National Fair Housing Alliance v. Bank of America, National Association

Opinion

IN THE UNITED STATES DISTRICT COURT FOR THE DISTRICT OF MARYLAND

NATIONAL FAIR HOUSING ALLIANCE, et al.,

v. Civil Action No. SAG-18-1919

BANK OF AMERICA, N.A., et al.

MEMORANDUM OPINION

In this lawsuit, a coalition of fair housing advocacy groups and individual homeowners seeks to hold Bank of America, N.A. (“BANA”) and its servicer, Safeguard Properties Management, LLC (“Safeguard”), responsible for allegedly discriminatory maintenance and marketing of real estate owned properties in the wake of the 2008 financial crisis. The allegations are predicated on the plaintiffs’ proprietary investigation of BANA-owned, and Safeguard- serviced, properties in 37 cities between 2011 and 2018. That investigation purports to demonstrate a statistically significant racial disparity in the defendants’ marketing and maintenance of real estate owned properties, or “REOs,” which, the plaintiffs claim, provides evidence of disparate impact and disparate treatment in violation of the Fair Housing Act. The scope of the plaintiffs’ claims is vast and their theory of liability novel. Accordingly, after the defendants’ motions to dismiss were denied, the parties commenced targeted expert and fact discovery aimed at testing the methodology and validity of the plaintiffs’ investigation. Mem. & Order at 1, ECF 115. With the initial discovery process complete, the defendants now move for summary judgment, arguing that the plaintiffs’ investigation methodology is inadequate to carry their claims further. See Bank of America Mot. for Summ. J., ECF 167; Mem. in Supp. of Bank of America Mot. for Summ. J., ECF 169-1 (“BANA Mem.”); Safeguard Mot. for Summ. J., ECF 166; Mem. in Supp. of Safeguard Mot. for Summ. J., ECF 173-1 (“Safeguard Mem.”). After careful consideration of the defendants’ challenges, the Court concludes that there are genuine issues of material fact relating to the methodology and validity of the plaintiffs’ investigation. The defendants’ motions for summary judgment on these issues will therefore be denied.1

I. BACKGROUND

The Court begins by providing the factual and procedural background necessary to resolve the pending motions. A. The Plaintiffs’ Investigation As the 2008 financial crisis unfolded, BANA quickly assumed ownership of a large number of foreclosed upon properties across the country, many of which it contracted with Safeguard to service on its behalf. See Decl. of Lindsay Augustine ¶ 2, ECF 184-1 (“Augustine Decl.”). Not long after BANA began managing its newly acquired properties, the National Fair Housing Alliance, or “NFHA,” started receiving complaints that BANA was not maintaining and marketing REOs “in communities of color in the same manner they were maintaining and marketing REOs in majority-white communities.” Id. ¶ 2. “As a result of those concerns, NFHA developed a large- scale investigation to determine whether the racial makeup of a neighborhood had an impact on the quality of [the defendants’] maintenance and marketing of REO properties.” Id. According to the plaintiffs, the investigation found “that there is a racial disparity between BANA’s and Safeguard’s exterior property maintenance in white census tracts and their exterior property maintenance in non-white census tracts” that is “highly statistically significant and practically significant.” Revised Expert Report of Jacob Rugh at 3, ECF 183-7 (“Revised Rugh Report”).

1 The Court refers to Bank of America and Safeguard collectively as “the defendants.” It uses “the plaintiffs” to refer to the organizational plaintiffs; the individual homeowners’ claims are not at issue in this motion. The plaintiffs’ investigation methodology is the focus of the pending motions, so the Court discusses it in some detail. The plaintiffs began by identifying metropolitan areas to serve as the sample set for their inquiry. Augustine Decl. ¶ 3. They did this using “RealtyTrac to identify metropolitan areas with high foreclosure rates as compared to the rest of the state and the national

average” so they could hone in on “communities that were affected by the foreclosure crisis and were most likely to have a high number of REOs to evaluate.” Id. They then narrowed in on specific, “highly-segregated” communities using the Brown University Dissimilarity Index, 2010 U.S. Census data, and the software mapping tool ArcGIS. Id. ¶ 4. From this pool of potential testing sites, they applied a final filter based on the availability of local fair housing organizations to contribute resources to the investigation. Id. ¶ 5. The result was a list of 37 metropolitan areas across the United States in which to investigate the defendants’ maintenance and marketing of REOs. Id. ¶ 6. After selecting cities for investigation, the plaintiffs next had to pick specific comparator communities and, ultimately, individual properties. They “first identified zip codes within the

metro areas that were both racially concentrated and had a high foreclosure rate.” Id. ¶ 7. They classified a zip code as “predominantly” white or non-white if it was composed of at least 50.1% of the relevant racial demographic. Id. ¶ 8. They filtered those results to neighborhoods that qualified as “working- or middle-class” based on relative median income rates. Id. ¶ 10. Using publicly available data, they next identified specific BANA REOs within the target communities to investigate. Id. ¶ 11. The plaintiffs “used a concentric circles methodology” to ensure their boundaries encompassed a sufficient number of white and non-white properties as close as practicable to the relevant city center. Id. ¶ 12. Every identified property within the selected zip code was then subject to investigation so long as it “had been in Bank of America’s possession for more than 30 days or [was] actively being marketed,” subject to certain exceptions. Id. ¶ 13. After creating a master list of properties for inspection, the plaintiffs developed the evaluation methodology. They ultimately curated a list of 35 criteria to serve as an objective baseline against which to measure and compare properties.2 See Decl. of Shanna L. Smith ¶ 5,

ECF 183-5 (“Smith Decl.”). Each criterion represented a different deficiency in the “exterior maintenance, marketing, [or] overall curb appeal of” a property. Id. The plaintiffs identified deficiencies for evaluation based on their “common sense, knowledge, and experience regarding how an owner maintains a house and how real estate is marketed to homebuyers” in combination with input from “real estate and property preservation and maintenance industry stakeholders and [other] fair housing organizations.” Id. These 35 criteria supply the comparator metric for the plaintiffs’ investigation. For purposes of their assessment, the more deficiencies present on a property, the worse its score. And, in the plaintiffs’ view, the greater the disparity between the average number of observed

deficiencies in white and non-white neighborhoods, the stronger the evidence of disparate impact and disparate treatment. Using this testing mechanism, the plaintiffs conducted data collection from 2011 to 2018 through individual assessments of BANA-owned properties. Augustine Decl. ¶ 17. To conduct their evaluations, the plaintiffs dispatched teams of trained investigators to the pre-identified REOs. Id. ¶¶ 14-17. The investigators visited each property in pairs, tallied each observed deficiency on an REO evaluation form, and photographed each deficiency to document their

2 The investigation initially considered 37 criteria, but the plaintiffs later removed two utility- related criteria to reduce the total number to 35. Compare REO Field Evaluation Form, ECF 183- 41, with REO Field Evaluation Form, ECF 184-3. conclusions. Id. ¶ 17. After completing an assignment, investigators returned to a local field office to upload the results and supporting photographic evidence to a centralized database. Id. ¶ 19. NFHA executives reviewed these uploads for quality control purposes, and in some cases edited the results for accuracy and standardization. Id. ¶ 25; see Mem. Op., ECF 226. Once the results for

each property were uploaded, the plaintiffs analyzed them to compare deficiencies across white and minority neighborhoods. Augustine Decl. ¶¶ 26-30. All told, 1,405 of the properties the plaintiffs inspected remain at issue in this litigation. Id. ¶ 28. The plaintiffs’ database contains over 30,000 photographs of these properties. Id. ¶ 33. B. The Plaintiffs’ Experts The plaintiffs retained four experts to offer opinions related to their investigation. The experts are University of Michigan Professor Dr. Michael D. Fetters, M.D., M.P.H., M.A., Fair Housing Center of Southeast & Mid Michigan Executive Director Pamela A. Kisch, former A&D Property Services, Inc., Executive Vice President and Chief Financial Officer Deavay Tyler, and Brigham Young University Professor Dr. Jacob S. Rugh, Ph.D.

