Connecticut Fair Housing Ctr v. CoreLogic Rental Property Solutions, LLC

Opinion

UNITED STATES DISTRICT COURT DISTRICT OF CONNECTICUT

CONNECTICUT FAIR HOUSING CENTER ET AL. : Plaintiffs, : : v. : No. 3:18-CV-705 (VLB) : CORELOGIC RENTAL PROPERTY : SOLUTIONS, LLC : Defendant. : January 24, 2020

ORDER ON MOTIONS TO COMPEL [Dkts. 65, 84, 97]

On April 24, 2018, Plaintiffs Carmen Arroyo and the Connecticut Fair Housing Center (collectively, “Plaintiffs”) filed a complaint against Defendant CoreLogic Rental Property Solutions, LLC (“RPS”). [Dkt. 1]. Plaintiffs alleges that RPS, through use of its criminal tenant screening product, CrimSAFE, violated the Fair Housing Act, 42 U.S.C. §§ 3601 et seq. (“FHA”), the Connecticut Unfair Trade Practice Act, Conn. Gen. Stat. §§ 42-110a et seq. (“CUTPA”) and the Fair Credit Reporting Act, 15 U.S.C. §§ 1681 et seq. (“FCRA”). Pending before the Court is the Plaintiffs’ July 19, 2019 motion to compel production of documents pursuant to Federal Rule of Civil Procedure 37. [Dkt. 65 (7/19/2019 Mot. to Compel), 65-1 (Mem. Supp.)]. As modified by the Plaintiffs’ subsequent Reply, [Dkt. 80], Plaintiffs seek an order compelling the RPS to produce documents in response to 14 requests for production: RFPs 31, 32, 47, 48, 49, 50, 51, 52, 53, 54, 56, 58, 72, and 73. RPS objected to the motion. [Dkt. 74 (Opp. to 7/19/2019 Mot.)]. Plaintiff replied. [Dkt. 80 (Reply Supp. 7/19/2019 Mot.)]. RPS also filed a sur-reply in response to the first motion to compel. [Dkt. 84-1].1

Also pending before the Court is the Plaintiffs’ September 24, 2019 motion to compel production of a response to RFP 30. [Dkt. 97 (9/24/2019 Mot. to Compel Resp. to RFP 30)]. RPS objected. [Dkt. 102 (Opp. To 9/24/2019 Mot.)]. Plaintiffs replied. [Dkt. 109 (Reply Supp. 9/24/2019 Mot.))]. For the reasons set forth below, the Court GRANTS in part and DENIES in

part Plaintiffs’ July 19 motion [Dkt. 65] and DENIES Plaintiffs’ September 24 motion [Dkt. 97]. Also, as noted, the Court also grants RPS leave to file a sur-reply per Local Rule 7(d). [Dkt. 84]. I. Relevant Factual Background A. CrimSAFE RPS’s product CrimSAFE screens the criminal history of housing rental applicants in three discrete steps: record acquisition, record-applicant matching. ,

and record categorization. In the first step of its process, RPS obtains criminal records from various government entities and organizes and stores them in its database. Inaccuracy may be introduced at this stage if the criminal records are incorporated incorrectly into the database or if they themselves contain mistakes.

1 The Court here grants RPS leave to file a sur-reply pursuant to Rule 7(d). The Court finds that good cause for filing a sur-reply exists in light of the new evidence raised in Plaintiffs’ reply. [Dkt. 84]. In the second step of its process, when a CrimSAFE report is requested, an automated search, or “matching logic,” attempts to match criminal records to the applicant based on identifying information. Incorrect matches may be generated because of record inaccuracies or because the records contain too little

information to be ensure they uniquely match to the individual to whom they pertain and instead may also match to other individuals with similar common names. These mistakes and how they are handled by the matching logic will generate higher or lower rates of “false positives,” where RPS attributes a record to the wrong person, and “false negatives,” where RPS fails to attribute a record to the right person. In the third step of its process, RPS sorts the matched records into categories and assigns an age to each. It then retrieves any records for a given rental applicant which fall within the landlord’s search parameters; if it retrieves

