By Martha J. Svoboda
The traditional focus of HMDA compliance has been, and continues to be, on the annual monitoring of mortgage loan origination and mortgage loan purchase activity. This article discusses possible benefits and challenges from the recent amendments to the regulations implementing HMDA, specifically those pertaining to fair lending analysis of the data that is required to be collected by home mortgage lenders.
HMDA, as enacted and amended over the years, has three primary purposes. The first purpose is to help assess whether home mortgage lenders are serving the needs of their communities. The second purpose is to assist public officials in the distribution of public-sector investment and the attraction of private investment. The third purpose is to assist in the identification of possible discriminatory lending patterns and the enforcement of antidiscrimination statutes.
It is primarily this third purpose, commonly referred to as “fair lending,” for which the HMDA Loan Application Register (LAR) was developed through HMDA’s implementing regulation, Regulation C. Fair lending law requires fair, impartial, and unprejudiced access to credit for qualified persons by prohibiting discrimination on the basis of any characteristic that is protected by law. These protected classes include race, religion, national origin, sex, marital status, age, those who have received public assistance income, familial status, persons with a handicap, and any applicant who, in good faith, exercised any right under the Consumer Credit Protection Act.
The theory of discrimination generally associated with a fair lending analysis that utilizes LAR data is known as “disparate treatment.” One type of disparate treatment using LAR data occurs when one individual is accorded different, less preferential, treatment when compared to a similarly situated individual on account of the disparately treated person’s membership in a protected class. Because the treatment is deemed to relate to the protected basis, intent is often implied, and regulators and enforcement agencies will look for comparative evidence in LAR data to prove a fair lending violation. However, if an institution can show that, in making the credit decision, the outcome was actually based on a legitimate difference that justified treating one applicant more favorably than the other, the appropriate conclusion is that the applicants were not similarly situated, such that no unlawful disparate treatment occurred.
Another type of disparate treatment analysis that utilizes LAR data occurs on a geographical basis. Sometimes there exists a statistically significant difference in LAR data between similarly situated, or peer, mortgage lenders in a geographic area. This statistical difference forms the basis for allegations of a type of disparate treatment known as “redlining.” In contrast to the type of disparate treatment revealed by a comparison of treatment between individuals, redlining may be alleged where a high-minority-concentration geographic area appears to have been disparately treated, or redlined, without regard to the fact that some individual borrowers residing within that geographic area may be qualified for a mortgage loan.
With respect to LAR data and its adequacy in a disparate treatment analysis, whether on an individual basis or a geographic basis, a 2015 press release from the Federal Financial Institutions Examination Council (FFIEC) observed that the current LAR data does not provide adequate information on its own to assess a mortgage lender’s compliance with fair lending law because many “potential determinants of loan application and pricing decisions” are absent from the current LAR data set. These perceived inadequacies now appear to be eliminated with the recently adopted amendments to Regulation C, referred to by the Consumer Financial Protection Bureau (CFPB) as the “Final Rule.” However, as discussed below, the Final Rule brings with it a potential set of new challenges.
The Final Rule, which is generally effective January 1, 2018, adds 25 new data elements to the existing LAR data set, and modifies and expands many other existing elements. Certain of the new elements, such as the combined loan-to-value ratio (CLTV), credit score, debt-to-income ratio (DTI), and automated underwriting system (AUS) results, will provide regulators and enforcement agencies information about lending practices that is currently only available in a loan file-by-loan file review such as occurs during a regulatory examination or under a formal document request by an enforcement agency. The positive aspects of this additional insight stand in tension with new challenges, particularly with respect to the ability of mortgage lenders to tell their own story about the meaning of their data.
The lack of certain loan-level information in the current LAR data set has resulted in regulatory and enforcement agencies’ allegations of disparate treatment redlining based primarily on a statistical analysis of a mortgage lender’s application or origination rate in comparison with that of other mortgage lenders that are deemed to be its peers. The increased visibility to loan-level information provided by the Final Rule, as opposed to mere statistical comparisons in the formulation of allegations of fair lending violations, should allow every mortgage lender to be judged on its own merits, rather than on comparisons to the mortgage lender’s deemed peers.
Further, the results of a fair lending analysis based on actual data should be more predictable and allow for better business planning. The expanded LAR data set should allow a fair lending inquiry to be more detailed, comprehensive and focused on the activity of a mortgage lender and its origination channel. This should facilitate a better understanding of disparities in both underwriting and pricing outcomes. New data elements such as credit score, DTI, CLTV, interest rate, interest rate spread, discount points, origination fees, and lender credits will now be analyzed, not just in isolation, but also in totality, to form a more clear overall picture. Similarly, denial ratios based on AUS results and action taken will replace the old matched-pair review previously performed by agencies only under an examination or investigation scenario.