Dr. Fetters, a social scientist with a specialty in mixed methods research, offered three central opinions: (1) “NFHA conducted a well-designed investigation that would qualify as a rigorous and compelling mixed methods study”; (2) the defendants’ experts mistakenly assumed that the results of the investigation were intended to be extrapolated to the defendants’ entire nationwide stock of REOs; and (3) the “defendants’ experts” critiques of the investigation’s “methodological design flaws” were “unwarranted.” Expert Report of Dr. Michael Fetters at 4-5, ECF 183-2 (“Fetters Report”). In particular, Dr. Fetters endorsed the plaintiffs’ sampling, including their concentric circles methodology, data collection instrument, data collection training, and data collection procedures. See id. at 11-18, 21-23. Dr. Fetters then offered a point-by-point rebuttal to the defendants’ experts’ testimony. See id. at 23-31. Ms. Kisch also opined on the validity of the plaintiffs’ investigation, drawing on her experience as a fair housing advocate to discuss the role of fair housing organizations in enforcing

the FHA and the use of testing in uncovering FHA violations. As they were admitted in this Court’s prior opinion on the motion to strike certain expert testimony, see ECFs 229-30, Ms. Kisch provided four overarching opinions: (1) fair housing testing is an important FHA enforcement tool; (2) “Plaintiffs’ investigation is consistent with traditional enforcement testing;” (3) fair housing testing can unearth discrimination without perfectly comporting with all “social science methodological principles;” and (4) “[r]equiring FHA testing to comport with social science methodological principles, in the manner suggested by Defendants’ experts, would . . . harm fair housing efforts.” Expert Report of Pamela A. Kisch Report at 3-4, ECF 183-6 (“Kisch Report”). Ms. Kisch devoted special attention to the importance of providing plaintiffs a path to proving FHA violations within the practical constraints imposed by the realities of field observation and

the resources available to the victims of housing discrimination and the groups that advocate on their behalf. Mr. Tyler provided additional expert testimony corroborating the plaintiffs’ testing methodology based on his experience working in the property maintenance and marketing industry. Mr. Tyler offered nine categories of admissible opinions between his initial and supplemental expert reports. In his initial report, he opined that: (1) the plaintiffs’ REO evaluation criteria corresponded to industry standards; (2) the plaintiffs properly documented marketing and maintenance deficiencies; (3) the defendants should have resolved many of the deficiencies the plaintiffs recorded; and (4) the defendants’ quality control was inadequate. Expert Report of Deavay Tyler ¶ 9, ECF 183-3 (“Tyler Report”). The Court admitted five more opinions from Mr. Tyler’s supplemental expert report. Specifically: (1) many of the plaintiffs’ REO evaluation criteria were independent of a property’s

condition at the time the defendants assumed responsibility of the property; (2) the plaintiffs’ REO evaluation criteria are independent of the “features of the neighborhood”; (3) the “national model of property preservation and maintenance followed by Bank of America” is inferior to the “local model”; (4) a defense expert, Mr. Bryar, “lack[s] experience in the property preservation and maintenance industry”; and (5) banks and servicers may have more maintenance and marketing responsibilities than typical property preservation and maintenance vendors. Suppl. Report of Deavay Tyler ¶¶ 1-6, ECF 183-3 Ex. B (“Tyler Suppl. Report”). Ultimately, the thrust of Mr. Tyler’s two reports was that the plaintiffs’ 35 chosen criteria served as an accurate proxy for assessing the quality of the maintenance and marketing of that property. Finally, Dr. Rugh, a sociologist, provided the critical expert opinion analyzing the results

of the plaintiffs’ investigation. Dr. Rugh’s revised report explains that among “the properties investigated by NFHA, there was an average of 3.06 more deficiencies among properties in non- white tracts than in white tracts.” Revised Rugh Report at 2. He also found that in “nonwhite tracts, REO properties [were] 3.3 times as likely to exhibit a high number” of “deficiencies than properties in white tracts,” and that “properties in white tracts were 2.9 times as likely to exhibit a low number” of “deficiencies as in non-white tracts.” Id. He concluded that this racial discrepancy was “highly statistically significant, practically significant, and” could be explained only by race. Indeed, in controlling for as many as “40 other variables,” he found “no variable or set of variables that eliminates or comes close to accounting for the racial disparity.” Id.; see id. at 3-4. Dr. Rugh used the plaintiffs’ deficiency data, collected through their in-person evaluations, to conduct his analysis. He measured statistical significance using a p-value of 0.001 (i.e., the odds the disparity is due to chance is less than one in one thousand) and assessed practical significance using Cohen’s d, which he calculated as .86 (i.e., the disparity amounted to almost a full standard

deviation, a large effect). Id. at 6-7. Dr. Rugh then employed a regression analysis to measure the strength of race as “an explanatory factor” in the outcome of the plaintiffs’ investigation. Id. at 7.3 After conducting his analysis, Dr. Rugh concluded that “no other variable or set of variables”— besides race—“can account for most of the racial difference in deficiencies between white and non-white tracts.” Revised Rugh Report at 10; see id. at 7-17. Race, he found, explained 78% of the disparity in deficiencies—a statistically and practically significant difference. See id. at 10. Dr. Rugh later complemented his revised report with a supplemental report in which his ultimate conclusions that race is a statistically and practically significant factor in the disparity between deficiencies in white and non-white tracts remained the same. See Rugh Suppl. Report, ECF 183- 7 Ex. B, (“Rugh Suppl. Report”). This report controlled for additional variables offered by the

defendants’ experts and found that it “still [left] 61% of the disparity,” or 1.97 deficiencies, “unexplained by anything other than race.” Id. at 4. C. The Defendants’ Experts The defendants offered, between them, eight expert reports from four experts: Charles River Associates Vice President Dr. David M. Skanderson, Columbia University School of Law

3 A regression analysis is “a statistical tool used to understand the relationship between two or more variables” that “is often employed in the analysis of data about competing explanations for the relationships among a number of explanatory variables.” In re Titanium Dioxide Antitrust Litig., No. 10-cv-0318, 2013 WL 1855980, at *14 n. 14 (D. Md. May 1, 2013) (citing Daniel L. Rubinfeld, Reference Guide on Multiple Regression, inReference Manual on Scientific Evidence 305 (Fed. Judicial Ctr., 3d ed. 2011)). Professor Dr. Justin McCrary, Georgia State University Professor Dr. Vincent Yao, and Managing Director of Bryar & Wollner LLC, Housing Gears LLC, and Mortgage Policy Advisors LLC, Marcel A. Bryar, J.D. Dr. Skanderson’s initial report concluded “that the statistical regression analysis and other

statistical analysis produced by Plaintiffs from their property inspections is not a reliable basis for drawing conclusions regarding discrimination in BANA’s REO property maintenance activities.” Expert Report of David Skanderson ¶ 11, ECF 169-2 (“Skanderson Report”). He explained that while the plaintiffs’ “regression analysis explains part of the difference in average inspection results between majority white and majority minority neighborhoods,” it “does not provide a legitimate statistical basis for inferring that discrimination in REO maintenance is the cause of the remaining unexplained difference . . . because there are additional plausible explanatory factors not accounted for in Plaintiffs’ regression model, and because of the biased and unrepresentative nature of their inspection sample.” Id. ¶ 14. For instance, he observed that the “analysis does not account for relevant economic differences or differences in property condition at the time BANA

acquired them.” Id. ¶ 13. Dr. Skanderson then opined further that even accepting the validity of the plaintiffs’ statistical analyses, the ultimate disparity the plaintiffs identify is likely too small to be practically significant. Id. at ¶ 11. Dr. Skanderson later provided a supplemental report responding to the plaintiffs’ expert reports and maintaining his initial position. See Suppl. Expert Report of David Skanderson, ECF 169-9. Dr. McCrary opined that the plaintiffs’ investigation was too methodologically flawed to determine whether the defendants caused the observed disparities identified in the plaintiffs’ investigation. Expert Report of Justin McCrary ¶ 20, ECF 169-3. At “the outset,” Dr. McCrary opined, the plaintiffs’ conclusions are based on the flawed premise “that any adverse difference” in “the external condition of Bank of America REO properties . . . constitutes evidence that Defendants discriminated in the maintenance of properties[,] ignor[ing] the common-sense fact, backed up by empirical evidence, that the external property conditions of any property may differ from any other property for a large number of reasons that have nothing to do with either