any relevant records, it reports that a disqualifying record exists and the applicant is denied. B. The Instant Case The instant case arises out of a tenant screening report ordered by RPS’s customer WinnResidential on Mikhail Arroyo in April 2016. The parties agree that the CrimSAFE portion of the screening report that RPS provided to WinnResidential accurately reflected that RPS had located a criminal record for Mr. Arroyo, and that the criminal record identified by RPS was correctly attributed to Mr. Arroyo. Rather, the Plaintiffs’ disparate impact claim arises out of the following policy: Defendant has a policy or practice of contracting with owners and managers of rental housing to screen applicants by: (i) searching its national database to locate applicants’ criminal records; (ii) determining that applicants’ criminal records, including but not limited to arrests and/or charges that do not lead to convictions, disqualify them from tenancy based on an automated evaluation; (iii) making these determinations without individualized assessments that examine relevant mitigating information outside the criminal records themselves; (iv) reporting to housing providers that applicants’ criminal records are “disqualifying”; and (v) not providing to housing providers any information about the nature, recency, or seriousness of the offense, or information sufficient to locate the criminal record (“Defendant’s Automated Criminal Records Screening Policy”).

[Dkt. 1 Compl. ¶106].

Plaintiffs allege that this Automated Criminal Records Screening policy has a disparate impact on Latinos and African-Americans. [Dkt. 1 ¶194]. II. Relevant Procedural Background Plaintiffs served their first Requests for Production on October 26, 2018, and RPS responded on December 17, 2018. [Dkt. 65-1 at 3]. The parties conferred over email and telephone over the next several months, narrowing their dispute to Nos. 31 and 32. Ibid. Plaintiffs served their Second Requests for Production on April 19, 2019, and RPS responded on June 6, 2019. Ibid. The parties conferred telephonically and in person throughout June. Ibid. To date, RPS has produced close to 1800 pages of documents, around the same number as Plaintiffs. [Dkt. 109 at 8]. Plaintiffs have taken 8 depositions, including three Rule 30(b)(6) witnesses, with Rule 30(b)(6) testimony for more than 20 hours. Ibid.; [Dkt. 102 at 10]. With regard to the question of accuracy, RPS has produced a two-page Excel table with the overall number of consumer disputes it receives each year, broken down into six broad categories (including “other”) and with four general resolution categories (including, apparently, ”no resolution”). Id. at 1-2. III. Standard for Motion to Compel & Scope of Discovery A. Standard for Motion to Compel & Scope of Discovery

“[A] party may move for an order compelling disclosure or discovery. The motion must include a certification that the movant has in good faith conferred or attempted to confer with the person or party failing to make disclosure or discovery in an effort to obtain it without court action.” Fed. R. Civ. P. 37(a). Pursuant to District of Connecticut Local Rule 37(a), the movant must first confer with opposing counsel in person or via telephone and discuss the discovery issues between them

in order to arrive at a “mutually satisfactory resolution.” In the event a resolution is not reached, the movant must attach an affidavit certifying that, despite a good faith effort, he was unable to resolve the discovery issue with opposing counsel. Id. The Federal Rules of Civil Procedure set forth the scope of discovery:

“Parties may obtain discovery regarding any nonprivileged matter that is relevant to any party’s claim or defense and proportional to the needs of the case, considering the importance of the issues at stake in the action, the amount in controversy, the parties’ relative access to relevant information, the parties’ resources, the importance of the discovery in resolving the issues, and whether the burden or expense of the proposed discovery outweighs its likely benefit. Information within this scope of discovery need not be admissible in evidence to be discoverable. Fed. R. Civ. P. 26(b)(1); see Fed. R. Civ. P. 26(b)(2)(C). A district court has discretion in ruling on motions to compel. "A district court abuses its discretion only when the discovery is so limited as to affect a party's substantial rights." In re 650 Fifth Ave., 934 F.3d 147, 157 (2d Cir. 2019).

