HMDA Reality Check: What You Can and Cannot Conclude from New Mortgage Loan Data
Extensive data about mortgage lending activity collected pursuant to the Home Mortgage Disclosure Act (“HMDA”) was just made available to the public for the first time on March 29, 2019. More detail about borrowers, about underwriting, and about loan features is now available than ever before, and that information also is easier for the public to access than it ever has been. The mortgage lending industry should expect that the expanded HMDA data will receive significant attention and scrutiny from private organizations and individuals, and the data is certain to spark controversy about the racial, ethnic and gender fairness of mortgage lending.
Aggregate HMDA data has generally revealed gaps in lending outcomes that could correlate to factors such as race and ethnicity, and the newly enhanced data will aid enforcement officials and advocacy organizations by facilitating a more precise identification of lenders that might warrant detailed investigations and analyses of the reasons for any such gaps. The enhanced data, by itself, however, still cannot serve as a legally sufficient basis for launching a charge of unlawful discrimination. It is critical to recognize and put into perspective the continued limitations of HMDA data. Absent the proper perspective, unsupported accusations of discrimination, however spurious, may nevertheless arise, and given this risk, all lenders should be prepared to address the predictable claims.
The mortgage landscape is now being shaped by major advancements in technology that advance nondiscrimination. During HMDA’s life-span, the industry has evolved from employing largely manual loan underwriting to decision-making based on sophisticated, automated loan underwriting engines, including those devised and required by Fannie Mae, Freddie Mac and HUD. This evolution has largely eliminated enforcement agency discrimination challenges to loan underwriting, notwithstanding certain continued aggregate disparities revealed by HMDA data. Major legal reforms have also been instituted in the recent years. For instance, concerns about unlawful discrimination in loan pricing were reduced substantially after the federal government prohibited lenders from compensating their loan originators based on the terms or conditions of a loan. It is not reasonably disputable that root causes of possible discrimination in underwriting and pricing have been addressed, which, in turn, provides important context for careful evaluation of remaining gaps that HMDA data may show.
In this context, the principle that HMDA data alone does not and cannot establish discrimination has long been recognized by the federal agencies and regulators tasked with enforcing fair lending laws. For two decades HMDA’s implementing regulation has stated that a “purpose” of the statute is to provide “loan data that can be used” to “assist in identifying possible discriminatory lending patterns.” The key words here are “assist” and “possible.” Webster’s Third New International Dictionary defines “assist” as “to give usually supplementary support or aid to,” Def. 1 (1981), and “possible” as “being something that may or may not be true or actual,” Def. 2 (1981). Thus, in plain English terms, the purpose of HMDA data is to serve as a supplementary tool that allows “identif[ication of] lenders that potentially are at heightened risk of having violated the fair lending laws and target[ing of] investigations and examinations accordingly.”
The enhanced HMDA data provides additional fields that might be relevant to a credit decision. These fields can be useful in identifying red flags and areas that may require deeper scrutiny. They do not, however, shed light on how the myriad factors of risk are layered and evaluated by the complex engines that predict performance. Also, legitimate non-statistical factors specific and proprietary to each lending institution are regularly considered in evaluating compliance with fair lending laws. As the FDIC has explained, the agency “fully investigates all the institutions that appear on its outlier lists to determine the source of the disparities reflected in the HMDA data. For most of the outlier reviews to date, the FDIC found that non-discriminatory reasons explain the disparities.”
Beyond HMDA data, the non-statistical factors considered by federal regulatory agencies conducting fair lending examinations or investigations include the lender’s “organizational structure”; “culture”; “business lines”; “products”; “services”; “customer base”; “fair lending compliance program”; “consumer complaints”; “loan files”; and “underwriting and pricing policies.” Moreover, OCC guidance instructs that due to factors such as “smaller loan volumes,” for certain “midsize institutions and for community banks, statistical analysis is not appropriate or feasible”; thus, “a fair lending examination for midsize and community banks is typically a comparative file review, rather than the use of statistical analysis.”
This holistic approach is consistent with the 2015 mandate of the U.S. Supreme Court, directing that “[r]acial imbalance … does not, without more, establish a prima facie case” of discrimination under the Fair Housing Act. That means that any statistical differences or disparities in HMDA data (i.e., “racial imbalance”) do not and cannot prove discrimination standing alone. Even the enhanced data will not provide the information necessary to level the serious change of unlawful discrimination.
