The Fair Lending Risk Review: What Does Your Fair Lending “DNA” Reveal About Your Institution? | Wolters Kluwer
  • Insights

  • The Fair Lending Risk Review: What Does Your Fair Lending “DNA” Reveal About Your Institution?

    By Jeffrey Robb, Senior Regulatory Consultant and Manager

    Published Summer 2017

    Much like DNA testing that can reveal very detailed information unique to you, a Fair Lending Risk Review (FLRR) provides an overall picture of your institution’s lending performance based on a specific dataset from a statistical point of view. It is a key part of “telling your story” to your regulators, especially in preparation for an exam. The FLRR can also help identify any fair lending gaps or problems that exist on a day-to-day basis, and can be used to dig deeper into your data, especially if requested by a regulator.

    In general, fair lending represents compliance with laws and regulations that have the common goal of promoting fairness and impartiality for prohibited basis groups throughout the lending process. Fair lending obligations apply to all credit products and credit-related services for both consumer and business purposes, across all aspects of credit transactions, including marketing, application-taking, underwriting, pricing, servicing and collections.

    The FLRR focuses on a specific method of data analysis to help your institution determine lending patterns for certain “prohibited basis” groups, including those based on race, ethnicity or gender. The data analysis will help your institution understand its overall fair lending risk from a statistical point of view. 

    During a FLRR, data relating to race and ethnicity (or proxy data), and gender are combined with other available data, such as age, and analyzed against an institution’s underwriting and pricing criteria to determine if disparate impact and/or disparate treatment is suggested. If so, this may indicate potential fair lending problems.

    This is a multi-step process, starting with data quality, performance compared to demographic data, and peer performance. Major components include regression analysis that identifies outliers, and comparative file reviews that identify “matched-pairs” for further analysis. Both regression outliers and matched-pairs may need further exploration to determine if issues can be satisfactorily explained.

    Type of Data Analyzed

    The data can come from any line of business, but it typically involves the following:

    • Mortgage – Up through 2017, HMDA data are used as a base and additional data, known as “HMDA Plus” and/or proxy data, are added in. Starting in 2018, the number of HMDA fields will increase significantly, particularly around race, ethnicity and gender, which will make statistical analysis even more complicated – and important.
    • Consumer – This includes direct motor vehicle, credit card, secured credit, unsecured credit, overdraft protection programs, specialty lending products, and other lending.
    • Indirect Auto – An industry hot-button for the last few years, an indirect auto FLRR includes portfolio-wide and dealer-specific analysis.
    • Small Business – Small business loans are also getting more attention from regulators, especially the Consumer Financial Protection Bureau (CFPB) as it explores its obligations under Dodd Frank Act section 1071. 

    Multi-step Review Process

    The following steps comprise a typical Fair Lending Risk Review:

    • Data Integrity – Initially, it is critical to ensure that all data that will be analyzed have been scrubbed or validated for accuracy. Analysis of data is complicated, sometimes tedious, and always time-consuming. You don’t want to waste all that effort on underlying data that are inaccurate.
    • Lending Performance Comparison – Next, compare your institution’s lending performance, applications and originations, within your assessment area, to the demographics of the communities you serve. (If your institution is not subject to CRA, you can create a market area for the base analytics.) Also, consider the lending performance of your peers. With redlining back on the radar of the regulators, include an evaluation of your institution’s lending performance in low-to-moderate (LMI) tracts and minority tracts. There are limitations to aggregate and peer data review for certain types of analysis, such as indirect auto lending.
    • Focal Point Review – A focal point review is used to determine if there is a prohibited basis group or groups that seem as though they have been treated differently in terms of denial rates or pricing. By themselves, those denial rates or pricing differences do not indicate disparate impact without a control for legitimate creditworthiness characteristics, such as through regression analysis. If the rates for a prohibited basis group are found to fail a benchmark test, or are tested to be statistically significant, that could become a focal point during your next exam.
    • Regression Analysis – Regression analysis is a statistical operation that calculates the relationship between various underwriting factors and the decision made by the underwriters. For example, for pricing, the factors are used to determine whether the price received was within a calculated range of possible prices. This methodology helps to identify “outliers,” i.e. applicants or borrowers who appear to have been treated differently. Regression analysis also helps to Identify whether one or more prohibited basis groups were more likely to have had a higher proportion of outliers than their respective control group. Such results would indicate a higher risk of potential fair lending issues and warrant further review, such as through comparative file analysis of control group comparators.
    • Comparative File Review – While Regression analyzes entire groups, a Comparative File Review compares individuals to one another. A Comparative File Review is used to identify a control group of “similarly situated” applicants, i.e. those whose applications were approved or who received better pricing, even though their credit qualifications appeared comparable to, or worse than, those for specified prohibited basis group. Where such comparators exist, this may be used to prove the existence of disparate treatment. These source files should then be explored to identify whether there is additional information available to explain the apparent disparity.

    Indirect Auto and Proxy Data

    The CFPB has identified a significant risk among some indirect auto lenders that have policies that allow auto dealers to markup lender-established buy rates, and that compensate dealers for those markups in the form of a reserve. These policies create incentives, and permit discretion, resulting in potential pricing disparities on a prohibited basis. Lenders may, therefore, be liable as a creditor for disparities within a dealer’s transactions, or across different dealers within the lender’s portfolio.

    A FLRR includes analytics of both dealer-specific and portfolio-wide loan pricing data. However, the challenge is that the data typically available in HMDA data, collected as part of the Government Monitoring Information (GMI), are not readily available for indirect auto loans.

    The CFPB supports the “BISG” (Bayesian Improved Surname Geocoding) proxy methodology of analysis, which assigns race and ethnicity based on the surname of the individual combined with geographic data. This approach is somewhat controversial, although generally accepted, as BISG is not 100% accurate and does not have all names included in the database. The OCC is also said to be working on a new proxy–“BISGF” –which includes a first name. A proxy for gender generally relies on a Social Security database for names, and has similar criticisms surrounding accuracy.

    Telling Your Story

    The identification of the double helix by Watson and Crick has ultimately unleashed an explosive amount of data that can now be used to solve ancient mysteries and identify intimate details about yourself. While not yet quite this expansive, the data contained within your lending files contain a lot of information, and contains even more data HMDA fields with the expanded regulatory collection requirements that took effect in 2018. The revised HMDA requirements not only will double the amount of data most institutions report, the new data required provides exponentially more analytical possibilities, raising more questions around potential discrimination. It’s critical that you take control of the information you have available to you, slice and dice it as the regulators are doing–and be prepared to tell your story. Information is power.

    About the Author

    Jeffrey Robb, CRCM, MBA, is a Senior Regulatory Consultant and Manager of the Fair Lending Risk Review team at Wolters Kluwer. Jeff supports clients of all sizes and types, including indirect auto lenders, with statistical analysis of their lending data. This includes identifying areas of risk and focal points for ongoing fair lending analysis, including regression analysis, and creating ongoing analytical programs. He brings over 24 years of experience to help clients pre-and-post exams, and also conducts fair lending regulatory training for staff and board members. Jeff can be reached at

  • Please take a moment and tell us what you think of our content.