B. Claims and Defenses to which the Contested RFPs May Be Relevant i. Disparate Impact Claim To prove their disparate impact claim, Plaintiffs must show (1) “the occurrence of certain outwardly neutral practices” and (2) “a significant adverse or disproportionate impact on persons of a particular type produced by the defendant’s facially neutral acts or practices.” MHANY Mgmt., Inc. v. Cty. of Nassau, 819 F.3d 581, 587-88, 619 (2d Cir. 2016) (internal citation and quotation

marks omitted.) “Plaintiffs must present a statistical analysis to meet the second prong of their prima facie case.” Winfield v. City of N.Y., No. 15-cv-05236 (LTS) (KHP), 2018 U.S. Dist. LEXIS 56160, at *6 (S.D.N.Y. Mar. 29, 2018). Further, their expert analysis must link the policy at issue – here, the Automated Criminal Records Screening Policy – to the disparity shown. “A robust causality requirement ensures that ‘[r]acial imbalance… does not, without more, establish prima facie case of disparate impact’ and thus protects defendants from being held liable for racial disparities they did not create.” Tex. Dep't of Hous. & Cmty. Affairs v. Inclusive Cmtys. Project, Inc., 135 S. Ct. 2507, 2523, (2015) (citing Wards Cove Packing Co. v. Atonio, 490 U.S. 642, 653, (1989)).

Once Plaintiffs make out their prima facie case, the burden shifts to RPS to show the challenged practice is necessary to achieve a substantial, legitimate, non- discriminatory interest. Id. at 2514-15, 2522-23 and 24 C.F.R. § 100.500(c)(3). If RPS establishes a business justification, Plaintiffs may prevail by establishing a less discriminatory alternative exists which would meet the business needs. Tex. Dep't of Hous. & Cmty. Affairs 135 S. Ct. at 2515.

ii. Business Justification The parties contest how to characterize RPS’s business jusitifcation defense. Plaintiffs contend that RPS’s legitimate business purpose is that CrimSAFE is necessary to keep communities safe because it screens applicants in

a more consistent and accurate manner than a human decision-maker. E.g. [Dkt. 97 at 9 (citing RPS Expert Kacirk Report) (“CrimSafe is necessary to provide… insight into whether the applicant has the propensity for criminal behavior” and is “necessary to ensure the timely, objective and accurate categorization of criminal records.” )]. RPS responds that Kacirk did not independently assess the accuracy of RPS’s matching algorithm. Dkt. 102 at 2-3. RPS argues that, in fact, Kacirk concludes that the legitimate business purpose of CrimSAFE lies in its categorization and interpretation of criminal records. Ibid. (CrimSAFE is “necessary to ensure the timely and accurate categorization of criminal records in today’s environment of mobility where there are large volumes of electronic

applications and a large number of unique crimes across federal, state, and local ordinances.”). IV. Analysis A. RFP 30 (Dkt. 97) RFP 30 reads: “Produce all documents concerning any legitimate business purpose you assert that CrimSAFE fulfills.” In the Docket 97 motion to compel, Plaintiffs argue that, in response to RFP 30, the Court should compel RPS to produce a study of its record applicant matching logic and other documents in produced in Williams v. CoreLogic Rental Property Solutions, Inc., No. PX-16-58, 2016 WL 6277675 (D. Md. Oct. 26, 2016) and Witt v.

CoreLogic SafeRent, No. 3:15-cv-386 (E.D. Va. Dec. 14, 2016). [Dkt. 97 at 1-2]. In Williams, plaintiffs alleged that CoreLogic violated the FCRA by offering criminal record screening products that did not use procedures to assure maximum possible accuracy. Williams, 2016 WL 6277675, at *2-3. Specifically, the Williams plaintiffs alleged that RPS inaccurately attributed to them serious criminal records that belonged to others. Ibid. The Williams Court ordered RPS to produce entries on RPS’s privilege log, describing them a “internal commentary regarding CoreLogic’s knowledge of a matching problem.” Williams, 2016 WL 6277675 at *3- 5. Plaintiffs have consistently characterized the Williams discovery as describing

RPS’s “matching logic.” See, e.g., [Dkt. 97 at 7, 8 & n.7, 10, 11, 13]; [Dkt. 109 at 4, 7, 9]. Plaintiffs and RPS dispute the relevance of CrimSAFE’s matching logic, and thus the relevance of the Williams justification. Plaintiffs make three arguments about relevance. First, Plaintiffs argue that RPS has consistently alleged that its business justification for CrimSAFE is that CrimSAFE is more accurate and reliable than a human decision-maker and is needed to ensure community safety. E.g. [Dkt.

97 at 8-10 (quoting RPS’s expert’s report stating CrimSAFE is necessary to provide “relevant information on applicant’s criminal history (if any) and… insight into whether an applicant ahs the propensity for criminal behavior”)] , [Dkt. 109 at 4]. They argue that, since record-applicant matching accuracy affects CrimSAFE’s overall accuracy, the Williams production, which contains documents and internal production that speak to record-applicant matching accuracy, are therefore relevant for RPS’s defense and are within the scope of discovery. [Dkt. 97 at 9-10].

Second, Plaintiffs argue that, even if the record-applicant matching accuracy of CrimSAFE is not relevant to the parties’ defenses, it is at least “an important issue at stake” in this action, so records that speak to it are within the scope of discovery. See Fed. R. Civ. P. 26(b)(1). Finally, Plaintiffs argue that RPS has conceded that the accuracy of its product is at issue. [Dkt. 109 at 4]. They argue that RPS has produced the overall number of consumer complaints in response to Plaintiff’s RFPS requesting documents concerning CrimSAFE’s accuracy, and that RPS questioned Plaintiffs’ expert on the basis of her opinion that criminal record screening reports were not accurate. Ibid.

In response, RPS asserts that the requested information is not relevant because the accuracy of its matching algorithm is not at issue in the litigation. [Dkt. 102 at 7]. RPS argues that the accuracy of its algorithm’s matching of criminal records to rental applicants is not at issue because Plaintiffs do not claim that the criminal offense record for Mr. Arroyo was inaccurate, or that RPS’s matching procedures were unreasonable, or that RPS’s matching algorithm is inaccurate with respect to a group of individuals. [Dkt. 102 at 7-9]. RPS argues that the business justification

for CrimSAFE is not its accuracy broadly, but ratherits “categorization and interpretation of criminal records.” Id. at 11. After reviewing the briefing, the Court agrees with RPS that the accuracy of its algorithm’s matching of criminal records to rental applicants is not at issue. The reported criminal offense record for Mr. Arroyo was accurate: the record was Mr. Arroyo’s, and the record contained an accurate criminal proceeding status. The

accuracy of the algorithm’s matching of criminal records is not relevant to the claims. Further, the Court is not persuaded record-applicant matching is relevant to CrimSAFE’s business justification defense. RPS characterizes its business justification defense as the “categorization and interpretation of criminal records,” rather than its overall accuracy. The Court credits RPS’s characterization because,

given the underlying complaint, any business justification offered by RPS must have a nexus with the categorization of criminal records. That is, even if CrimSAFE did accurately match criminal offense records to applicants, such accurate matching would not provide a business justification for the claimed policy in this case, which regards categorization and type and amount of information disclosed. See [Dkt. 1 at ¶106]. For the same reason, the Court does not see how the requested discovery on matching are relevant to an “important issue at stake” in the action.

The Court is also not persuaded that RPS conceded the accuracy of its matching logic when it produced the overall numbers of consumer complaints in response to Plaintiffs’ RFPs requesting documents concerning CrimSAFE’s accuracy. RPS has consistently maintained that requests relating to the accuracy of its software are irrelevant, even while agreeing to provide consumer complaint statistics. E.g. [Dkt. 65-2 at 11-13 (RFP request responses objecting “to the extent that [this Request] calls for irrelevant information because the degree of accuracy of CrimSAFE reports… is not at issue in this action)]; [Dkt. 74 at 19]; [Dkt. 102 at 7]. Therefore, the Court does not find RPS’s production significant with regards to the relevance of matching logic accuracy.