The industry is rightfully concerned that the newly released HMDA data will cause unwarranted allegations. Notwithstanding the repeated warnings issued by knowledgeable enforcement officials, the purpose and limitations of HMDA often get lost and there is a propensity for third-party groups, including the media, to rely solely on HMDA’s raw numbers to paint exaggerated and invalid depictions of lending patterns. These mischaracterizations are almost always misleading and incomplete, and they harm consumers and communities by undermining sincere efforts by institutions to offer financing to all borrowers.
As the federal government recognized more than 10 years ago in the debate about expanding HMDA data fields, at most, “expanding HMDA data to include certain underwriting data could facilitate regulatory and independent research efforts to assess the potential risk for mortgage discrimination,” thus “enabling them to better target investigations and examinations toward the lenders most at risk of having violated the fair lending laws.” But even “[w]hile certain key underwriting data, such as borrower credit scores, DTI ratios, and LTV ratios, generally would benefit regulatory screening efforts and independent research, advocacy, and private enforcement, they may not be sufficient to resolve questions about potential heightened risk for discrimination by individual lenders or in the industry generally.”
It is no answer to say that lenders facing public indictments of discrimination based solely on HMDA data can prevail by proving their innocence. That concept itself stands our civil system of justice, which is founded on the maxim “innocent until proven guilty,” on its head. Moreover, a primary objective of lenders, as with most businesses, is to minimize the risk of ever facing a legal claim. Accusations of discrimination present a grave charge, are expensive to defend, and can occasion an immediate reputational injury and business disruption. These concerns are not limited to legal claims that might be filed in a court of law, where ethical and pleading rules set at least minimum standards for bringing suit. The court of public opinion is open to any person capable of using a computer or smart phone, and concerns regarding reputational injury occasioned by studies reaching improper conclusions of discrimination based on HMDA data alone are only heightened by the reality of our modern society’s connection to social media.
These circumstances impact all mortgage lenders as the new data is made easily available to any interested person. Reputational damaging allegations of discrimination can come from left field, and allegations of this type often receive front-page media attention. The best defense for any lender is to promptly conduct its own analysis of the new data, and be prepared to tell its own account of its efforts to promote and achieve nondiscrimination in lending.
 12 C.F.R. § 226.36(d) (2010).
 Home Mortgage Disclosure, 54 Fed. Reg. 51362 (Dec. 15, 1989) (codified at 12 C.F.R. 1003.1(b)(1)).
 U.S. GOV’T ACCOUNTABILITY OFFICE, GAO-09-704, FAIR LENDING: DATA LIMITATIONS AND THE FRAGMENTED U.S. FINANCIAL REGULATORY STRUCTURE CHALLENGE FEDERAL OVERSIGHT AND ENFORCEMENT EFFORTS, at 14 (2009) [hereinafter Fair Lending Data Limitations].
 Letter from Sandra Thompson, Dir., Fed. Deposit Ins. Corp., to Richard J. Hillman, Managing Dir., Fin. Mkts. and Cmty. Inv., U.S. Gov’t Accountability Off. (July 10, 2009) (describing the FDIC’s fair lending supervisory process).
 Letter from John C. Dugan, Comptroller of the Currency, Off. of the Comptroller of the Currency: Adm’r of Nat’l Banks, to Orice Williams Brown, Dir., Fin. Mkts. and Cmty. Inv., U.S. Gov’t Accountability Off. (July 10, 2009) (describing the OCC’s fair lending supervisory process); Letter from Sandra Braunstein, Dir., Div. of Consumer and Cmty. Affairs, Bd. of Governors of the Fed. Reserve Sys., to Orice Williams Brown, Dir., Fin. Mkts. and Cmty. Inv., U.S. Gov’t Accountability Off. (July 9, 2009) (describing the Federal Reserve’s fair lending supervisory process).
 Letter from John C. Dugan, Comptroller of the Currency, Off. of the Comptroller of the Currency: Adm’r of Nat’l Banks, to Orice Williams Brown, Dir., Fin. Mkts. and Cmty. Inv., U.S. Gov’t Accountability Off. (July 10, 2009).
 Texas Dep’t of Hous. & Cmty. Affairs v. Inclusive Communities Project, Inc., 135 S. Ct. at 2507, 2523 (2015) (quoting Wards Cove Packing Co. v. Atonio, 490 U.S. 642, 653 (1989)). The Fair Housing Act prohibits discrimination in lending and, along with the Equal Credit Opportunity Act, is the primary source of federal fair lending requirements.
 Fair Lending Data Limitations, supra note 2, at 20.
 Id. at 21.
 The evidentiary analysis required to respond to an alleged fair lending violation is inherently complex. As recognized by the government, “lenders often hire law firms that specialize in fair lending to assist the lender in its response.” Fair Lending Data Limitations, supra note 2, at 58.