Home / Fair Lending / Fair Lending Conundrum: Can Capturing Too Much Data Cause Fair Lending Violations?

Fair Lending Conundrum: Can Capturing Too Much Data Cause Fair Lending Violations?

By Alan Bercovitz

All businesses are increasingly driven by data, perhaps none more so than the mortgage industry. The operational challenge for lenders is to efficiently capture, retrieve, and organize the flow of data throughout the mortgage production cycle, from origination, processing, underwriting, closing, post closing, sale, and on to servicing.

Superimposed over the data demands of the mortgage production cycle are those required to remain compliant with a seemingly unending barrage of both new and fluctuating regulations.

Lenders Usually Want to Capture and Retain More Data

The most thorough capturing and ease of retrieving applicant data are generally viewed as helpful in both the operational and regulatory sides of the mortgage business.

On the operational side, handling data properly reduces underwriter touches per file, reduces the number of days to close, avoids repetitive document requests from consumers, ensures that initially approved mortgages qualify to close under the originally disclosed terms, and results in higher customer satisfaction leading to repeat and referral business.

On the regulatory side, capturing and retaining more data on every applicant helps a lender defend Ability To Repay (ATR) claims, fends off UDAAP issues regarding the sufficiency of the underwriting process, prevents consumer complaints to the CFPB since fleshing out more data at the time of application makes it more likely that the consumer will have a positive experience, and avoids false claims and FIRREA violations because when all pertinent data is considered it’s less likely that a defective loan will be manufactured.

When Do Lenders Fear Retaining More Data than Regulations Require?

Many lenders are reluctant to retain certain data that they rely upon to aid their operations in a format that makes it easily analyzed for fear that statistical variances could appear that give rise to potential fair lending violations. While they directly consider this data in their production of mortgage loans, they hesitate to make it part of their easily retrievable database for analysis.

Non-Discretionary Data

All lenders strive to produce every mortgage to meet the applicable Agency guidelines as well as any overlays the lender has chosen to apply. This is the data that the authors will define as “non-discretionary”. When a lender does not treat every applicant in the same manner when considering this non-discretionary data, they leave themselves open to a potential fair lending violation.

Here is a recent example of a mortgage application where a lender could have unwittingly created such a potential violation. The borrower wanted to remove funds from their business to use as part of the down payment for their mortgage transaction. The Fannie Mae guidelines state that “Business assets may be an acceptable source of funds for the down payment, closing costs, and financial reserves when a borrower is self-employed and the individual federal income tax returns have been evaluated by the lender, including, if applicable, the business federal income tax returns for that particular business (non Schedule C).”

While the wording of this guideline could be clearer, the accepted meaning is that because a Schedule C business does not require a separate income tax return, there is no need to examine a business tax return that does not exist. The loan officer submitted the file to underwriting, where the file was rejected because the underwriter claimed that the guideline meant that it was not possible to use assets earned within a Schedule C business in conjunction with any residential mortgage loan. And for good measure, she told the loan officer that Fannie Mae never allows an exception to this guideline!

Thankfully, the loan officer asked for guidance on a mortgage industry discussion board and was advised that the underwriter had a mistaken interpretation. Eventually, the loan officer was able to get the decision reversed and the loan closed, but if the loan officer was new or hadn’t persevered, the loan could have easily been declined. Even though this was simply a case of underwriter error, if the applicant happened to come from a protected class, there could have been a potential claim of disparate treatment if other non-protected class consumers had their files properly underwritten under this guideline.

Discretionary Data

The trickiest situations for both originators and underwriters to analyze are those cases not explicitly covered by a guideline or overlay, which the authors will refer to as “discretionary data”. This data is often the root of a reason for denial, as loan officers learn hard lessons after they originate and send a file to underwriting that they can rightfully claim meets all of their company’s published guidelines, yet still runs afoul of one of the “off the books” rules that our industry typically teaches to loan officers in the same experiential manner in which wilderness tribal groups pass lessons onto their young as each unique situation presents itself.