Finally, Plaintiffs allege that RPS misled Plaintiffs as to the existence of these studies. Specifically, Plaintiffs allege that RPS represented that RPS had never studied the accuracy of CrimSAFE or its matching algorithm beyond responding to individual consumer complaints. [Dkt. 97 at 5]; [Dkt. 102 at 3 n.2]. RPS responds, saying that it did not so represent, and pointing out that the Williams order is public. [Dkt. 102 at 4 n. 3]. Without more information, the Court declines to consider

this dispute as a factor in its evaluation of the motion to compel. Therefore, the Court DENIES the Docket 97 motion to compel as to RFP 30. B. RFP 31 • RFP 31 reads: “Produce all documents concerning the scheme, schema documentation, or the like for any database or system that may reasonably be expected to contain information relating to CrimSAFE or the subject matter of this lawsuit, including information about consumers’ criminal records or their race or ethnicity.” [Dkt. 65-2 (Ex. A, Disputed Discovery Responses) at 1]. Plaintiffs argue that this information is necessary to necessary to determine whether the criminal records databases backing CrimSAFE include a disproportionate number of criminal records belonging to Black and Latino people, and to evaluate the claim that CrimSAFE provides more accurate results than other background check methods. [Dkt. 65-1 at 5]. RPS argues that this information is not relevant, that it has offered to answer questions about specific fields, and that it has already produced a data dictionary, the CrimSAFE structure, the number of criminal offense records in RPS’s database that contain information on the race of the offender. [Dkt. 74 at 10-11].

The Court DENIES the motion to compel as to RFP 31. RPS has already disclosed sufficient information for RPS to determine whether criminal records databases backing CrimSAFE include a disproportionate number of criminal records belonging to Black and Latino people. And, as the Court discusses above, the Court is not persuaded that CrimSAFE’s accuracy is relevant.

C. RFP 32 • RFP 32 reads: “Produce all documents concerning the nature, structure, appearance, format, or electronic searchability of any standard, enhanced, or custom report that can be generated by a database containing information relating to CrimSAFE or individuals’ criminal records or their race or ethnicity.” [Dkt. 65-2 at 2]. Plaintiffs argue that RPS has withheld information regarding the types of reports that its products can generate, including the reports that it advertised to WinnResidential. [Dkt. 80 at 4, Dkt. 80-2 (Ex. B, CoreLogic Proposal to WinnResidential) at 19]. RPS responds that it has already informed Plaintiffs about which systems are searchable in what manner, and that the reports it advertised to WinnResidential went beyond those regarding the CrimSAFe product, and that Plaintiffs have not identified further information that was undisclosed. [Dkt. 74 at 4], [Dkt. 84-1 at 2-3 (citing Dkt. 80-2 at 19)]. After reading RPS’s proposal to WinnResidential, the Court agrees with RPS that the reports it advertised to WinnResidential went beyond those relating to CrimSAFE. It therefore DENIES the motion to compel as to RFP 32. D. RFPs 47, 51, 53, and 54 • RFP 47 reads: “Produce all documentation relating to the CoreLogic RPS algorithm(s) that match a criminal offense record to a CrimSAFE category/subcategory and a degree/level of crime, including, but not limited to, documents relating to whether the model(s) is/are deterministic or probabilistic, missingness, techniques for dealing with missing data, techniques for dealing with misspellings or other data errors, and the variables used for matching.” [Dkt. 65-2 at 5].

• RFP 51 reads: “Produce all documents relating to the degree of accuracy of CrimSAFE reports.” [Dkt. 652 at 10].

• RFP 53 reads: “Produce the number of instances of which you are aware where a a CrimSAFE “decline” decision for a property (a) in Connecticut and (b) anywhere in the United States was based on a criminal record that was sealed, expunged, vacated, inaccurate, or erroneously attributed to a housing applicant.” [Dkt. 65-2 at 12].

• RFP 54 reads: “Produce the number of instances of which you are aware where a background report produced by CoreLogic RPES included information about criminal records that were sealed, expunged, vacated, inaccurate, or erroneously attributed to a housing applicant.” [Dkt. 65-2 at 13]. RPS has asserted that the requested information in each of these requests for production is not relevant because the accuracy of its matching algorithm is not at issue in the litigation. See [Dkt. 74 at 12 (re: RFP 47, “RPS’s matching algorithm has no relevance to this case…. There is no claim that RPS has inaccurately reported criminal records for Plaintiff or anyone else”); (re: RFPs 51, 53, 54, “there is no claim that the record reported for Mr. Arroyo was in any way inaccurate or did not reflect the current status of the proceeding a the time it was reported. None of these requests, therefore, relate…”). As discussed above in the analysis of RFP 30, the Court agrees. Therefore, the Docket 65 motion to compel is DENIED as to RFPs 47, 51, 53 and 54. E. RFPs 48, 50, 52 and 58

