Weighing CFPB’s HMDA resubmission guidelines

Most institutions are already starting to gear up for the HMDA Regulation C changes that are coming in 2018. The regulation is now final and covered financial institutions are working on understanding the complexities of the regulatory changes, making adjustments to various operating systems, forms, procedures, and determining resource needs.

In the middle of that process, the Consumer Financial Protection Bureau issued a Notice and Request for Information to consider changes to its HMDA Resubmission Guidelines.

The CFPB HMDA Resubmission Guidelines outline how and when CFPB would typically require supervised institutions (their supervised institutions) to correct and resubmit inaccurate HMDA data.

The guidelines are part of CFPB examination procedures. They describe the review process used to verify the accuracy of HMDA data, including sample sizes and thresholds for requiring correction and resubmission of the HMDA LAR (loan application register).

We know that it’s pretty close to impossible to always have a perfect HMDA LAR, particularly if your bank is a larger institution that has a sizeable number of transactions and when you have a certain portion of HMDA reporting that is subject to interpretation by the examiners who are reviewing the LAR.

Right now there are 26 data points for each transaction. Even a bank that only has 50 reportable transactions per year has 1,300 opportunities to make an error in its HMDA LAR each year. (And the regulators will go back multiple years at an examination).

And the number of data fields is increasing substantially when the new HMDA regulation becomes effective, which will greatly add to the error opportunities.

CFBP seeks input

On Jan. 12, 2016, CFPB published a Notice and Request for Information concerning its HMDA corrective action and resubmission guidelines in the Federal Register. Comments were due by March 14.

There were no specific changes proposed by CFPB. The bureau was seeking input on how it might change the existing guidelines.

One of the big questions is whether the existing error percentage thresholds should continue to be used to determine whether resubmission of HMDA data is required.

Currently, if an institution reports fewer than 100,000 total transactions per year, that institution is required to correct and resubmit its HMDA LAR when errors are found in 10% or more of the sampled entries, or in 5% or more of the sampled entries within an individual data field.

For example, if a bank has 2,000 total transactions and CFBP sampled 79 transactions (normal sample size) during the exam and found 8 transactions that had an error, the bank would be required to resubmit the LAR. Those percentage thresholds are set lower (4% and 2%-4%, respectively) for larger institutions that report 100,000 or more transactions.

In addition, CFPB has discretion within these “guidelines” to require corrective and resubmission regardless of the percentage of the error rate, if it believes the errors make analysis of the institution’s lending performance unreliable. (Or, in actuality, whenever they want to.)

According to CFPB’s current exam procedures, a typical streamlined HMDA review for an institution with fewer than 100,000 HMDA transactions would initially sample 32 loans. If no errors were found, the review would stop. If an error was found examiners would sample more files. If an error was found and CFPB staff continued sampling more transactions, correction and resubmission would be required if the error rate reached the 10% or 5% threshold.

Specific questions from the bureau

CFPB, through the notice and request for public comment, is looking at whether and how to change these guidelines, including:

• Should the threshold percentages be changed?

• Should they use some method other than percentage of total errors and data field errors?

• Should larger institutions with large numbers of transactions have a different threshold than smaller institutions with fewer transactions? If so, where should that dividing line be?

• Should some data fields receive different treatment because they are more or less important than others? (For example, are errors regarding the race or ethnicity of the borrower worse than if you mess up the occupancy code for the loan?)

• Are sample sizes that CFPB uses to review HMDA LARs appropriate with the changed regulation?

Key question not asked

What CFPB’s notice is not asking is how the guidelines will affect non-CFPB supervised institutions:

• Will other regulators follow CFPB’s lead?

• Will other regulators have a more or less stringent process that will create inequalities among regulated financial institutions? 

• When will guidance come out—before or after exams start? 

Changes to the guidelines are almost assured.

Think about it. Expansion of HMDA data fields by 12, plus changes to many of the existing data fields, significantly increases the potential for errors.

The old threshold percentages applied to more data fields automatically makes it more difficult to avoid resubmission level.