One example of a discretionary situation is a borrower who has received AUS approval on a conventional loan for an owner-occupied purchase, even though they recently closed on another owner-occupied transaction. The note they signed at their first closing included their intention to remain in the property for at least 12 months. Many lenders do not have an explicit written guideline covering this scenario. Some underwriters may (using their own various set of criteria) allow applicants to close on a second transaction under owner-occupied guidelines, others will deny the loan, while some will counter-offer to handle the transaction as an investment and then list “approved but not accepted” if the borrower does not want to move forward under the loan terms available for an investment property purchase. Because DU and LPA will typically only issue a “verify occupancy” feedback message, it is very difficult for an examiner to see the discretion that the lender applied.

Underwriter Discretion is Not Unlimited

Underwriters like to retain the right to exercise their discretion on every file they review. How many times have we all heard an underwriter say they “don’t like” or “don’t feel comfortable” with a file? On a manually underwritten file they may get more latitude, but on a file that has received an AUS approval, they are generally required to come up with a more concrete reason. There is a large but still finite number of these specific discretionary issues that a loan officer and underwriter may face, and it’s possible for a lender to institute a written policy covering all of them.

New HMDA Data

Leonard Ryan, president of Questsoft, wrote an article entitled “Piecing Together the New HMDA Puzzle” in the April 2016 edition of Mortgage Banking.  In that article, he wrote that Home Mortgage Disclosure Act (HMDA) and fair lending will become synonymous due to the automated nature of HMDA data. In this regard, Rob Chrisman pointed out in his daily blog on June 26, 2016, that lenders are busy analyzing the new HMDA data they will be required to report beginning January 1, 2018, to see if there are any patterns that could give regulators the statistical appearance that there may be fair lending violations. These newly required reportable fields contain data that lenders have been seeing but not capturing in a manner that made the data easily examinable by regulators (prior to the new requirements, these data points and all data points not subject to HMDA reporting, were often referred to as “HMDA+”). Now that this data will be readily retrievable and reportable, lenders are scrambling to see what patterns this data may reveal.

Reason for Denial Poses a Big Regulatory Risk for Lenders

The Reason for Denial is a modified data point effective January 1, 2018 and the regulations now require that the reason(s) a lender reports must be specific and accurately describe the principal reason(s) the financial institution denied the application. However, the examples the CFPB provides show that the specific reason is still very general in nature (for example “debt-to-income ratio” or “employment history”).

Consumer discrimination complaints caused by the poor handling of application data during the mortgage production process often triggers fair lending exams. The MLO promised one set of terms and the borrower is ultimately offered different ones or is rejected outright when it turns out that the MLO and underwriter were not on the same page regarding one or more pieces of discretionary data. Consumers who are initially told they are “all set” end up wondering if they were denied because they are members of a protected class. If a lender does not have a policy to provide some underwriter guidance on discretionary guidelines (and capture and analyze those data points), they are at risk of fair lending violations when a “HMDA+” audit digs deeper into the Reason For Denial and uncovers a pattern of possible discrimination due to the grossly uneven treatment of discretionary data.

Is It Disparate Impact or Disparate Treatment?

Let’s consider a lender with 100 loan officers and 10 underwriters. Say that three of those underwriters have a strict policy of only allowing one owner occupied transaction per 12 month period. In a given year, those 100 loan officers may move 50 such borrowers into the mortgage production cycle seeking to complete a second transaction. If 10 of those borrowers are members of a protected class, and by the luck of the draw seven of them end up with one of the three strict underwriters, there is going to be a strong case for disparate impact.

An even bigger risk occurs if one of the more lenient underwriters also denies a member of a protected class who had a clearly a superior financial profile (and possibly a better “reason” for “needing” to move if such explanations are retained and retrieved) compared to a member of a non-protected class who they approved. In isolation, this might be explained as a nebulous case of underwriter discretion or even underwriter error, but taken in the overall context, a lender could easily end up accused of a devastating finding of disparate treatment.

Making a Policy to Obtain Regulatory Benefits While Avoiding Regulatory Risk

The CFPB said it added the new data points beginning January 1, 2018 to reduce false positives based on credit characteristics not currently reported under HMDA.

Regulators continue to grow in their understanding of the intricacies of the mortgage business. As they see some lenders develop a policy to assemble a set of guidelines to handle each potentially discretionary situation their underwriters face, regulators could formally change HMDA to require a more specific “Reason for Denial” or they could just be on the lookout for how lenders handle discretionary scenarios when they conduct fair lending examinations.