The Court follows the parties in considering Requests for Production 48, 50, 52, and 58 together as they each request extracts of data from the CrimSAFE database. • RFP 48 reads: Produce the total number of people of each race (based on the race code filed in the table ‘CRMMASTP’ in database ‘CRIMNALDB’) who have at least one criminal offense record that was in CoreLogic RPS’s database on April 26, 2016 that would trigger a CrimSAFE decline decision using WinnResidential’s CrimSAFE configuration for ArtSpace Windham on April 26, 2016, disregarding dates/events after April 26, 2016 associated with those criminal records (e.g. a subsequent disposition date, sentence date, custody date, etc.). Provide this number for people residing in (a) people residing in each zip code in Windham CT (which includes Willimantic, Windham Center, North Windham, and South Windham); (b) people residing in each zip code in Connecticut; and (c) people residing in each zip code in Windham County, CT. [Dkt. 65-2 at 5-6].

• RFP 50 reads: Produce the total number of people of each race (based on the race code filed in the table ‘CRMMASTP’ in database ‘CRIMINALDB’) residing in each zip code in Connecticut for whom CrimSAFE would return a decline decision using CrimSAFE settings uniformly set, for all categories and degrees/levels, at: A. 1 year; B. 3 years; C. 5 years; D. 7 years; E. the lesser of 10 years or the maximum permissible number for a given category/degree/level; and the F. the lesser of 20 years or the maximum possible number for a given category/degree level; G the lesser of 99 years or the maximum permissible number for a given category/degree/level; H. the maximum permissible number for each category/degree/level. [Dkt. 65-2 at 8].

• RFP 52 reads: For each CrimSAFE category/subcategory and degree/level of crime, produce the total number of people of each race (based on the race code filed in the table ‘CRMMASTP’ in database ‘CRIMINALDB’) with at least one offense record within that category and degree/level where the age of the offense (determined in the same manner as CrimSAFE) is less than (a) 1 year, (b) 3 years, (c) 5 years, (d) 7 years, (e) 10 years, (f) 20 years, and (g) 99 years. Produce this information for: (a) people residing in each zip code in Connecticut; (b) people residing in each zip code in Windham County, CT; and (c) people residing in each zip code in Windham CT (which includes Willimantic, Windham Center, North Windham, and South Windham). [Dkt. 65-2 at 11]. • RFP 58 reads: For each degree/level of crime in CrimSAFE (e.g. “Felony Convictions Only,” “Other Felony Charges,” etc.), produce the total number of people of each race (based on the race code field in the table ‘CRMMASTP’ in database ‘CRIMINALDB’) with at least one offense record within that degree/level where the age of the offense (determined in the same manner as CrimSAFE) is less than (a) 1 year, (b) 3 years, (c) 5 years, (d) 7 years, (e) 10 years, (f) 20 years, and (g) 99 years. Produce this information for (a) people residing in each zip code in Connecticut; (b) people residing in each zip code in Windham County, CT; and (c) people residing in each zip code in Windham CT (which includes Willimantic, Windham Center, North Windham, and South Windham). [Dkt. 65-2 at 15]. Plaintiffs argue that the requested that extracts prove disproportionate housing rental denials more directly than the public demographic data available to Plaintiffs’ expert since it would show, broken down by race, (a) whether the settings selected by the Arroyo Plaintiffs’ landlord would, within the local area that contains the likely applicant pool, disproportionately deny African-American and Latino rental applicants; (b) whether various CrimSAFE settings would disproportionately deny African-American and Latino rental applicants within the likely applicant pool; and (c) may also go to whether less discriminatory alternatives exists. [Dkt. 65-1 at 13]. Plaintiffs argue that the requested data is the most probative statistical evidence given that RPS has no record of the race of actual applicants except when the application was denied. Id. at 14; [Dkt. 80 at 7]. RPSs make three arguments in response: that the data requested by the Plaintiffs are irrelevant because they concern the hypothetical pool of potential applicants [Dkt. 74 at 14-15]; that the reports requested are irrelevant because they concern a pool of people which does not reflect the potential applicant pool [Dkt. 74 at 14], [Dkt. 84-1 at 5]; and that it is technically infeasible to respond to the request. While the Court agrees with the Plaintiffs that the requested reports are relevant,2 it finds that the requested data do not exist, and that RPS therefore cannot be compelled to produce them.