foreclosure or REO status.” Id. ¶ 21. Dr. McCrary continued to state that this problem was compounded by the plaintiffs’ “fail[ure] to apply reliable analytical methods to control for differences in neighborhoods” and instead treating “any two surveyed properties as if they were comparable.” Id. ¶ 23. Dr. McCrary also identified issues with the “point in time” nature of the plaintiffs’ investigation, the difficulty of replicating the plaintiffs’ results, the bias of the plaintiffs’ investigators, and “numerous additional methodological problems.” Id. ¶¶ 25-28. Dr. McCrary also filed a supplemental report responding to the plaintiffs’ expert reports in which he maintained that “[n]othing in Plaintiffs’ Experts’ Reports changes my fundamental opinion that Plaintiffs’ investigation methodology suffers from numerous methodological flaws that individually and collectively render its conclusions unreliable.” Suppl. Report of Dr. Justin McCrary ¶¶ 1, 4, ECF

169-10. Dr. Yao’s initial report focused on explaining the property foreclosure lifecycle and how the condition of a property typically fluctuates in connection with each stage of the cycle. Expert Report of Vincent Yao, ECF 169-4 (“Yao Report”). He explained “that the negative effect of foreclosures starts when homeowners become delinquent, grows in magnitude during the delinquency period and reaches a maximum before the REO period.” Id. ¶ 16. He continued to state “that there is a stabilization and improvement in property conditions following the foreclosure sale, or when the lender takes ownership of the foreclosed property. This stabilization and improvement occurs as the lender resumes maintenance and prepares the property for sale.” Id. As a result, there are “many differences in maintenance patterns across properties” which “can be driven by various factors, such as property characteristics, the length of vacancy, and local economic conditions.” Id. ¶ 17. Nevertheless, the plaintiffs made “no attempt to take into account the initial conditions of each property at the time Bank of America assumed ownership,” and “no

attempt to measure any change in property conditions after Bank of America assumed ownership.” Id. ¶ 40. Thus, their investigation is “not sufficient to draw conclusions about any patterns in Bank of America’s REO maintenance activities.” Id. After Dr. Rugh submitted a report responding in part to Dr. Yao’s initial report, Dr. Yao filed a reply report maintaining his original position and rebutting Dr. Rugh’s opinions. Reply Report of Vincent Yao, ECF 169-11. Finally, Mr. Bryar offered an opinion on the applicability of the plaintiffs’ findings to Safeguard. Expert Report of Marcel A. Bryar, ECF 169-5 (“Bryar Report”). He opined “that it is not reasonably possible to conclude based on NFHA’s surveys and regression analysis that Safeguard’s conduct was discriminatory” because “NFHA’s surveys do not show that Safeguard had discriminatory policies or that Safeguard’s work on the At-Issue Properties (through its

subcontractors) had a discriminatory impact.” Id. ¶ 10. It was also his opinion “that it is not possible to substantiate NFHA’s claim that Safeguard engaged in intentional discrimination.” Id. This is because, in Mr. Bryar’s opinion, the investigation did “not actually assess Safeguard’s preservation work.” Id. ¶ 11. Critically, in his view, a “proper assessment of a property preservation company’s work requires” examining “the particular work the company was authorized to perform (through its subcontractors) and then only immediately after the work is completed.” Id. “NFHA did not perform such examinations, however.” Id. ¶ 12. Moreover, the plaintiffs’ deficiency criteria assess Safeguard’s work without regard for the scope of services Safeguard was contracted to perform, and thus penalizes Safeguard for disparities in deficiencies they did not have authorization to address. Id. ¶¶ 13-15. Similarly, the deficiency criteria are not, in fact, consistent with industry standards. Id. ¶¶ 16-18. Last, the investigation did not show that Safeguard employed any of the allegedly discriminatory policies. Id. ¶¶ 18-20. Dr. Bryar also provided a supplemental report responding to the plaintiffs’ experts and standing by his initial opinions. Suppl. Report of Marcel

A. Bryar ¶¶ 7, ECF 173-5. D. Procedural History The plaintiffs initially pursued their claims through administrative complaints with the U.S. Department of Housing and Urban Development, commonly referred to as “HUD.” NFHA and other groups filed an administrative complaint against BANA on September 25, 2012. See Fair Housing Compl., ECF 169-15. Around the same time, they filed administrative complaints against other banks, including U.S. Bank, based upon the same methodology. See NFHA v. U.S. Bank N.A., No. 01-12-0283-8 (HUD Jan. 8, 2016), ECF 167-6 (“U.S. Bank”). In January 2016, HUD concluded that there was “no reasonable cause to believe that” U.S. Bank “engaged in a pattern or practice of discriminatory treatment as alleged, engaged in individual instances of discrimination

as would be indicated by testing, or violated Sections 804(a), (b), (c), or (d) of the Fair Housing Act.” Id. at 29. HUD’s conclusion rested largely on its determination that the plaintiffs’ testing methodology was fatally flawed. Id. After HUD’s decision in U.S. Bank, NFHA withdrew its administrative complaint against BANA and instead filed a civil complaint in this Court on June 26, 2018. See Compl., ECF 1. The case was assigned to District Judge Catherine C. Blake. The defendants ultimately moved to dismiss the complaint, and in July 2019, Judge Blake denied their motions. Mem. Op., ECF 66; Order, ECF 67. After denying the motions to dismiss, Judge Blake initiated the discovery phase of the litigation. On November 6, 2019, Judge Blake entered an initial case management order limiting Phase I discovery to (1) the defendants’ marketing and maintenance of identified properties in 6 of the 37 metropolitan areas and (2) determining which properties surveyed by the plaintiffs were

actually the defendants’ responsibility at the time they were assessed. Stipulated Initial Case Management Order, ECF 83. Almost one year later, on November 2, 2020, Phase II of discovery began. See Mem. & Order, ECF 115; Phase II Case Management Order, ECF 117. That phase permitted the parties to “complete discovery on the properties in the six metropolitan areas that have been the focus of examination since the initial stipulated case management order entered in November 2019, and resolve the status of disputed properties; take depositions on matters of general relevance; and engage in expert discovery for the purpose of summary judgment briefing directed at the methodology and validity of the theories on which the plaintiffs’ experts’ conclusions are based.” See Mem & Order at 1, ECF 115. Judge Blake explained that no broader scope of discovery was

necessary because the defendants’ methodological challenges did “not obviously depend on a factual analysis of all the properties in the 37 metropolitan areas.” Id. Judge Blake also observed that “a ruling on the validity of the experts’ approach, while [not] entirely case-dispositive, could narrow the issues significantly, to the benefit of either the plaintiffs or the defendants depending on the outcome, and to the overall benefit of this complex litigation.” Id. at 2. At the close of Phase II discovery, the defendants filed several motions. First, they moved for sanctions due to evidence spoliation. ECF 153. Second, they moved for summary judgment. ECFs 166-167, 171. Third, they moved to exclude, at least in part, the opinions of each of the plaintiffs’ experts. ECFs 205-206. After these three sets of motions were filed, but before they were decided, this case was reassigned to the undersigned judge. Paperless Entry, Jan. 18, 2023. On January 23, 2023, this Court granted in part the defendants’ motion for sanctions, ECF 227, and on February 8, 2023, it granted in part and denied in part the defendants’ motions to exclude certain expert testimony, ECF 230. The Court now turns to the summary judgment motions.

