should anticipate that behaviors during this large wave of loan modifications or workouts will be subject to high levels of scrutiny. Internal compliance departments monitoring fair lending may need to turn more of their attention to whether customers are getting equal treatment on loan forbearance and modifications and receiving equitable levels of early intervention and contact or options on payment deferrals, loss mitigation, or foreclosure sale activities once the suspension of foreclosure activity comes to an end on June 30, 2020. Servicers will want evidence that they have applied standards and policies fairly across their portfolios. Clarifying these policies and procedures before the coming wave of loan modifications and foreclosure activity is, therefore, of paramount importance.
It all sounds good, but how does this work out in today’s reality?
Forbearance arrangements will include a plan for repaying the missed monthly payments. The options usually include paying in one lump sum, making multiple payments, or modifying the mortgage. There could also be an option to recast the loan with monthly payments remaining the same but extending the term.
Most loan servicing people are no longer working in a facility with supervisors or colleagues by their sides. They are working remotely, probably in cramped or repurposed rooms in their homes, full of distractions. Internet and phone connections may not be the most efficient (or secure). In any case, it probably is not easy for such people to ask questions or get decisions on specific loan modification requests, the number of which undoubtedly has skyrocketed.
It’s a given that these people do not mean to discriminate against any protected class of customers, but mistakes happen and backup oversight by senior lending officers can fail. Human nature being what it is, there can be a reluctance, case by case, to either grant a loan modification or deny one out of fear of not making that decision the “right” way.
Even before the COVID-19 era, fair lending compliance loomed large. The most effective industry stance was to find potential violations in-house as soon as possible, fix them, and if material, report the matter.
With a higher potential for mistakes to happen, and with the near certainty that someday, post-COVID-19, regulators and inspectors will comb back-files for infractions, what lenders need to do is install and/or beef up their technological and review systems. Such systems subject loan modification and servicing activities to tests that can reveal potential fair lending violations quickly, allowing for prompt mitigation. Ongoing monitoring and testing is the best safeguard against inappropriate behavior and regulatory action.
Even in these confusing, threatening, and perilous days, lenders cannot afford to let fair lending violations slip through the cracks.