An old adage is that statistical analysis can only be as good as the data that goes into it. Garbage in, garbage out. In the realm of fair lending analysis, there is a dilemma that every analyst must deal with, namely, how to handle and analyze HMDA records where race and/or ethnicity are reported as “not provided”.

“Why should I be concerned about ‘not provided’ reporting?”

HMDA rules allow the applicant to be coded as “not provided” in several situations and will pass the CFPB edit checks. However, the more “not provided” is used, the more uncertainty will surround fair lending analysis of those data. Remember, statistical analyses take representative samples to draw inferences about the population. The cause for concern is that a non-random pattern of unknown values with respect to race/ethnicity can bias conclusions made about the population for whom race and ethnicity are reported.

If the distribution of unknown race/ethnicity varies by “origination source” (e.g., online, retail branch, wholesale broker loans), loan officer, income, branch, or racial composition of the census tract, then the sample with reported race and ethnicity will be less representative of the true applicant population.  And if that’s the case, it becomes a matter that needs to be controlled for in the analysis, which is a subject beyond the scope of this blog.

Takeaways for your institution

Given the above, it is paramount to scrutinize the HMDA data prior to submission to identify any patterns of “not provided” data. LendingPatterns™️, HMDA Ready™️, and Fair Lending Magic™️ are the tools of choice by many users to tackle the issue.  Here are a few things that can be done in relation to “not provided” values:

  1. Determine whether your institution’s percentage of “not provided” is above your peers, especially those with similar business models. (For those who want to delve even deeper, you can also look at whether the share of applications where race/ethnicity is determined based on visual observation or surname is roughly in line with peers.)
  2. If your institution is an outlier relative to peers, trace the source of the reporting back to where it first appeared. You may find it’s a function of geography (e.g., MSA), origination source, branch, or loan originator.
  3. Determine if your institution is requesting GMI in a standardized and correct manner. Can it be done in a more effective way?
  4. Identify some peers that over-perform in this reporting, and inquire about how they go about it.
  5. Help the loan origination side of the institution to understand the value of this data for accurate fair lending monitoring. This requires communication on the topic, and may require new internal controls.

We can help you monitor and evaluate your fair lending risks.  The ComplianceTech Suite of popular fair lending software products include:

If you’re not sure which fair lending software suits your organization’s needs best — or whether you need more than one solution — request a demo so you can experience the features that come with each option.