While the new data elements in the Final Rule should improve fair lending inquiries of the type mortgage lenders have generally experienced, the new data elements, however, may allow for other areas of fair lending analysis to emerge. These other areas, if they emerge, will present new challenges for mortgage lenders.
Some of these new challenges may come from the inclusion of home equity lines of credit (HELOCS) as “covered loans” in the LAR data set and from the additional new data that will be included in the LAR data set. The inclusion of HELOCS and, by way of illustration, data on cash-out refinancings and interest rate types, could result in regulators and enforcement agencies renewing their focus on instances of loan/application steering. The new LAR data sets for HELOCs, cash-out refinancings and interest rate types will enable analysis of the rate of incidence among various class members of HELOCs versus home equity loans; home equity loans versus cash-out refinancings; prime versus subprime loans; adjustable-rate (ARM) versus fixed-rate loan products; and rate spreads for all loans, not just the current “higher-priced mortgage loans.” Other new LAR data elements, including the NMLSR identifier and the application channel (retail versus wholesale), will enable assessment of disparities that correlate with an individual loan originator or particular channel. The LAR data set can be analyzed for predatory lending patterns based on loan product attributes in totality, e.g., teaser rates, non-amortizing features, ARMs, high-DTI borrowers, high-fee loans, and prepayment penalties. Availability of this type of information may, for example, create a shift in the current redlining focus on overall branch and loan production office activity to a more delineated focus on individual loan originators or specific channels.
Another challenge from the new data that will be included in the LAR data set relates to the ability to identify originators who are especially successful at penetrating minority communities and neighborhoods. This capability, while a positive for some originators, will enable regulators and enforcement agencies to aggregate total loan application activity of individual wholesale originators/brokers (Broker or Brokers) across all of the mortgage lenders to whom the Broker brokers loans. This capability to look at a Broker’s total application activity and thereby identify outliers as to particular application activity directed to a single originating institution might possibly facilitate implication of a discriminatory disposition on the part of that originating mortgage lender, rather than a discriminatory intent by a singular Broker.
The granular nature of the new LAR data set may create its own challenges, too, especially for national mortgage lenders. Even if a lending institution’s numbers appear in-line on a national basis, they can be analyzed and sorted at more and more delineated levels, down to regional and local levels, until disparities are found. Similarly, the Final Rule’s framework for reporting race and ethnicity by disaggregated sub-categories may present challenges. An answer of “Puerto Rican” or “Cuban” will be captured in the disaggregated sub-categories for ethnicity, rather than in the current aggregate category of “Hispanic.” This granularity may be used by regulators and enforcement agencies, as well as consumer groups, in unforeseen ways in a quest for evidence of disparate treatment on an individual level, as well as in a peer analysis relative to redlining – particularly with respect to answers provided in the new free-form, self-identified “other” data fields and in combinations that may be accumulated based on a number of protected characteristics. By way of illustration, the CFPB has already stated that it will be looking for “improper consideration of age in credit decisions.”
The availability of new, more discrete data may also enable other changes in the approach to a fair lending analysis, all of which may present additional challenges for mortgage lenders. Access to loan data for a particular property address might allow a more narrowed neighborhood lending focus, and may also inform analysts as to lending patterns on the fringes of certain government-designated census tracts. This could give rise to a gradual evolution in the geographic census tract-based analysis that is the current focus of redlining investigations.
A final, but extremely important, challenge that may emerge relates to the ability of a regulator or enforcement agency to slice and dice the data on its own. This means that a fair lending analysis can be performed from an examiner’s or investigator’s desk without any hint to the mortgage lender that there may be a possible fair lending concern. Further, the richness of the new information that will be available to an off-site analyst because of the Final Rule may result in a conclusion that the typical business justifications that are allowed to be made by a mortgage lender when a possible loan-level disparity is identified during the course of an examination or investigation are not necessary for a proper analysis. Consequently, the regulator’s or agency’s own arguments can be fully formulated and developed without asking for the mortgage lender’s own view on a particular lending decision or lending pattern. As anyone who has been in an examination or investigation situation knows, once a regulator or an agency has made up its mind, it is much more difficult to persuade them otherwise. On this note, it is critical that each mortgage lender carefully scrub its data set prior to submission, as inaccurate LAR data will lead to inaccurate fair lending conclusions.
Despite the new challenges for mortgage lenders, opportunities may still be found. The expanded LAR data set required by the Final Rule provides a sensible occasion for each mortgage lender to better know its own data and to correct institutional deficiencies on a prospective basis prior to detection of past lapses by outside users of the data, including regulatory and enforcement agencies. Finally, depending on the extent and substance of the public data set published by the CFPB—a determination that has not yet been announced—the availability of more data on competitor’s underwriting and pricing schemes may allow an institution to gauge the competition and seek opportunities to differentiate its product offerings and better compete in the marketplace.
Martha J. Svoboda is Of Counsel at Poyner Spruill