The question here is whether RPS would need to generate new documents to produce the requested reports or if instead it would only need to formulate new queries for an existing database. “While ‘a party should not be required to create completely new documents, that is not the same as requiring a party to query an existing dynamic database for relevant information. Courts regularly require parties to produce reports from dynamic databases, holding that the technical burden... of creating a new dataset for [ ] litigation does not excuse production.’”

2 Where, as in this case, reliable applicant flow data is not available, data on the pool of potential applicants is relevant. See Hazelwood Sch. Dist v. United States, 433 U.S. 299, 308-309 n.13 (1977) (holding data which demonstrates the pool of eligible candidates is appropriate to consider where reliable applicant flow data unavailable). RPS’s argument that any statistics must be based on actual applicants relies on mis-cited cases. [Dkt. 74 at 15]; see Davis v. N.Y. Housing Authority, No. 90 Civ. 628, 1997 U.S. Dist. LEXIS 10451 at *20 (S.D.N.Y. July 17, 1997) (quoting Dothard v. Rawlinson, 433 U.S. 321, 330 (1977)) (“There is no requirement… that a statistical showing of disproportionate impact must always be based on analysis of the characteristics of actual applicants." (emphasis added)); Fair Hous. Ctr. of the Greater Palm Beaches, Inc. v. Park Partners Residential, LLC, No. 16-80740-MIDDLEBROOKS, 2017 U.S. Dist. LEXIS 35546, at *18 (S.D. Fla. Mar. 9, 2017) (quoting Hallmark Developers, Inc. v. Fulton Cty., 466 F.3d 1276, 1286 (11th Cir. 2006)) (“statistics based on the general population [should] bear a proven relationship to the actual applicant flow.”). The requested data would be an estimate of the number of people of each race for whom disqualifying criminal records would likely be found in the potential applicant pool – that is, the number of individuals excluded for each race by CrimSAFE. The Court understands that this set of numbers, together with estimates of the whole potential applicant pool for each race, would generate an estimate of the rates of exclusion for the local potential applicant pool. N. Shore-Long Island Jewish Health Sys., Inc. v. MultiPlan, Inc., 325 F.R.D. 36, 51 (E.D.N.Y. 2018) (quoting Apple Inc. v. Samsung Elecs. Co., No. 12-CV-0630, 2013 WL 4426512, at *3 (N.D. Cal. Aug. 14, 2013)); see Hallmark v. Cohen & Slamowitz, Midland Funding LLC, 302 F.R.D. 295, 299 (W.D.N.Y. 2014) (“It is basic that in

responding to a document production request, pursuant to Fed.R.Civ.P. 34(a)…a party is not required to create documents meeting the document requests, only to produce documents already in existence.” (collecting cases)). RPS argues that it cannot apply the identified CrimSAFE categories against its underlying database of criminal records because more than 95% of all offenses in its database do not have an associated CrimSAFE “category,” and RPS would

have to create new coding, or mapping, to generate the requested reports. [Dkt. 74 at 15-16, 18; Dkt. 84 at 4-5]. Plaintiffs respond that RPS has not “suggest[ed] the difficulty of [coding to apply the categories] would be high or prohibitive,” and that RPS’s 30(b)(6) designee, Naeem Kayani, testified that the categories were included with the criminal record database. [Dkt. 80 at 8-9 (citing Dkt. 81-1 (Ex. 1, Kayani Dep.)]. RPS replies that Plaintiffs mischaracterize Kayani’s deposition testimony, as Kayani also testified that he did not know how the categorization process happened. [Dkt. 84 at 4-5]. RPS does not provide any estimates of the time or expense that would be needed to write such code.

