Kravitz v. U.S. Dep't of Commerce

366 F. Supp. 3d 681

• Court: District Court, D. Maryland
• Decided: April 5, 2019
• Docket: Case No.: GJH-18-1041; Case No.: GJH-18-1570
• Status: Published

Opinion

GEORGE J. HAZEL, United States District Judge

TABLE OF CONTENTS

I. Findings of Fact...692

A. The Secretary's Decision (Administrative Record)...693

1. Genesis of Secretary Ross's Interest in Including a Citizenship Question...693

2. Manufacturing DOJ's VRA Rationale...695

3. Census Bureau Research, Analysis, and Recommendations...698

4. Secretary Ross and Staff Persist...701

B. The Secretary's Decision (Extra-Record Evidence)...706

C. Standing...712

1. Differential Decline in Census Participation...713

2. Differential Undercount of at Least 2 Percentage Points...716

3. Effect of a Differential Undercount...722

4. Additional Harms...732

II. Conclusions of Law...733

A. Standing...736

1. Injury to Individual Plaintiffs and Members of Organizational Plaintiffs...736

2. Causation and Redressability of Individual Plaintiffs' and Member Plaintiffs' Injuries...740

3. Organizational Plaintiffs' Standing...741

B. APA Claims...742

1. The Decision to Add the Citizenship Question Was Arbitrary and Capricious and Must Be Set Aside Under APA § 706(2)(A)-(D)....743

2. Secretary Ross's Stated Rationale was Pretextual in Violation of the APA....749

C. Enumeration Clause Claims...751 *691D. Equal Protection and 42 U.S.C. § 1985 Claims...752

E. Remedies...755

III. Conclusion...756

Plaintiffs in these related cases challenge Commerce Secretary Wilbur Ross's decision to include a citizenship question on the 2020 Census.² Plaintiffs contend that the addition of a citizenship question violates the Administrative Procedure Act (APA) and the United States Constitution. The LUPE Plaintiffs also allege that the question was added as part of a conspiracy to violate their civil rights in violation of 42 U.S.C. § 1985. The Court held a six-day bench trial between January 22 and January 31, 2019. Pursuant to Federal Rule of Civil Procedure 52(a)(1), this memorandum opinion contains the Court's findings of fact and conclusions of law, resolving Plaintiffs' claims. For the following reasons, the Court concludes: Plaintiffs have standing to assert their claims; the decision to add a citizenship question to the 2020 Census was arbitrary and capricious in violation of the APA; the Defendants' actions violate the Constitution by unreasonably compromising the distributive accuracy of the Census contrary to the Enumeration Clause's³ mandate; and Plaintiffs did not meet their burden to prove Defendants' actions violate the Due Process Clause or amount to a conspiracy to violate civil rights because Plaintiffs failed to show that the addition of the citizenship question was motivated by invidious racial discrimination.

Discovery here was coordinated with the discovery conducted in two consolidated cases before Judge Jesse M. Furman of the United States District Court for the Southern District of New York (together, the "New York Case"), and the parallel Census cases pending before Judge Richard G. Seeborg in the United States District Court for the Northern District of California (together, the "California Case"). See *692New York v. U.S. Dep't of Commerce , No. 1:18-cv-02921-JMF (S.D.N.Y.) (N.Y. Docket); State of Cal. v. Ross , No. 3:18-cv-01865-RS (N.D.C.A 2018) (California Docket). In those cases and here, Defendants objected to consideration of any discovery beyond the material in the Administrative Record. Given the limited time for appellate review of these matters prior to the start of the 2020 Census, this Court borrows from the approach taken by Judges Furman and Seeborg and will attempt to distinguish facts and conclusions of law based exclusively on the Administrative Record from those supported by extra-record evidence. Plaintiffs and Defendants have stipulated that the Administrative Record includes both 1) the initial materials compiled and submitted by Defendants, PX-1 (AR 1-1320) and 2) the supplemental materials, PX-3 to PX-14 (AR 1322-AR 13024), that Defendants produced in compliance with Judge Furman's order that they complete the initial administrative record, N.Y. Docket, ECF No. 199 (Jul. 5, 2018), after Secretary Ross's supplemental memorandum demonstrated that the initial record was deficient, PX-2 (AR 1321). Joint Stipulation Admitting Trial Exhibits, ECF No. 134.⁴

The Court will now begin its analysis by discussing the facts found to be proven by a preponderance of the evidence at trial, followed by the conclusions of law reached by applying those facts to the applicable legal principles, and, finally, the appropriate remedy.

