As published in ABA Bank Compliance (now ABA Risk and Compliance) by Jennifer Paradise, September/October 2021

When it comes to managing fair lending risk, banks should consider all loan products offered, not just mortgage loans reported under the Home Mortgage Disclosure Act (HMDA). Regulation B generally only permits collection of government monitoring information for residential mortgage loans. Without this demographic data, determining fair lending risk exposure outside the realm of home mortgage lending can be challenging, although it is not an impossible mission.

Importance of Managing Fair Lending Risk in Non-mortgage Products

Managing fair lending risk begins with a strong compliance management system (CMS) that enables management to prevent, or to identify and self-correct compliance violations. An effective fair lending CMS entails that the bank’s board and management meet the same objectives as the general CMS. To know what kind of picture their lending activity paints, it is important for banks to manage fair lending risk adequately in non-mortgage products.

Managing risk exposure in products such as consumer auto loans and credit cards may not receive the same level of attention as real estate secured loans, but they do present fair lending risk. This point is made clear in the FFIEC Interagency Fair Lending Examination Procedures (“Exam Procedures”), which state, “The procedures emphasize racial and national origin discrimination in residential transactions, but the key principles are applicable to other prohibited bases and to nonresidential transactions.” (See www.ffiec.gov/PDF/fairlend.pdf.)

Discriminatory Practices occur when an applicant is treated less favorably on a prohibited basis. There are two primary theories of discriminatory practices: disparate treatment and disparate impact.

Disparate treatment occurs when a lender treats a credit applicant differently based on a prohibited basis. It can be established in either of two methods:

• Overt Evidence—where a lender openly discriminates on a prohibited basis which includes making discriminatory statements or having a policy that expressly discriminates on a prohibited basis. It can also encompass when a lender expresses, but does not act on, a discriminatory preference. (Note that Overt Evidence is sometimes considered its own fair lending theory rather than part of Disparate Treatment.)
• Comparative Evidence—where similarly situated applicants or borrowers are treated differently and these differences in treatment are not fully explained by legitimate nondiscriminatory factors.

Disparate impact occurs when a lender applies a facially neutral policy or practice equally to all applicants, but the policy or practice disproportionately excludes or burdens certain persons on a prohibited basis. (For more information on prohibited bases, see Fair Lending and Prohibited Bases—New Developments on page 16 of this issue.)

When evaluating the potential for these discriminatory practices, examiners will consider various indicators of discrimination for each of the risk factors outlined in the Exam Procedures: underwriting, pricing, steering, redlining, and marketing. The CFPB also looks at servicing and collections. Being familiar with these factors and their potential presence in loan products is crucial for managing fair lending risk. When developing new loan products or making changes to an existing product, these discrimination risk factors should also be taken into consideration before roll-out.

Presence of Fair Lending Risk in Non-Mortgage Products

Fair lending applies to every step of the credit process, and non-mortgage products present different types and degrees of fair lending risk exposure. In particular, if you have discretionary decision-making and/or pricing, then you have fair lending risk. Manual, judgmental underwriting that permits discretion may present more fair lending risk than reliance on a validated and tested automated underwriting system.

The risk of disparate treatment also increases if the bank does not use standard underwriting and pricing guidelines, or if the bank does not have a written policy on exceptions, does not adequately document exceptions, or fails to monitor for potential disparities on a prohibited basis. The greater the permissible lender discretion, the higher the risk exposure. Disparate treatment by steering occurs when applicants are guided toward or away from a specific loan product or feature because of their race, sex, or other prohibited characteristic.

For example, add-on products such as “payment protection” and “credit monitoring” that are sold without regard to the applicant’s needs or other legitimate factors, exposes the bank to fair lending risk. If the bank fails to properly consider the treatment of limited English proficiency (LEP) and non-English-speaking consumers in its product and service offerings, the greater the risk of potential steering into less-advantageous products. (See Speaking to Their Hearts: Considerations for Serving Limited English Proficiency Customers on page 28 of the May–June issue of ABA Bank Compliance.)

Small Business Lending

A CFPB blog post from April 27, 2020, entitled “The importance of fair and equitable access to credit for minority and women-owned businesses” highlights potential warning signs of lending discrimination. While the blog discusses the purpose of the Coronavirus Aid, Relief, and Economic Security Act and the Small Business Administration’s Paycheck Protection Program, it emphasizes that minority and women-owned business owners are protected by Regulation B and includes the following warning signs:

• Refusal of available loan or workout option even though you qualify;
• Offers of credit or workout options with a higher rate or worse terms than the one you applied for, even though you qualify;
• Discouragement from applying for credit;
• Denial of credit without given a reason or told how to find out why; and
• Negative comments about race, national origin, sex, or other protected statutes.

Banks should evaluate their lending activities to ensure the warning signs in the CFPB’s blog post are not present and that loan officers and other staff involved in small business lending receive appropriate fair lending training.