While the Court remains uncertain about the difficulty of generating the reports, the Court finds that such reports would be new documents, rather than queries of old ones. According to RPS, how each offense is categorized— information necessary to answer the requests—does not currently exist for the vast majority of offenses in CrimSAFE’s database. Therefore, the Court DENIES the Motion to Compel as to RFPs 48, 50, 52, and 58.

F. RFP 49 • RFP 49 reads: For each property in Connecticut for which CrimSAFE has been used in the past 10 years, produce the name and address of the complex (e.g. ArtSpace Windham), the landlord or management company (e.g. WinnResidential), the CrimSAFE configuration, the total number of people of each race (based on the race code filed in the table ‘CRMMASTP’ in database CRIMINALDB’) for whom CrimSAFE has returned a decline decision in each calendar year between 2010 and 2019 (inclusive), and the total number of Connecticut residents of each race shoe criminal records would trigger a CrimSAFE decline decision based on the property’s current CrimSAFE configuration. [Dkt. 65-2 at 7]. For Connecticut-based properties and Connecticut-based offenders, RPS has produced the number of instances where a CrimSAFE product has identified a record based on a property’s settings, whether that record was associated with the offender’s race, and (if so) the race associated with that record, of the period from 2016. [Dkt. 74 at 17]. It has also provided the CrimSAFE settings for each Connecticut property that uses CrimSAFE. Ibid. Plaintiffs requests additional information identifying the specific property and property addresses, so that they can be prepared if RPS argues that different rental markets in different parts of the state would attract different demographic mixes of tenants. [Dkt. 80 at 10]. In light of the relevance of the additional information, the Court GRANTS in part and DENIES in part the motion to compel as to RFP 49. RPS must additionally produce to Plaintiffs the zip code of each property in Connecticut for which CrimSAFE has been used in the past 10 years. G. RFP 56 • RFP 56: Produce the number of criminal offense records in CoreLogic RPS’s criminal records database that (a) lack a disposition date, (b) lack a disposition date and a sentence date, (c) lack a disposition date, sentence date, and custody date, and (d) lack a disposition date, sentence date, and arrest date. Produce this information for criminal records associated with (a) Connecticut residents and (b) U.S. residents. Plaintiffs argue that documents related to RPS’s decision to prioritize the disposition date are important to issues such as RPS’s business justification and the viability of alternative methods, and that RFP 56 could illuminate how often and to what extent CrimSAFE treats a rental applicant differently than other applicants because of the lack of an available disposition date. [Dkt. 65-1 at 22]. RPS responds that the request is not linked to any actual reporting of records by RPS. [Dkt. 74 at 21]. It also responds that this RFP, like the RFPs relating to accuracy, is deficient because no claim in the case implicates specific dates associated with the underlying criminal records held in RPS’s database. Ibid. Plaintiffs do not reply to this response. [Dkt. 84-1]. Plaintiffs separately note that RPS has produced documents or deposition testimony which resolve other requests related to RPS’s decision to prioritize the disposition date. Ibid In light of RPS’s production of other related documents and the tangential relationship between this request and the claims in this case, the Court DENIES the motion to compel as to RFP 56. H. RFPs 72 and 73

• RFP 72: Produce any document that has ever been available at the URL http://corporate.corelogic.com/landing-pages/asset_upload_file691_14887.pdf. [Dkt. 65-2 at 20]. • RFP 73: Produce any document with the filename asset_upload_file691_14887.pdf. Ibid. The Court DENIES the motion to compel as to RFPs 72 and 73 as RPS represents that it has already performed extensive searches for these documents and Plaintiffs have represented that they will not further address the matter. [Dkt. 74 at 24, Dkt. 80 at 2]. V. Conclusion

For the reasons given, the Court GRANTS in part the First Motion to Compel [Dkt. 65] as to RFP 49, but otherwise DENIES it. The Court also DENIES the Second Motion to Compel [Dkt. 97] as to RFP 30. Also, as noted, the Court grants RPS leave to file a sur-reply. [Dkt. 84].

SO ORDERED. ________________________ District Judge Vanessa L. Bryant

Dated: January 24, 2020 at Hartford, Connecticut