I. Findings of Fact

1. To conduct the modern person-by-person Decennial Census count, the Census Bureau sends a short questionnaire with only a handful of questions to virtually every housing unit in the United States. ECF No. 103-1, Joint Stips. ¶ 20. For the 2020 Census, in addition to responding by mail, households will also be given the option to complete the questionnaire online or over the phone. Id. ¶ 22. If a household does not self respond, the Census Bureau then sends a staffer, known as an enumerator, to the housing unit to attempt to collect data via an in-person interview. Id. ¶ 23. This process is the first step in the Census Bureau's Non Response Follow Up (NRFU) operation. Id.

2. For the 2020 Census, the Census Bureau has proposed using administrative records to enumerate a limited number of those households for which there are high quality administrative data about the household if the initial NRFU visit does not result in the collection of complete data for that household. Id. ¶ 24. The Census Bureau will have enumerators attempt to recontact households for which high-quality administrative records do not exist. Id. ¶ 25. If by the third contact attempt a household has not responded, the housing unit will become "proxy-eligible." Id. ¶ 26. A proxy is someone who is not a member of the household-such as a neighbor, landlord, postal worker, or other knowledgeable person who can provide information about the unit and the people who live there. Id. ¶ 27. After three proxy attempts, a household becomes eligible for what is known as "whole-person imputation" or "whole-household imputation," in which the Census Bureau imputes the characteristics of the household, including in some circumstances the household member count. See id. ¶ 29. After the NRFU process is completed, the Census Bureau counts the responses from every household to determine the population count in each state. Id. ¶ 30. Data from the Decennial Census are reported down to the census block. Id. ¶ 31.

3. Between 1970 and 2000, the Census Bureau used both the short-form questionnaire *693(described above) and a long-form questionnaire (containing inquiries on the short-form questionnaire as well as additional questions), which was distributed to only a sample of the population. See U.S. CENSUS BUREAU, MEASURING AMERICA: THE DECENNIAL CENSUSES FROM 1790 TO 2000 , http://www2.census.gov/library/publications/2002/dec/pol_02-ma.pdf, 77-78, 84-85, 91-92, 99-100. During that time, the long-form questionnaire contained a citizenship question, but the short-form questionnaire did not. Id.

4. In 2005, the Census Bureau began administering the American Community Survey (ACS), a yearly survey of approximately 3.5 million households or 1 in every 38 households across the United States. ECF No. 103-1, Joint Stips. ¶ 72. The ACS replaced the long-form questionnaire. Id. ¶ 70. While the Decennial Census is intended to provide an actual enumeration for apportionment and other purposes, the ACS is intended to provide information on characteristics of the population, and the social and economic needs of communities. Id. ¶ 80. Unlike the short-form Census questionnaire, which asks only a handful of questions, the ACS asks more than 50 questions, including a citizenship question. Id. ¶ 73. The data collected by the ACS allows the Census Bureau to produce estimates of Citizen Voting Age Population (CVAP). Id. ¶ 77. CVAP data based on responses to the ACS are reported by the Census Bureau down to the census block group level. Id. ¶ 78.

A. The Secretary's Decision (Administrative Record)

5. On March 26, 2018, Secretary Ross issued a memorandum directing the Census Bureau to add a citizenship question to the 2020 Census. See PX-26 (AR 1313-20) (Ross Memo). He asserted that the decision was prompted by a December 12, 2017 letter from DOJ, requesting the addition of a citizenship question to facilitate enforcement of Section 2 of the Voting Rights Act (VRA). PX-26 (AR 1313). The Ross Memo states that DOJ needed to obtain CVAP block level data from Decennial Census data because "current data collected under the ACS are insufficient in scope, detail, and certainty to meet its purpose." PX-26 at 2 (AR 1314). Secretary Ross further states that the DOJ's "need for accurate citizenship data and the limited burden that the reinstatement of the citizenship question would impose outweigh fears about a potentially lower response rate." Id. at 5 (AR 1317). As described in more detail below, the genesis of the DOJ letter demonstrates the VRA rationale was a pretext, and the statements in the Ross Memo contradict the unanimous opinion of the Census Bureau that DOJ's stated goal could be achieved through a superior alternative and that adding a citizenship question to the Census would result in less accurate and less complete citizenship data while harming the overall quality of Census data and increasing costs and respondent burden. PX-22 (AR 1277); PX-132 (AR 9812); PX-102 (AR 5500); PX-147 (AR 11634); PX-100 (AR 5473).