II. LEGAL STANDARDS Under Federal Rule of Civil Procedure 56(a), summary judgment is appropriate “if the movant shows that there is no genuine dispute as to any material fact and the movant is entitled to judgment as a matter of law.” The moving party carries the burden of showing there is no genuine dispute of material fact. See Casey v. Geek Squad Subsidiary Best Buy Stores, L.P., 823 F. Supp. 2d 334, 348 (D. Md. 2011) (citing Pulliam Inv. Co. v. Cameo Props., 810 F.2d 1282, 1286 (4th Cir. 1987)). Should the moving party establish that no evidence supports the non-moving party’s case, the burden shifts to the non-movant to proffer specific facts that create a genuine factual dispute. Id. The non-moving party must provide enough admissible evidence to “carry the burden of proof in [its] claim at trial.” Mitchell v. Data Gen. Corp., 12 F.3d 1310, 1316 (4th Cir. 1993).

A mere scintilla of evidence supporting the non-moving party’s position is not enough; there must evidence sufficient for a reasonable jury to find in its favor. Casey, 823 F. Supp. 2d at 348 (citing Anderson v. Liberty Lobby, Inc., 477 U.S. 242, 251 (1986)). Moreover, a genuine issue of material fact cannot rest on “mere speculation, or building one inference upon another.” Id. at 349 (quoting Miskin v. Baxter Healthcare Corp., 107 F. Supp. 2d 669, 671 (D. Md. 1999)). Summary judgment is also warranted if the non-moving party fails to provide evidence that establishes an essential element of the case. Id. at 352. The non-moving party “must produce competent evidence on each element of [its] claim.” Id. at 348-49 (quoting Miskin, 107 F. Supp. 2d at 671). If the non-moving party fails to do so, “there can be no genuine issue as to any material fact,” because the failure to prove an essential element of the case “necessarily renders all other facts immaterial.” Celotex Corp. v. Catrett, 477 U.S. 317, 322-23 (1986); Coleman v. United States, 369 F. App’x 459, 461 (4th Cir. 2010). In ruling on a motion for summary judgment, a court must view all the facts, and any reasonable inferences that can be drawn therefrom, “in the

light most favorable to the party opposing the motion.” Matsushita Elec. Indus. Co., Ltd. v. Zenith Radio Corp., 475 U.S. 574, 587-88 (1986) (quoting United States v. Diebold, Inc., 369 U.S. 654, 655 (1962)). III. DISCUSSION The plaintiffs pursue their claims under various provisions of the Fair Housing Act (the “FHA”) on theories of both (A) disparate impact and (B) disparate treatment, and the defendants move for summary judgment on both grounds. The Court addresses each theory in turn. A. Disparate Impact An FHA disparate impact claim proceeds under a three-step burden-shifting framework: (1) the plaintiff must establish a prima facie case of disparate impact; (2) the defendant may then

offer a legitimate objective that justifies the policy responsible for the disparate outcome; and (3) the plaintiff must ultimately identify a less discriminatory alternative to serve the defendant’s legitimate goals. See Reyes v. Waples Mobile Home Park Ltd. P’ship, 903 F.3d 415, 424-27 (4th Cir. 2018); NFHA v. Fannie Mae, 294 F. Supp. 3d 940, 947 (N.D. Cal. 2018) (quoting NFHA v. Travelers Indem. Co., 261 F. Supp. 3d 20, 28 (D.D.C. 2017)). At step one, the plaintiff must satisfy three elements to establish a prima facie case based on a statistical disparity. The plaintiff must produce evidence of (1) a statistical disparity; (2) a specific policy of the defendants; and (3) a robust causal connection between the policy and the disparity. See NFHA v. Deutsche Bank Nat’l Tr., 18-cv-839, 2019 WL 5963633, at *13 (N.D. Ill. Nov. 13, 2019). The defendants’ motions for summary judgment argue that the plaintiffs fail to satisfy any of those three elements.4 i. Statistical disparity At the outset, the plaintiffs must show a significant statistical disparity to provide the

backbone for their prima facie disparate impact claim. A disparity is sufficiently significant if it demonstrates a statistically and practically significant discrepancy between “valid . . . comparative populations.” See Sw. Fair Hous. Council, Inc. v. Maricopa Domestic Water Improvement Dist., 17 F.4th 950, 963-64 & n.11 (9th Cir. 2021) (quoting Brnovich v. Democratic Nat’l Comm., 141 S. Ct. 2321, 2343 n.17 (2021)). The plaintiffs here have established a genuine dispute of fact as to whether a significant statistical disparity exists. Dr. Rugh’s expert report provides the plaintiffs’ affirmative evidence of a disparity. The investigation’s top-line conclusion is that properties in minority communities exhibited “an average of 3.06 more deficiencies” than properties in white communities. Revised Rugh Report at 2. Moreover, REOs in non-white tracts were “3.3 times as likely to exhibit a high number (11 or

more) of deficiencies than properties in white tracts (33.6% vs. 10.2%, respectively); likewise, properties in white tracts were 2.9 times as likely to exhibit a low number (5 or fewer) deficiencies as in non-white tracts (47.1% vs. 16.3%, respectively).” Id. Dr. Rugh concluded that the “difference in average deficiencies between non-white and white tracts” was “highly statistically significant”: the p value was less than 0.001, in other words, the odds that chance produced the

4 Because Phase II focused on the methodology and validity of the plaintiffs’ investigative techniques, it is not clear that steps two and three of the disparate impact framework have yet been fully explored through discovery. This Court’s ruling that the plaintiffs’ methodology and validity survive summary judgment, then, does not necessarily foreclose subsequent motions for summary judgment on other issues. This Court will set a status conference with counsel to discuss the next phase of this litigation. result is less than 1 in 1,000. Id. at 6. He also found the disparity to be practically significant: the Cohen’s d, a metric social scientists use to contextualize statistical significance across datasets with varying standard deviations, was 0.86, a large effect size. Id. at 6-7. All told, he found that “no variable or set of variables,” aside from race, “comes close to accounting for the” observed

disparity in REO deficiencies. Id. at 2. After the defendants’ experts responded to his report, Dr. Rugh ran another set of regression analyses to accommodate their critiques. This analysis controlled for 20 variables: 17 that Dr. Rugh had controlled for in his initial report and 3 suggested by the defendants’ experts. Id. at 4. He found that these 20 variables could explain 39% of the disparity between REOs in white and non-white areas, leaving 61% of the disparity, or 1.97 deficiencies, “unexplained by anything other than race.” Id. This race-based disparity of 1.97 deficiencies, in his opinion, “remains highly statistically significant and practically significant.” Id. Further, it is unchanged by removing areas “with fewer than 30 properties” or properties for which BANA “argues it was not responsible.” Id.

The defendants argue that this evidence is insufficient to show a statistical disparity for six main reasons. Specifically, they argue the disparity (1) is based upon an unrepresentative set of properties; (2) compares properties that are not similarly situated; (3) measures invalid deficiency criteria; (4) lacks evidence of maintenance practices over time; (5) relies upon subjective and unreliable evaluations; and (6) is based on inadmissible hearsay. First, the defendants say the REOs the plaintiffs evaluated “are not a random or representative sample of” the defendants’ nationwide portfolios, or even of the defendants’ portfolios within “the metropolitan areas” they “chose to investigate.” BANA Mem. at 27; see Safeguard Mem. at 31. The plaintiffs do not dispute this, but instead explain that this criticism misstates the objective of their investigation. The investigation did not seek “to extrapolate findings in the selected metropolitan areas to all BANA REOs nationwide (or to all REOs within each metro area).” Pls.’ Opp’n to Mots. for Summ. J at 25 n.29, ECF 184 (“Opp’n”). Instead, “the intent of the investigation was to assess whether there was a recurrent pattern of disparities in the

maintenance and marketing of REOs in middle [and working] class majority White neighborhoods as compared to middle [and working] class majority African American, Latino, and non-White neighborhoods” in the zip codes they investigated. Fetters Report at 23; see Opp’n at 8; Augustine Decl. ¶ 10. So while the defendants may ultimately seek to limit the extent of the available remedy based on the scope of the plaintiffs’ investigation, it does not provide a basis for granting summary judgment.5 Second, the defendants contend that the investigation did not compare similarly situated properties. BANA Mem. at 29-30; Safeguard Mem. at 30-31. Where a discrimination claim depends “upon a comparison” between members of a protected and “a non-protected class,” the “validity of” the “prima facie case depends upon whether that comparator is indeed similarly

situated.” Haywood v. Locke, 387 F. App’x 355, 359 (4th Cir. 2010). A “comparator need not be an exact match,” but must be “similar in all relevant respects.” Id. at 359-61. To satisfy this standard, the defendants say NFHA was required to compare properties sharing various characteristics identified in HUD’s U.S. Bank opinion. BANA Mem. at 29; see Safeguard Mem. at 30. As HUD explained, “similar houses cannot be moved from one neighborhood to another to assess their treatment,” so the plaintiffs’ “testing model involves finding similarly situated REOs