After all, fair lending examination guidelines specifically state that vague underwriting criteria, lack of clear guidelines on making exceptions to underwriting criteria, and lack of clear loan file documentation regarding reasons for any exceptions to standard underwriting criteria are indicators of potential disparate treatment.

Part of a regulator’s fair lending compliance review is to determine whether the policies and procedures of the institution enable management to prevent, identify, and self-correct illegal disparate treatment. They are also tasked with identifying the manner by which management addresses its fair lending responsibilities with respect to the institution’s lending practices and standards, inspecting their training and other application processing aids, and exploring the guidance lenders provide to employees in dealing with customers.

During the exam process, a regulator may ask the institution to account for differences in customer treatment. A lender may be able to remain compliant if they point out a specific difference between the applicants’ qualifications, show some factor not captured in the application that legitimately makes one applicant more or less attractive to the institution, or explain the difference by pointing out some non-prohibited factor related to the processing of their applications. Since discretionary data is often the reason that applications are approved or denied, it would be better if a lender systematically retains and is prepared to use this data to explain any situations that could give concern to a regulator.

Disparate treatment may be more likely to occur in applicants neither clearly qualified nor clearly unqualified because these files have more need for discretion, so discretion is more likely to alter the ultimate underwriting decision. However, even if a loan is still approved, discretion can cause fair lending pricing violations if, for example, discretion is used to eliminate income that causes a pricing adjustment because of a resulting higher DTI.

Remember that lenders are considering this discretionary data in the process even if it’s not uniformly captured and/or easily retrieved and subject to statistical analysis. Some files may have notes placed in the LOS by an MLO and/or underwriter while others may not. Fortunately, guidance from compliance experts is available to help lenders develop a policy to handle discretionary scenarios and software has been developed to capture it so lenders can prove to regulators that borrowers are all treated uniformly and fairly.

How Can Lenders Protect Themselves?

A lender should operate as though all of their HMDA+ data, especially data that relates to their underwriting decisions, will eventually be captured and analyzed by regulators even though it is not currently reportable. Lender best practice is to develop a policy to provide guidance to their underwriters on discretionary scenarios and to capture this data so it can be analyzed to prevent fair lending violations. Underwriter discretion will rightfully always exist, but an underwriter should be provided policies that allow them to “anchor” or “give them a place to put their foot down” when they make a decision.

A business reason is a valid defense against claims of fair lending violations, but if the reason given is not part of a policy, the defense is much weaker. Regulators treat lenders more leniently if they have a policy in place, even if an underwriter misses it on a particular file.

One policy regarding a second owner occupied transaction within a 12 month timeframe would be to never allow them. Another option would be to allow them as long as the first transaction wasn’t used to either raise the down payment funds and/or pay off bills that brought the DTI in line to allow the applicant to qualify for the second transaction. A third choice would be for a lender to adopt an FHA type guideline for conventional loans and only allow a second transaction within a one year period for a new job requiring a move of a certain number of miles or an increase in family size that was not contemplated at the time of the original transaction. Any of these policies are fine and except for the first option still allow room for underwriter discretion.

The authors have seen individual underwriters adopt each of the above mentioned possibilities as their individual policy. We have even heard of underwriters abdicating their responsibility and simply conditioning for an okay letter from the current servicer of the initial owner occupied transaction! This choice is likely problematic as it could easily result in a lender denying more qualified protected class members while allowing less qualified non-protected class members to close, but it is an example of the kind of decisions that get made when management does not institute written guidelines for discretionary scenarios and effectively lose control over this aspect of their pipeline.

There is always some risk that a court or regulator won’t accept a lender’s business reason, but as long as a lender follows their policy, the more granular data they retain the better off a lender will be both operationally and regulatorily, even when the issue of fair lending is considered.


Alan Bercovitz

Alan Bercovitz is the author of The Complete 1003 Software, an artificial intelligence tool designed to deliver a more thorough, accurate, and uniform application for every mortgage borrower. He can be reached at Alan@GuaranteedMortgagequote.com.



Be Sociable, Share!
(Visited 18 times, 1 visits today)

Check Also

Avoiding Surprises

Servicing Quality Assurance and Quality Control Monitoring Systems can be Management’s Best Friend By Jim …