HMDA Data Re-submission
By: Greggory B. Oberg, Esq.
Once again, HMDA time is upon us. Examiners are gearing up for 2020 exams, and HMDA data from from 2018 to current is fair game. This is nothing new to us, but every year I find clients unsure of the process regulators will use to examine their HMDA Data Validity for upcoming exams.
This Week, I take a deep dive on the Interagency Examination Procedures used under the 2018 HMDA Rule, and most recently updated in April 2019. These procedures, despite all rumors to the contrary following the infamous Dec 2017 CFPB "good faith compliance" letter, will dictate whether the efforts you took to adapt to 2018 Reporting (and beyond) were successful--or whether you're back to the drawing board and resubmitting corrected data.
Transaction Testing Procedure
As examination procedures go, this one is relatively straight forward. Flip down to page 74 of the CPFB Supervision and Examination Manual linked at the end of this article. The next 13 or so pages will be the focus here. While I won't focus on every step in the process, this should serve as a decent summary of what you can expect with 2018 and 2019 data reviews pending for examinees.
Sample Selection and Error Testing
First, your regulator will request a copy of the LAR and files for a "Initial Sample" population, which is a function of the overall volume contained on the LAR. As the screenshot below shows, see Column B for the initial sample size appropriate to the volume of your institution.
For most of us, this means an initial sample of less than 35 loans. If anybody is having an initial sample of 61 loans, I need to do more business with you, call me.
From this sample, your examiner will perform a "Stage 1" review of the LAR and corresponding files pulled for review. While the examiner is technically free to investigate any and all fields within the sample for errors, there is a general consensus on what constitutes a "critical field" for review.
it is likely that the examiners will focus on 37 fields, or 21 if you are a "partially exempt" HMDA reporter under EGRRCPA. Those fields are shown below, using the 2019 OCC Bulletin's language (linked at bottom of article).
OCC: To evaluate financial institutions' compliance with HMDA requirements, OCC examination staff will focus on identified key data fields during transaction testing pursuant to HMDA for data collected on or after January 1, 2018. Examination staff will focus on the 37 fields listed below for banks that are subject to collecting, recording, and reporting information for all HMDA data fields.
Error Testing
The process of file review is well documented in examiner workpapers. As stated in the examination procedures:
Examiners should document in their work papers any differences between the data in the HMDA LAR and information in files, and determine whether the differences may be explained by any additional information that the financial institution may provide.
Differences that are not adequately explained should be identified as errors.
This process is designed to favor a finding of error where inconsistent documentation practices make it difficult to conclusively validate the date. Essentially, it is safe to assume that the tie will not go in your favor if additional documentation cannot clear up the issues. Better practice is to have already ensured the documentation needed is present.
Examiners WILL, however, allow some error tolerance in certain fields.
Application Date and Action Taken Date: subject to +/-3 calendar day tolerance, provided such error doesn't cause a HMDA reportable record to shift into a different calendar year.
Gross Income and Loan Amount: Allow rounding errors in rounding to the nearest thousand, and a $1000 tolerance, respectively.
Note that all four of these items are "critical fields," which are the most likely to result in re-submission requirements.
Stage Two and Beyond (if necessary)
Once the examiner has reviewed the Initial Sample, if the number errors within each Critical Field (or theoretically any field, but safe to assume at least the critical fields) exceeds the Initial Sample Threshold in Column C, then the remainder of the Total Sample (column A) will be reviewed as a "Stage Two" review.
During stage two, the examiner must review all critical field for which error exceeded the threshold. However, the examiner may choose to review all fields, even those without error in Stage One testing.
Re-Submission Thresholds
After reviewing the Total Sample (likely <60 loans, inclusive of the Initial Sample), the Column D Re-Submission thresholds will be compared to the actual errors found. If exceeded, the institution must resubmit.
Fell below re-submission, or better yet won in Stage 1? Well guess what…they can STILL make you resubmit based solely on the "reasonable" discretion of the examiner that the errors are "likely to make analysis of the HMDA data unreliable."
Fortunately, we get a pretty decent insight as to what will be considered "likely to make analysis … unreliable." The examples provided in the Examination Procedure pertain particularly to the key objective HDMA data is meant to monitor--lending disparity on the basis of prohibited characteristics. AKA Fair Lending. For example, the Guide states that consistent incorrect coding of action taken as withdrawn--opposed to denied--would be likely to cause material issues in underwriting disparity analysis.
CFPB and other regulators have spoken on what they consider to be "material" or "critical" data fields (see OCC example, above) and the logic there largely lines up to this fair lending analysis purpose.
Another class of resubmission is less important--sometimes. Here, examiners may add back in loans which were previously omitted from the HMDA LAR, such as those erroneously coded as exempt for temporary finance, or where the security property is improperly classified as commercial…a common issue in mixed use property.
Good Faith Efforts?
This is a good place to talk about certain buzz we've been hearing from clients. If you recall, the CFPB notified industry at the 11th hour that 2018 HMDA data was, to paraphrase heavily, not a big deal if you tried your best. More specifically:
"Recognizing the … significant systems and operational challenges needed to adjust to the revised regulation, for HMDA data collected in 2018 and reported in 2019 the Bureau does not intend to require data resubmission unless data errors are material. Furthermore, the Bureau does not intend to assess penalties with respect to errors in data[.] Collection and submission of the 2018 HMDA data will provide financial institutions an opportunity to identify any gaps in their implementation … and make improvements in their HMDA compliance management systems[.] Any examinations of 2018 HMDA data will be diagnostic to help institutions identify compliance weaknesses and will credit good faith compliance efforts."
CFPB Statement with respect to HMDA implementation, Dec. 21, 2017.
Many of us--me included--took this at face value; but that might not have been the right read. Despite the language above tending to preclude re-submission and other negative consequences in 2018 data reporting, we have heard from multiple sources within the New England mortgage community that at least one federal regulator has required re-submission of 2018 data.
If you're scheduled for an exam in this calendar year; be aware of the fact that 2018 HMDA data may play a bigger role than you expect it to, promises of "good faith credit" notwithstanding.
Resources:
