With HMDA Changes Implemented, Time to Update Your Fair Lending Analytics
(As published in March 2019 issue of Scotsman Guide, Residential Edition)
Now that the Home Mortgage Disclosure Act (HMDA) changes have been implemented, is it time to relax and reflect on the completed project? Actually, no.
Although it may seem that the HMDA regulatory reform is a significant enough regulatory burden all on its own, let’s remember that the HMDA data is the fuel used for fair lending and Community Reinvestment Act (CRA) analytics. Clearly the regulators, community groups, and the HMDA savvy public will have a greater level of detail relating to mortgage lending applications and originations than ever before. History shows that changes to HMDA reportable data are soon followed by increased regulatory and civil activity against lenders for possible fair lending and CRA violations.
To some in the industry, fair lending analytics might seem like the work of an illusionist. It’s especially intimidating if the mystery is left to the regulators to solve. Those familiar with working with data, numbers and statistics usually know all of the tricks behind the smoke and mirrors. However, even the most seasoned statistician will need to re-evaluate the fair lending implications of all of the new HMDA data fields.
Set the table
Initially, the most important precursor to performing fair lending analysis is to be sure that the data is accurate. No amount of magic can turn bad data good. Therefore, it is critical to ensure that all HMDA data is timely and accurately collected. Additionally, it would be a mistake to think the new analysis will be the same as the previous HMDA Plus analysis. There are a lot of nuances around the new HMDA reporting, even though the data is familiar.
Divide the changes in half
So, what is the best way to approach the HMDA changes? Perhaps it might be easier to divide the changes into two key areas: one relating to the change in the number of data records that have been collected, and the other relating to specific fields of data were collected.
Reporting requirement changes
Perhaps the biggest influence on the overall analysis is the influx of newly reportable records into the HMDA-reportable definition. For consumer transactions, reportable transactions are determined based on the securitization of a “dwelling,” and no longer purpose-based. This new standard has the potential to increase the HMDA reportable population for many institutions. These changes are important because fair lending analytics is based on numbers and ratios. Even small changes to this overall population of data can unmask potential problems.
These types of changes will also likely impact the overall risk profile, as well as how an institution will perform a trend analysis, as year over year will no longer involve an apples-to-apples type of comparison. Also, core business, compliance and marketing strategies that have been put in place based on the 2017 data profile may need to be re-visited based on the data profile that will be created from this new data. Let’s discuss some of the biggest changes that should be considered as part of your analytics toolkit.
Home Equity Lines of Credit (HELOCs)
Previously optional to report, the required reporting of dwelling-secured, open-end lines of credit (commonly known as a Home Equity Lines of Credit) may be the single most impactful change of all of the new reporting requirements. This is for two reasons: the portfolio nature of the product, and the overall volume of reporting for these transactions.
For products such as HELOCs that are likely to be retained on portfolio, credit exceptions are believed to be more likely to occur, as an institution would not have to answer to an investor on these determinations. Additionally, underwriting and pricing criteria may not have been developed in as rigorous a process as traditional first mortgage loans. This added level of discretion without an appropriate level of controls could result in a higher degree of risk for potential discrimination on a prohibited basis now that this data is readily available. Therefore, lenders should examine the credit and pricing policies and controls to ensure that the criteria are empirically derived. This would be in addition to performing fair lending analysis, monitoring, and testing to ensure that the policies are equally applied to all applicants.
Additionally, the sheer number of HELOCs could potentially shift ratios and statistics for an institution in dramatic fashion. For example, the introduction of a large population of HELOCs within the HMDA data will undoubtedly impact redlining analysis, which has been a hot topic for regulatory scrutiny for the past several years. It may take a lot of time for institutions to adjust to these new volume measures, so it’s critical to start doing analytics as soon as possible to determine the impact of this new data for one’s institution.
Specific to the State of New York, applications in which the transaction was completed involving a CEMA (Consolidation, Extension, and Modification Agreement) are now reportable. For New York lenders, this could be a significant portion of the application population. Institutions should begin to perform analysis on the collective, new HMDA reportable population to determine if the addition of CEMA reporting results in any disparities.
Unsecured home improvement loans
For consumer transactions (not business purpose), these will no longer be reportable. If previous analysis is included in this population, analysis should be updated to review the populations separately in order to understand the impact and changes year-over-year.
Pre-Approval requests—approved but not accepted
Previously optional to report, these records are now required. Further adding to the transparency of the lending data, the addition of pre-approval data will add more records to the reporting for those institutions that participate in a pre-approval program.
Adjustments based on new reported data fields
Certainly, the expanded data fields are the area of significant focus, as the data can now be evaluated in many ways, singularly and combined, and can thus yield much more information about each application. It is critical to be proactive, and consider the fields that have the highest potential impact. The following are of some of the fields that should be considered in that analysis.
The Application Channel added into the mix of the HMDA data introduces an interesting wrinkle into the fair lending analytics. If a lender participates in both direct and indirect lending, separate analysis by channel may yield disparities that were previously hidden within the analysis of the aggregate data. Because only a small number of loans could mask problems, lenders who engage in both direct and indirect lending should begin to understand how their data may be viewed separately—before the regulators have access to the data. If currently performing regression or other robust analysis, these institutions should also begin to bifurcate the approach in order to determine if a single channel may surface hidden issues.
Fees & Interest rate
The expanded HMDA reporting includes several pricing fields, including the fees associated with the loan (Origination Charges, Discount Points, and Total Loan Costs), Lender Credits, and the Interest Rate. Because of the impact of these fields on the APR calculation and pricing decisions, data integrity controls and robust analysis needs to be put in place to ensure that these criteria are assessed individually, as well as in the aggregate APR calculated field. For example, lenders should know which of their borrowers are charged Discount Points and the degree to which these points benefit the borrower. There may also be some trends or customs by regions that may need to be accounted for to ensure that one can accurately “tell your story” when interpreting the data and discussing any apparent differences with the regulator.
Prepayment penalties & non-amortizing features
These loan features have had a negative perception associated with them over the past few years. Institutions need to understand the distribution of these elements to ensure that there is no disproportionate distribution of these components across prohibited basis factors.
Automated underwriting system results
The automated underwriting results will provide the regulators a clearer view into exceptions to underwriting policies. In order to limit scrutiny around exceptions, lenders may want to consider tighter controls around exceptions and general underwriting procedures before the data is reported.
CTLV, Credit Score, DTI
Data fields such as Combined Loan to Value (CLTV), Credit Score, and Debt-to-Income (DTI) ratio are important factors that may be used in credit decisions and loan pricing, and thereby are typically involved in focal point reviews and regression analysis. If the denial 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 the next exam, and may involve redlining or perhaps reverse redlining concerns. Therefore, it is critical for these fields to be accurate.
This may be the most challenging and complicated area for data collection, data integrity, and data analysis—and could impact marketing, redlining, pricing, underwriting and servicing fair lending analytics. Lenders need to be confidant that they understand the many nuances and complexities of the rules relating to the aggregate, disaggregate and free-form text categories. Lenders should monitor the volume of applications received from applicants that designate their race or ethnicity within these new subcategories in order to begin the analysis on these populations as soon as possible. Similar to many of these other newly reported fields, lenders should know whether there are issues within the more granular data that may be masked in the aggregate data reporting.
Pulling a rabbit out of a hat may seem like a much easier feat to accomplish than analyzing the new HMDA data. For all lenders, it is time to start analyzing this new data to better understand how this information will be viewed. The time to act is now to devise a strategy to ensure that fair lending analytics will be something to celebrate!