5 For example, whether the plaintiffs will be able to extrapolate the results of this investigation to any generalized populations (such as other middle-and-working class neighborhoods) remains in question, as does the issue of what any proposed relief will look like should the plaintiffs successfully prove one of their claims. and assessing their treatment as a proxy for Respondents’ treatment of the relevant neighborhoods.” U.S. Bank at 24. HUD found that given its “analysis of the variables involved in the assessment of any REO, at a minimum, [comparing] similarly situated REOs would require: (1) similar property value at the time of bank acquisition; (2) similar property condition at the point

of acquisition; (3) similar time in private ownership (since mortgage issuance); (4) similar time in possession of Respondent; and (5) similar age of property. In addition, any analysis would also have to control for neighborhood characteristics such as: (1) crime rate; (2) local zoning ordinances; (3) neighborhood income; and (4) environmental characteristics.” Id. The plaintiffs’ failure to compare properties sharing these attributes, the defendants urge, nullifies the significance of their results. The plaintiffs describe the burden that the defendants—and HUD—erect as impossible to satisfy and out of touch with “real world conditions.” Opp’n at 24-25 (quoting Fetters Report at 25). That is because “it was impossible for NFHA to find neighborhoods and properties that were identical in all ways except race.” Id. at 24. Instead, they “used a systematic approach to identify

targets,” id., that were “similar in all relevant respects,” Haywood, 387 F. App’x at 359-61. Dr. Fetters testified that “the plaintiffs have identified the best possible, or, at a minimum, appropriately comparable communities within the chosen cities.” Fetters Report at 25. To require “fair housing groups to find properties or neighborhoods that are identical comparators” the plaintiffs argue, “would thwart enforcement of the FHA for the claims at issue in this” and other cases. Opp’n at 26. The plaintiffs developed what they view as the only practical—yet still reliable—solution: On the front end, they investigated properties situated as similarly as possible; on the back end, Dr. Rugh controlled for the variables the defendants, and HUD, identified, and “succeeded in isolating the critical independent variable—race.” Opp’n at 23-24. The Court resists the plaintiffs’ suggestion that their burden of proof is somehow relaxed by the practical obstacles to their ambitious inquiry, but nevertheless concludes that, for the purposes of satisfying the summary judgment standard, they have created a genuine issue of material fact through the combination of their initial inspection criteria and their subsequent

regression analysis. Haywood, 387 F. App’x at 359. At the survey design stage, the plaintiffs selected for investigation only those homes that were unoccupied single-family units in working and middle-class neighborhoods and had been bank-owned for at least 30 days. Augustine Decl. ¶¶ 10-13. Afterward, Dr. Rugh controlled for many of the HUD factors, including property age and value, time in bank ownership, local crime rate, year and season evaluated, and neighborhood poverty level. Revised Rugh Report at 3; Rugh Suppl. Report at 11. As this case proceeds, of course, the defendants will be free to, among other things, challenge the adequacy of Dr. Rugh’s methods, maintain the importance of all of the HUD factors (including those Dr. Rugh did not incorporate), and dispute the appropriateness of the plaintiffs’ comparators. But at this stage, the issue presents a question of fact not amenable to disposition on summary judgment.

Third, the defendants dispute the validity of the plaintiffs’ inspection criteria. See BANA Mem. at 30-31; Safeguard Mem. at 12-17. They argue that “Plaintiffs’ inspection criteria do not correlate to the various contractual, legal, and other requirements governing [their] maintenance activities,” which they characterize as limited to health and safety metrics. BANA Mem. at 30. Thus, the plaintiffs’ inclusion of criteria like damaged siding, broken mailboxes, and dead grass “artificially inflated the purported disparity.” Id. at 32. Looking first to BANA, the Bank explains that its property preservation activities “var[ied] by investor, jurisdiction, and other property-specific criteria and were not static over the relevant period of time.” Reply in Support of BANA Mot. at 14, ECF 211-1. And to understand whether the plaintiffs’ criteria accurately measured the quality of its property preservation efforts, it continues, the criteria must accurately reflect the work BANA was actually responsible for—and was in fact—performing. In its view, sources like HUD’s 2010 and 2016 Preservation and Protection Requirements, ECF 209-23 (2010), ECF 167-34 (2016), Fannie Mae’s Property

Preservation Matrix, ECF 211-14, and Freddie Mac’s Approved Expense Amounts for the Preservation and Maintenance of Abandoned Properties, ECF 211-15, reflect these responsibilities. A review of these documents, and a comparison of the criteria they contain with those considered within industry standards by Mr. Tyler and Mr. Bryar, reveals approximately 10 criteria that would almost always have been within the scope of BANA’s responsibilities—leaving 25 always or often out of scope.6 BANA argues that the plaintiffs cannot attribute a disparity across these 25 inapplicable criteria to its work, and thus that the plaintiffs have failed to show a relevant statistical disparity to carry their prima facie case. Safeguard contends that the scope of its responsibility—and authority—for property preservation was even narrower. As Safeguard Chief Operating Officer Michael Greenbaum

explained in his declaration, “Safeguard is not a property manager or an asset manager,” does “not provide regular monitoring of the” relevant properties, and “generally did not have authority to determine what work would be done at a property.” Decl. of Michael Greenbaum ¶ 7, ECF 173- 14 (“Greenbaum Decl.”). Instead, “Safeguard was authorized” only to fulfill “specific, discrete

6 These ten criteria are: (1) trash removal; (2) overgrown grass or leaves; (3) overgrown or dead shrubbery; (4) unsecured or broken doors; (5) damaged steps and handrails; (6) damaged windows; (7) damaged roofs; (8) damaged fencing; (9) broken gutters; and (10) blocked gutters. See HUD 2010 Preservation and Protection Requirements, ECF 209-23; HUD 2016 Preservation and Protection Requirements, ECF 167-34; Fannie Mae Property Preservation Matrix, ECF 211-14; Freddie Mac Approved Expense Amounts, ECF 211-15; Tyler Report Ex. A, ECF 183-3; Bryar Report Ex. 5, ECF 169-5. The criteria are remarkably similar across these documents, though the 2010 HUD requirements are somewhat less demanding. tasks that BANA authorized, consistent with the contractual undertakings of Safeguard.” Id. ¶ 8. “Without specific instruction and authorization by BANA, Safeguard was not authorized to do any work at a property.” Id. ¶ 9. After Mr. Greenbaum reviewed the plaintiffs’ 37 deficiency criteria in light of these contractual obligations, he concluded that “23 of the criteria either always fell

outside the scope of work governed by the BANA contracts or frequently fell outside the scope of the contracts in certain circumstances.” Id. ¶ 38. The plaintiffs dispute that the scope of the defendants’ work was so limited. The minimum HUD, investor, and industry standards do not define the contours of the defendants’ potential liability, the plaintiffs say—the extent of their actual work does. And, in their view, the defendants’ internal documents and contracts show that they actually maintained properties at a much higher standard of care than the external standards they reference. To support this position, the plaintiffs look not to the standards set by outside entities but to BANA’s internal standards and its contracts with Safeguard. Specifically, they point to BANA’s own REO Inspection Checklist, see Smith Decl. Ex. 3, ECF 184-3, and a BANA inspection

agreement with Safeguard that contains BANA’s REO Inspection Checklist attached as an exhibit, see QC Inspection Statement of Work, ECF 184-8. The exterior maintenance criteria included in these documents mirror the criteria included in the plaintiffs’ investigation (though BANA’s version also contains interior inspection criteria). Moreover, the property level documents also provide evidence of Safeguard, on occasion, proposing bids to address criteria like peeling paint and damaged siding—criteria the defendants have claimed is wholly outside their scope of work. See, e.g., Opp’n Ex. 17 at REO Inspection Detail, Row 5, ECF 183-17.7

7 At this stage, the record is unclear as to whether the defendants actually performed such work on relevant properties. The defendants write off BANA’s REO Inspection Checklist as governing “inspection services, not property preservation services.” Reply in Support of Safeguard Mem. at 10, ECF 199 (emphasis added). The contract can reasonably be read otherwise, however, as it instructs Safeguard to “visually inspect properties specified by Bank of America to ensure that the current

property conditions conform to Bank of America standards,” as defined by the REO Inspection Checklist. See QC Inspection Statement of Work ¶ 1, ECF 184-8 (emphasis added). That suggests that BANA may have held itself, and Safeguard, to a higher standard that the plaintiffs’ criteria more closely reflect. Safeguard’s proposed bids can be viewed as corroborating that understanding. See, e.g., Opp’n Ex. 17 at REO Inspection Detail, Row 5, ECF 183-17. This documentary evidence is enough to create a factual dispute as to whether the plaintiffs’ statistical disparity appropriately represents the scope of defendants’ maintenance work. On top of this, Dr. Rugh’s wholesale regression analysis is not the only evidence that the defendants’ maintenance activities had a disparate impact on communities of color. Three additional datapoints add depth to his overall result: (1) Dr. Rugh concluded that the racial disparity

in deficiencies persisted across all criteria and was driven by no single deficiency; (2) Dr. Rugh found a “statistically significant and practically significant” racial disparity in “overgrown grass and leaves and the presence of trash/debris”—two criteria indisputably within the scope of the defendants’ work; and (3) the plaintiffs’ raw data demonstrates that a racial disparity persists across nearly every criterion they considered. See Revised Rugh Report at 4, 13, 15-16; Augustine Decl. Ex. F; Rugh Dep. at 242:5-245:14, ECF 183-11. At trial, the plaintiffs will of course be required to show that a significant racial disparity exists across only those criteria the defendants in fact maintained or should have maintained. For now, though, they have adduced sufficient evidence to suggest that whatever the ultimate scope of the defendants’ responsibilities, they may be able to show that a racial disparity exists across those particular criteria. Fourth, the defendants argue that the plaintiffs’ investigation cannot demonstrate an actionable disparity because the point-in-time nature of their evidence is incapable of

distinguishing between disparities in property maintenance and disparities in pre-existing property conditions. BANA Mem. at 32-34. As Dr. Yao’s expert report explains, “the negative effect of foreclosures . . . reaches a maximum before the REO period.” Yao Report ¶ 16. After the lender “takes ownership” and “resumes maintenance,” the condition of the property stabilizes. Id. This lifecycle suggests that the defendants inherited properties at the apex of their disrepair and were then responsible for rehabilitating them. Yet the plaintiffs failed to conduct a pre-acquisition inspection to determine whether there may have been a pre-existing disparity in deficiencies, so any disparity the plaintiffs purport to have observed after acquisition may simply restate the status quo ante—the plaintiffs cannot be sure it provides any insight on the quality of the defendants’ maintenance. Indeed, the defendants’ work could have improved a prior disparity and the

plaintiffs’ investigation would not detect their progress. HUD shared this concern with the plaintiffs’ investigation methodology. It explained in U.S. Bank that maintenance “is an ongoing, two dimensional, process, and a single test visit does nothing to show whether the condition of a property is improving or deteriorating.” U.S. Bank at 3-4. HUD illustrated the problem using the following chart, which captures the point. /d. at 4. Condition of two properties over time Photograph

Condition | - = men

Time

The chart shows the condition of two hypothetical properties on the Y-axis and time on the X-axis. “Although Property 1 (dashed line) is in better condition at the time photographs of it are taken, it is recerving less maintenance and falling into worse condition. Meanwhile Property 2 (solid line) is in worse condition at the time of the photographs, yet maintenance is taking place that will bring this second property into better condition over time.” /d. The plaintiffs’ investigation is incapable of capturing this dynamic. See id. As the defendants seem to recognize, their argument that point-in-time investigations misrepresent the quality of their maintenance performance due to pre-existing disparities depends upon the premise that properties in the surveyed minority communities exhibited more deficiencies at the time of acquisition than did properties in surveyed white communities. The defendants’ support for this premise is Dr. Skanderson’s testimony comparing “as-is” and “as-repaired” property values between white and non-white communities at the time they were acquired. Skanderson Report 94-98. When BANA acquired properties, it obtained a “broker price opinion” for each home. /d. § 94. These opinions included “an estimate of the market value of the property ‘in current condition with repairs limited to those posing safety and/or health issues only’ [] before any repairs,” or the “as-is value,” and “an estimate of the value after repairs that would 25

“significantly improve the condition of the property, such as structural and cosmetic repairs” or the “as-repaired value.” Id. The estimates are “generally obtained based on an exterior review by a marketing agent” and thus “align[] closely with the scope of Plaintiffs’ exterior-only inspections.” Id. Dr. Skanderson opined that the “difference between the initial as-is and as-

repaired BPO valuations can be used as a proxy for the amount of repairs that a property needs and, thus, for the extent of property damage at the time it was acquired by BANA.” Id. ¶ 95. Although it is “not a direct measure of the cost of needed repairs [] because the increase in a property’s market value yielded by a given repair expenditure may be greater than or less than the cost of repairs,” it “is the most reasonable gauge of repair needs available in the data that has been produced in this case.” Id. Dr. Skanderson applied this methodology to a subset of the properties at issue in this phase of the litigation. Analyzing a tract of 109 properties, he found that the “median as-is BPO value [was] $115,000 for properties in majority white tracts compared to $36,000 for properties in majority minority tracts.” Id. ¶ 96. The median as-repaired value was “$147,000 for the properties

in majority white neighborhoods versus $70,000 for the properties in majority minority neighborhoods.” Id. This reveals a nearly identical difference between as-is and as-repaired values: $21,125 in white neighborhoods and $22,000 in minority neighborhoods. Id. at 48, Table 9. Expressed as a percentage, though, “the median percentage difference between as-is and as- repaired values is 67% for the properties in majority minority neighborhoods compared to just 22% for properties in majority white neighborhoods.” Id. ¶ 97. Dr. Skanderson viewed this as evidence that “the properties Plaintiffs inspected in majority minority neighborhoods were in greater need of repair than the properties in majority white neighborhoods.” Id. at 98. As a result, the defendants argue, the plaintiffs’ investigation “improperly blamed Defendants for conditions they did not cause.” See BANA Reply at 18. This criticism of the plaintiffs’ point-in-time investigations does not warrant summary judgment for three reasons. To start, the defendants seek to ascribe the plaintiffs’ results to overall

preexisting property conditions. But some number of the plaintiffs’ deficiency criteria are largely independent of a property’s initial condition—that is, given the defendants’ obligations to perform routine and ongoing maintenance and marketing, the initial state of a particular criterion (for example, blocked gutters) should not be predictive of the state of that criterion at the time the plaintiffs investigated the property, which was at minimum 30 days, and on average nearly one year, later. See Revised Rugh Report at 20-21. Given the nature of the plaintiffs’ criteria, it is not clear that each of the maintenance disparities they observed can be attributed to prior-owner maintenance.8 Next, although the plaintiffs’ investigation may not have been two-dimensional on a per- property basis, it had two-dimensional features in the aggregate. That is because the plaintiffs

investigated the properties at issue at varying times after BANA acquired them: a property could have been inspected after two months or after two years. On average, “nearly a full year” passed between the defendants’ acquisition of a property and the plaintiffs’ inspection of it, “sufficient time in any reasonable estimation to remedy or ameliorate property deficiencies.” Id. at 21; see Tyler Report ¶¶ 34-37. On top of that, “if one accepts the premise that property defects were caused by the prior mortgagor and then remedied by BANA or its property preservation and maintenance agents like Safeguard, then we would expect there to be fewer deficiencies the longer BANA

8 Of course, the plaintiffs’ ability to rely on arguments about ongoing maintenance criteria may be hampered by the scope of their investigation, which given its breadth also encompasses many criteria much more susceptible to the defendants’ criticism. owned the property, certainly fewer than in the first weeks or months.” Revised Rugh Report at 17. As Dr. Rugh’s report shows, however, that “is not the case.” Id. Dr. Rugh explains “that the total number of REO property deficiencies observed is [] unrelated to how long it has been since BANA foreclosed upon the property.” Id. at 19. In fact, the plaintiffs’ evidence suggests that

property conditions improved slightly after the defendants initially acquired an REO, but then steadily declined the longer the defendants held it—suggesting that the conditions the plaintiffs observed were linked not to a property’s prior owner but to its new one. Id. at 22. Last, the plaintiffs dispute the premise that properties in minority tracts were in worse condition when BANA acquired them than were their white-tract counterparts. The plaintiffs counter the defendants’ evidence about initial home values with evidence that BANA paid approximately the same price per square foot for properties in white and non-white communities. Id. at 41. Were minority-tract properties truly plagued by a disproportionate number of deficiencies so serious they could not be remedied in the 30-day grace period the plaintiffs afforded each property (or the one-year average period between acquisition and inspection in minority

neighborhoods), they argue, then surely those homes would have had a lower value per square foot. See id. All told, the defendants’ critiques of the plaintiffs’ point-in-time investigations rest upon disputed factual premises that cannot serve as a basis for summary judgment. Moving on to the defendants’ fifth argument, they contend that the plaintiffs’ investigation cannot provide evidence of a statistical disparity because the deficiency criteria were overly subjective, the investigators were tainted by bias, and the results of particular investigations could not be replicated. BANA Mem. at 34-35; Safeguard Mem. at 10-12. The plaintiffs acknowledge that the “rating of social phenomena is a difficult task,” but their experts in social science research and FHA testing introduced evidence that the plaintiffs effectively mitigated the problems the defendants raise. See Fetters Report at 29. The plaintiffs trained investigators “rigorously,” providing “in-person” and “in-the-field apprentice training.” Id. Indeed, NFHA’s training programs are recognized as being “so distinguished that approximately 72% of all HUD Fair Housing Initiatives Program grantees rely on NFHA’s training”—and they submit that they

provided particularly “extensive” training in this case. Kisch Report at 11. The plaintiffs also used an observational checklist to “provide an objective measure of a highly complex phenomenon of interest,” and “deployed a consensus approach in the field between two observers, and through a quality check during the final data cleaning process.” Fetters Report at 29 (footnote omitted). At bottom, the potential for subjectivity to skew the accuracy of the plaintiffs’ deficiency data goes to the weight of the evidence; it is not a basis for summary judgment. Sixth, the defendants seek to exclude the plaintiffs’ investigation results altogether as based upon inadmissible hearsay. See Fed. R. Evid. 802. They argue that the evidence is not admissible under the business records exception to the rule against hearsay because, for the reasons explained in their motions for sanctions, the plaintiffs’ disposal of the original copies of their REO field

evaluation forms “indicate[s] a lack of trustworthiness.” Fed. R. Evid 803(6). The Court has already concluded that there is no evidence of bad faith or untrustworthy circumstances, however, so although they will be subject to this line of attack at trial, the records are nonetheless admissible. See Mem. Op. at 10-11, ECF 226. In sum, the plaintiffs have carried their burden of introducing sufficient evidence to create a genuine issue of material fact as to whether a statistical disparity exists in the defendants’ maintenance of the properties at issue. ii. Specific policy The plaintiffs’ burden on the second element of the prima facie case is to “identify a specific policy of the” defendants “that caused the disparity.” Deutsche Bank, 2019 WL 5963633, at *16; see Reyes, 903 F.3d at 425. The plaintiffs’ evidence here identifies a policy of undue

delegation and failure to supervise, which “can be the basis of . . . liability under a disparate-impact theory.” Wal-Mart Stores, Inc. v. Dukes, 564 U.S. 338, 355 (2011). In their briefing and interrogatory responses, the plaintiffs actually point to evidence of ten policies, practices, and procedures, “grouped into three main categories,” to which they attribute the statistical disparity in the defendants’ property maintenance and marketing. Opp’n at 43. These three categories are: (1) the defendants’ “national model” of property, preservation, and maintenance; (2) the defendants’ delegation of unchecked discretion to vendors and employees; and (3) a catchall category of “policies and practices [that] disparately impacted communities of color.” See id. at 43-62. Although the plaintiffs describe these categories as ten distinct practices in three discrete buckets, in truth, each is simply a different permutation of an overarching policy

of undue delegation and failure to supervise. The plaintiffs introduce both expert testimony and property-level documents as evidence of this policy. See id. The defendants seem to agree that the plaintiffs’ ten purported polices are better understood in “combination” and offer two responses. See BANA Mem. at 49. First, they argue that the plaintiffs’ theory is incognizable because it does not identify a “policy” but instead calls out, “at best, [] the ‘lack of a policy.’” BANA Mem. at 46 (quoting City of L.A. v. Wells Fargo & Co., No. 13-9007, 2015 WL 4398858, at *8 (C.D. Cal. July 17, 2015), aff’d, 691 F. App’x 453 (9th Cir. 2017)). The Supreme Court has made clear, however, that delegating undue discretion “can be the basis of [] liability under a disparate-impact theory—since ‘an employer’s undisciplined system of subjective decisionmaking can have precisely the same effects as a system pervaded by impermissible intentional discrimination.” Wal-Mart, 564 U.S. at 355 (emphasis added) (quoting Watson v. Fort Worth Bank & Tr., 487 U.S. 977, 990-91 (1988)). Every judge to have addressed the question in the context of lawsuits based on NFHA’s investigation, including Judge Blake, has

agreed. See Mem. Op. at 18 & n.11, ECF 66 (a policy of “undue delegation of maintenance discretion . . . or a failure to properly supervise” is cognizable); Deutsche Bank, 2019 WL 5963633, at *16 (a policy of “abdicating” responsibility is cognizable); Fannie Mae, 294 F. Supp. 3d at 948 (a “policy of delegation of discretion or failure to supervise” is cognizable). This Court will not depart from such uniform precedent. Second, the defendants vigorously deny the existence of such a policy, that is, they dispute that they in fact delegated undue discretion to their vendors and failed to supervise them. Indeed, the defendants proffer evidence of rigorous independent auditing mechanisms designed to ensure that vendors met their demanding quality assurance standards. Such protestations present a quintessentially factual dispute, however, and the plaintiffs have therefore succeeded in providing enough evidence of a specific policy to survive the defendants’

summary judgment motions. iii. Robust Causation The third element of the prima facie case is robust causation. “To establish causation in a disparate-impact claim,” plaintiffs must “‘demonstrate that the disparity they complain of is the result of one or more of the practices that they are attacking, specifically showing that each challenged practice has a significantly disparate impact’ on the protected class.” Reyes, 903 F.3d at 425 (quoting Wards Cove Packing Co. v. Atonio, 490 U.S. 642, 657 (1989)). This “robust causality requirement ensures that ‘racial imbalance does not, without more, establish a prima facie case of disparate impact’ and thus protects defendants from being held liable for racial disparities they did not create.” Inclusive Communities, 576 U.S. at 542. To satisfy this requirement, plaintiffs must show both that “the disparity they complain of is the result of one or more of the practices that they are attacking,” and that “the practice in question has caused the” disparity because of a protected characteristic. Reyes, 903 F.3d at 425.

The defendants say the plaintiffs have come up short on both fronts. The first requirement demands that the plaintiffs “connect the claimed disparity to discriminatory property maintenance policies, not to ‘race’ in general or to the wide variety of pre-existing social inequities correlated with race.” BANA Mem. at 43. The problem, the defendants say, is that, regrettably, racial disparities across a variety of factors permeate American society. Id. Such disparities—like the racial wealth gap—lead to, among other things, disparities in property conditions that cut along racial lines. See id. But this is an artifact of “lamentable pre-existing social inequalities” and the “shameful legacy” of “mortgage redlining”—not the defendants’ specific property preservation policies. Id. Dr. Rugh’s regression analysis purporting to link race with property conditions may provide evidence confirming the continuing effects of redlining, but it cannot attribute these effects

to the defendants themselves. And “[n]one of Plaintiffs’ four experts offers any opinion that a single alleged policy caused the purported disparity.” Id. at 42. This argument essentially reiterates the defendants’ earlier contention that the plaintiffs’ investigation fails to account for pre-existing property conditions and thus seeks to hold the defendants “liable for racial disparities they did not create.” Inclusive Communities, 576 U.S. at 542. It is insufficient to warrant summary judgment on the causation prong for the same reason it could not provide a basis for judgment on the disparity prong. The defendants also understate the plaintiffs’ evidence. Mr. Tyler’s reports catalog how BANA’s delegation of responsibility to Safeguard, and Safeguard’s subsequent outsourcing to third-party vendors, directly resulted in property deficiencies that should have been, but were not, remedied despite written Safeguard policies designed to catch the relevant problems. See Tyler Report ¶¶ 28-33; Tyler Suppl. Report ¶¶ 14-22; Opp’n at 44-46, 48-54. The plaintiffs have therefore introduced sufficient evidence, at this stage of the case, connecting the defendants’ policies to the racial disparity.

The second requirement asks the plaintiffs to show that the identified policy created a disparity because of race—not some other factor. The plaintiffs proffer this evidence through Dr. Rugh’s regression analysis. The defendants contend, however, that Dr. Rugh’s analysis shows only that race explains the disparity more strongly than any of the other “variables controlled for in his regression analysis.” BANA Mem. at 42 (brackets omitted) (quoting Revised Rugh Report at 3). The analysis cannot, in their view, eliminate the effects of variables Dr. Rugh did not control for— like, for example, preexisting property conditions or holdover occupants. So it shows only that race has a stronger correlation with the alleged disparity than the plaintiffs’ other selected factors— not that race in fact caused the disparity when considering all possible factors. Here, the defendants demand too much of the regression analysis. The Supreme Court has

made “clear that a regression analysis that includes less than “all measurable variables” may serve to prove a plaintiff’s case” so long as it “accounts for the major factors.” Bazemore v. Friday, 478 U.S. 385, 400 (1986) (Brennan, J., joined by all other Members of the Court, concurring in part). And whether a regression analysis accounts for the major factors is a “question[] of material fact.” Smith v. Va. Commonwealth Univ., 84 F.3d 672, 677 (4th Cir. 1996) (en banc). Dr. Rugh’s analysis controlled for approximately “40 different variables, other than race, that could potentially explain the observed disparities in maintenance” but found that “no variable on its own or in combination with others explained more than half the disparity.” Revised Rugh Report at 36-37. This included adding additional race-neutral variables proposed by defense experts to his original analysis and arriving at the same result. Id. at 4; see Rugh Suppl. Report. Dr. Rugh’s analysis is therefore sufficient to create a genuine issue of material fact as to robust causation. Finally, Safeguard argues that, for two reasons, even if the plaintiffs’ evidence is sufficient to show causation as to BANA, it is insufficient as to Safeguard: (1) Safeguard lacked the

contractual authority to remedy many of the plaintiffs’ deficiency criteria, and (2) Dr. Rugh ran his analysis on the full sample set of 1,405 properties for which the plaintiffs seek to hold BANA accountable but not the 1,333 properties the plaintiffs attribute to Safeguard (much less the 556 Safeguard believes it was actually responsible for at the time of investigation). See Safeguard Mem. at 41-46. The Court has already addressed the first contention and concluded that the plaintiffs have established a genuine dispute of fact as to whether their criteria are an appropriate proxy for Safeguard’s work. As for the second, Dr. Rugh testified in his deposition that he did run a regression analysis for the 1,333 Safeguard properties and that the results were “very similar.” Rugh Dep. at 6:19-7:2, ECF 173-16. And Safeguard may well deny responsibility for some subset of these properties, but that is a factual dispute that must be left for the jury. The plaintiffs’

evidence of robust causation is thus sufficient to survive the defendants’ motions. For all these reasons, the defendants’ motions for summary judgment on the plaintiffs’ disparate impact claims will be denied. B. Disparate Treatment The plaintiffs pursue a disparate treatment theory of liability in addition to their disparate impact theory. This Court concurs with Judge Blake’s assessment that “the nature of the plaintiffs’ claims seems closer to a disparate impact theory.” Mem. Op. at 15 n.8, ECF 66. Nevertheless, an FHA “plaintiff is not required to elect which theory the claim relies upon [] pre-trial,” Reyes, 903 F.3d at 421, so the Court considers the disparate treatment theory assuming the plaintiffs’ claims could “be recast as disparate treatment at a later stage,” Mem. Op. at 15 n.8, ECF 66. A disparate treatment claim under the FHA can be shown in at least three ways: (1) “directly, through direct or circumstantial evidence”; (2) “indirectly, through the inferential burden

shifting method known as the McDonnell Douglas test”; or (3) through the totality of the circumstances, applying the framework set out in Village of Arlington Heights v. Metropolitan Housing Development Corporation, 429 U.S. 252 (1977). Corey v. HUD, 719 F.3d 322, 325 (4th Cir. 2013) (quoting Kormoczy v. HUD, 53 F.3d 821, 823-24 (7th Cir. 1995)); see Mem. Op. at 15 n.8; McDonnell Douglas Corp. v. Green, 411 U.S. 792 (1973). The plaintiffs’ evidence of disparate impact is also sufficient at this stage to satisfy their burden under McDonnell Douglas. Under McDonnell Douglas, the plaintiff bears the initial burden of establishing a prima facie case of discrimination, which, if satisfied, shifts the burden to the defendant to show a “legitimate, nondiscriminatory reason for the” challenged policy, and then ultimately shifts the burden back to the plaintiff to prove that the defendant’s “stated

reason . . . was in fact pretext.” McDonnell Douglas, 411 U.S. at 802; see Graoch Assocs. # 33, L.P. v. Louisville/Jefferson Cnty. Metro Hum. Rels. Comm’n, 508 F.3d 366, 374 (6th Cir. 2007). In the context of the FHA, a plaintiff may make out a prima facie case of disparate treatment by “identifying and challenging a specific housing practice, and then showing an adverse effect by offering statistical evidence of a kind or degree sufficient to show that the practice in question has caused the adverse effect in question.” Graoch Assocs. # 33, L.P., 508 F.3d at 374 (brackets omitted) (quoting Kovacevich v. Kent State Univ., 224 F.3d 806, 830 (6th Cir. 2000)). For the reasons discussed above, the plaintiffs have introduced sufficient evidence to create a genuine issue of material fact as to whether the defendants’ practice of delegating undue discretion to third parties without appropriate supervision adversely impacted minority homeowners. The defendants’ motions for summary judgment on the disparate treatment claim will therefore be denied. IV. CONCLUSION

For the reasons explained above, this Court will deny the defendants’ motions for summary judgment.9 A separate order follows.

March 24, 2023 /s/ Date Stephanie A. Gallagher United States District Judge

9 In reaching its conclusion, the Court has not considered the declaration of Justin Chappell, ECF 183-52, which does not, on its face, demonstrate that Mr. Chappell worked on any of the properties at issue, or that he had personal knowledge of Safeguard’s operations or reputation in the relevant zip codes. The Court does not rely on his deposition testimony for the same reason, and the plaintiffs’ motion for leave to file a surreply, ECF 217, will therefore be denied.