Ensuring the accuracy of Home Mortgage Disclosure Act data has always been a challenge. Even with the changes from the recently enacted Economic Growth, Regulatory Relief, and Consumer Protection Act that exempts small volume depository institutions and credit unions from certain reporting requirements, the challenges for 2018 will be even greater.
Leveraging technology to perform data integrity reviews can increase data accuracy and reduce the chances of having to do costly manual file reviews and resubmissions.
Regulators expect HMDA data to be accurate and they expect mortgage lenders to implement a system of policies, procedures, training, and ongoing monitoring to confirm its accuracy. Assuring HMDA data integrity, whether manually or through systemic data integrity testing should be an integral part of the ongoing monitoring process.
Beginning with the 2018 HMDA data, regulatory agencies, including the Consumer Financial Protection Bureau and prudential regulators will begin using new HMDA testing procedures and resubmission thresholds when examining the accuracy of HMDA data. According to the FFIEC HMDA Examiner Transaction Testing Guidelines issued in August 2017, an institution with a Loan Application Register (“LAR”) with at least 501 and up to 100,000 loan applications can expect examiners to select and test up to 79 loans for accuracy during a HMDA exam.
If four or more errors are found in any one data field, the institution will be directed to correct that field, and resubmit its LAR. That represents only a 5.1% resubmission threshold. The resubmission threshold for smaller institutions is triggered at only three errors and ranges from 5.1% to 6.4% for LARs between 101 and 500 items. That does not leave much room for error. While the traditional method of manually comparing system and loan file information to the LAR data can be effective, it becomes less practical and costlier for lenders with larger HMDA LARs.
Systemic data integrity reviews can analyze the entire LAR without looking at a loan file. The process includes obtaining LAR data and data from the company’s loan origination system and comparing the data sets to validate fields reported and identify potential errors. These potential errors can then be further reviewed and resolved, allowing for a more effective use of resources, and resulting in better data integrity.
A MULTI-STEP APPROACH
An effective systemic data integrity process is performed using a multi-step approach and includes both omissions testing and conflict/error testing.
Omissions testing: Omissions testing is a reconciliation of the transactions reported on the HMDA LAR to transactions entered into the company’s loan origination system. Omitting reportable applications from the HMDA LAR is a common issue among HMDA filers, and a regulatory hot button. The 2017 FFIEC Examiner Transaction Testing Guidelines specifically indicate a lender may be directed to resubmit its HMDA LAR to include reportable applications or loans if it is determined that reportable applications or loans were omitted in error.
Traditional manual file review methodologies may fall short of identifying applications and loans that were omitted from the LAR because they do not use a comprehensive source for comparison. Instead, institutions rely on piecemeal sources such as credit bureau vendor activity, CALL Report information, or collateral code reports.
By comparing the application numbers on the LAR data to application records on the company’s loan origination system, applications that were received, but not reported on the LAR, can be identified.
Once identified, these records can be further reviewed to determine if the applications were incorrectly omitted from the LAR and need to be added. This testing will also identify any applications incorrectly reported on the LAR. This usually occurs if an application is included on the LAR for the wrong reporting year, or perhaps an inquiry or prequalification request was incorrectly reported.
Conflict/error testing: Systemic data integrity testing includes validity and quality edits similar to the validity and quality edits performed by regulatory software upon submission of the LAR. These tests will identify technical compliance exceptions and logic exceptions within the HMDA data. For example, a validity error would occur if the action taken date reported on the LAR occurs before the application date reported.
Systemic data integrity methods are designed to use data available within the loan origination system that can be used to cross-validate the accuracy of HMDA reportable data to identify conflicts or errors. This is performed through both a field-to-field validation process and a logic validation process.
• The field-to-field validation process includes matching LAR fields to various system fields that descriptively contain the same information. The system fields used for comparison need to be clearly defined based on the way the lender uses the field in its system. For example, the application date on the LAR would be compared to the “application date” and “disclosure application date” fields on the loan origination system to identify potential errors as long as the system field contained the defined data.
• Logic validation includes matching LAR fields to loan origination system data using logic validation comparisons. For example, the application date reported on the LAR would be compared to system data fields such as “loan created date,” “date file started,” “intent to proceed date,” etc. Another example would be comparing loans reported on the LAR as a refinance to system data fields indicating “first time home buyer.” Potential conflicts are identified for further review and correction.
Rate spreads and geocoding information: Systemic data integrity methods can also be used to validate rate spreads and geocoding information, which are derived from other loan data such as APR and property address. These derived fields can be independently calculated using third party software, and compared to the information reported on the LAR.
However, be sure that the third-party software is also periodically tested for accuracy by comparing its output to the FFIEC HMDA rate spread calculator and geocoding system. Any differences identified can be further researched and resolved.
PROS AND CONS OF SYSTEMIC REVIEWS
One of the major benefits of a systemic data integrity review process is that it can assess 100% of the items included on the LAR in an efficient manner. Once a validation program is built and customized for the company, the process can be done regularly throughout the year for ongoing monitoring and avoiding last minute reviews prior to submission.
It also allows for better trending analysis, which in turn facilitates better root cause analysis, corrective action, and accurate performance analysis throughout the year.
The process is not infallible as some reporting errors cannot be identified with system data, but instead can only be identified by manually reviewing the loan file information. The adage, garbage in, garbage out, applies here.
For example, some errors may be identified from inconsistencies in underwriter notes or originator comments. While system data and LAR data may indicate that a loan application was withdrawn, loan originator comments may provide insights indicating that denial reasons were communicated to the applicant prior to him or her withdrawing the application. Therefore, it is imperative that reporting procedures are clear and effective training is in place.
It may still be beneficial to perform targeted testing of a small number of files to address the potential for contradictory information in the loan file.
The accuracy and completeness of HMDA data is critical to understanding where and to whom loans are being made to manage fair lending risk and to ensure compliance with HMDA.
Regulators continue to scrutinize lenders, and inaccurate data can lead to costly file reviews, resubmissions, fines, and civil money penalties.
Leveraging technology to perform ongoing periodic reviews of HMDA data can help manage that risk and help ensure the accuracy of your HMDA data.