Credit Card Lending

Utilization of credit scoring models and pre-screened marketing campaigns are common in credit card lending. Some credit card issuers use custom scoring models to predict an applicant’s credit risk. Automated decision systems are often used, which may limit the potential for similar applicants to be treated differently; however, these systems can also be a source of fair lending risk if not properly developed and monitored. Credit card issuers often use pre-screened marketing campaigns. Consumers who receive these offers of credit, or invitations to apply for credit, are screened using objective criteria and/or statistical models to develop a targeted list of recipients. The potential risk of disparate impact arises if the demographic distribution of recipients for the pre-screened offer reveals that qualified applicants from a protected class were underrepresented or excluded.

Automobile Lending

Indirect auto lending carries a greater risk of disparate treatment in pricing if the bank’s dealer compensation policies permit discretionary pricing adjustments. Common methods of dealer compensation include a flat fee per loan or a discretionary dealer interest rate mark-up above the minimum base rate set by the bank. These mark-ups have nothing to do with the borrowers’ creditworthiness and if not limited and closely monitored, can result in disproportionately higher interest rates for borrowers in a protected class.

Additional sources of dealer compensation are from the sale of add-on products and services such as vehicle service contracts, guaranteed auto protection (GAP) insurance, extended warranties, and credit insurance products such as death, disability, or involuntary unemployment insurance. Failing to track and monitor the sale (and refunds when applicable) of these products and services exposes the bank to fair lending risk.

Challenges With Identifying Potential Fair Lending Risk Exposure (And Solutions)

Conducting a meaningful fair lending analysis to identify potential risk exposure requires demographic information, yet the data collection prohibitions make it challenging. Using surrogates to infer an applicant’s race, ethnicity, or gender is one way to bridge this gap. This point is also made in the Exam Procedures, which state, “…examiners need not attempt to calculate the indicated ratios for racial or national origin characteristics when the institution is not a HMDA reporter. However, consideration should be given in such cases to whether or not such calculations should be made based on gender or racial-ethnic surrogates.”

In conducting fair lending analysis of non-mortgage products, the CFPB relies on the Bayesian Improved Surname Geocoding method, which combines geography and surname-based information to produce a proxy probability for race and ethnicity. Applying this methodology requires the applicant’s physical address, the five-digit zip code and the applicant’s full name. The probability of belonging to a specific race and ethnicity is derived using census tract data and the list of Frequently Occurring Surnames from the U.S. Census. Because regulators are utilizing this method of fair lending analysis, banks should consider using it as well.

Another option that the CFPB has utilized lately is more of a redlining analysis. Also relying on zip codes, this method examines the census tract or county where the application or loan originated and draws assumptions based on the demographics of the location. Also, potential redlining can be exposed by using the number of small businesses in each census tract (available from Dun & Bradstreet), compared to the bank’s penetration in predominantly minority tracts.

Conducting analyses on small business loans (other than to sole proprietors) can present additional challenges. For example, knowing whether the business is minority or women owned. When the Dodd-Frank Wall Street Reform and Consumer Protection Act was passed, Section 1071 amended the ECOA to require small business data collection. The CFPB is planning to release its proposal later this year. In the meantime, since ECOA prohibits the collection of a business owner’s demographic characteristics, using surrogates to identify race and gender is the only viable option. This would require banks to maintain records of the principal business owner’s full name since it could serve as a surrogate for his or her race and gender. The applicant’s surname could be used to proxy for race and the applicant’s first name could be used to proxy for gender. The Social Security Administration’s Popular Baby Names list is often used for this purpose.

Another challenge in conducting a fair lending analysis could be the lack of information available for non-originated loans. Most loan origination systems include the data fields necessary to perform analyses, yet too often banks fail to populate all these fields. In the case of small business or commercial loans, non-originated loans may not even be tracked by the institution. Inadequate data leads to unreliable results, which inhibits the banks’ ability to identify potential fair lending risks accurately. Banks should have procedures in place to ensure the proper collection of all relevant data to conduct a fair lending analysis.

Available Resources

Software vendors and consultants perform fair lending data analysis, and there are resources available if your bank is not in a position to invest in software. For instance, during a 2013 Federal Reserve Outlook Live webinar entitled “2013 Interagency Fair Lending Hot Topics,” a Step-by-Step Guide to Coding for Gender and Ethnicity (“Guide”) was included in the presentation materials. The Guide offers a method for inferring an applicant’s gender and ethnicity by using their first and last name, based on female and Hispanic names lists from the U.S. Census. Also included in the materials are examples of hypothetical loan data along with the lists and the formulas required to perform the calculations in Excel. The only caveat is that an intermediate to expert knowledge of Microsoft Excel is necessary.

Conclusions

Whether it is understanding the bank’s fair lending risk, integrity of the data, or ability to conduct analysis, the goal of knowing what kind of picture your lending activity paints is not out of reach. Regardless of the method you use, it is possible to achieve the mission of overcoming fair lending challenges in non-mortgage products.