1. Genesis of Secretary Ross's Interest in Including a Citizenship Question

6. The Senate confirmed Secretary Ross on February 27, 2017, see 163 CONG. REC. S1421, S1455 (2017), and by March 10, 2017 the Secretary's interest in a citizenship question had surfaced. PX-55 (AR 2521). On that day, Deputy Chief of Staff and Director of Policy Earl Comstock emailed Secretary Ross in response to a question the Secretary had posed regarding the Census with confirmation that "undocumented residents (aliens) in the 50 states" are "included in the apportionment population counts." Id at 1 (AR 2521);

*694ECF No. 103-1, Joint Stips. ¶ 3. The email also included the text of a Wall Street Journal article titled "The Pitfalls of Counting Illegal Immigrants." PX-55 (AR 2521).

7. Although Secretary Ross spoke with various "senior Administration officials" and "other governmental officials" about adding a citizenship question to the Census around this time, PX-2 (AR 1321), the record is largely void of details regarding these conversations. The record does show, however, that in early April 2017, White House Chief Strategist Steve Bannon contacted Secretary Ross to ask the Secretary if he would speak to Kansas Secretary of State Kris Kobach about adding a citizenship question to the Census. PX-19 (AR 763); PX-58 (AR 2651). Thereafter, complying with Bannon's request, Kobach and Secretary Ross discussed Kobach's ideas about adding a citizenship question to the Decennial Census, and "the fact that the US census does not currently ask respondents about their citizenship." PX-19 at 2 (AR 764). Secretary Ross and Kobach discussed the potential effect adding "one simple question" to the Census would have on "congressional apportionment." Id.

8. On April 13, 2017, Comstock emailed Mark Neuman, the point person for President Trump's transition team on the Census, asking when the Census Bureau must "notify Congress regarding the questions that will be on (A) the ACS and (B) the decennial Census." PX-87 (AR 3709). Neuman responded that the notification to Congress "relating to questionnaire content additions for 2020 Census just took place." Id. The Department had transmitted to Congress the Census Bureau report entitled Subjects Planned for the 2020 Census Program two weeks earlier. PX-1 (AR 202-13). That report included five planned "subjects" for the 2020 Census: age, gender, race/ethnicity, relationship, and tenure (owner/renter). Id. It did not include citizenship. See id. Neuman reassured Comstock that there would be "another opportunity" to notify Congress about Census questions "next year." PX-87 (AR 3709).

9. On April 20, 2017, the Secretary emailed Comstock and Senior Advisor and Chief of Staff to Secretary Ross, Wendy Teramoto, noting that then-Census Director John Thompson had an upcoming meeting with the Census National Advisory Committee on Racial, Ethnic and Other Populations (NAC), and emphasizing: "We must get our issue resolved before this!" PX-81 (AR 3694) (emphasis in original). Secretary Ross's personal assistant sent this note from her account but explained in the subject line that she had tried unsuccessfully to send it from Secretary Ross's email and that "it's from him." Id.

10. Secretary Ross then complained in a May 2, 2017 email to Comstock: "Worst of all they emphasize that they have settled with congress on the questions to be asked. I am mystified why nothing ha[s] been done in response to my months old request that we include the citizenship question. Why not?" PX-88 (AR 3710). Comstock responded:

I agree Mr. Secretary.

On the citizenship question we will get that in place.... We need to work with Justice to get them to request that citizenship be added back as a census question, and we have the court cases to illustrate that DOJ has a legitimate need for the question to be included. I will arrange a meeting with DOJ staff this week to discuss.

PX-88 (AR 3710). Immediately after this exchange, Comstock and other Commerce officials began implementing the plan to solicit from DOJ a request to add a citizenship question. PX-84 (AR 3701); PX-537 (AR 12756); PX-51 (AR 2462).

*6952. Manufacturing DOJ's VRA Rationale

11. On May 4, 2017, Comstock emailed the White House Liaison at DOJ requesting a call. PX-51 (AR 2462); PX-537 (AR 12756). They then "met in person to discuss the citizenship question" and Comstock was put in touch with James McHenry, the head of DOJ's Executive Office of Immigration Review. PX-537 (AR 12756). After speaking with Comstock "several times," however, McHenry advised Comstock that "Justice staff did not want to raise the [citizenship] question given the difficulties Justice was encountering in the press at the time (the whole Comey matter)." Id.

12. Because DOJ was reluctant to request a citizenship question, McHenry referred Comstock to Gene Hamilton at the Department of Homeland Security (DHS).Id. But after "several phone calls" with Hamilton, Comstock was told that DHS "really felt that it was best handled by the Department of Justice." Id. After both DOJ and DHS declined to pursue a citizenship question, Comstock asked James Uthmeier, who had recently moved from Secretary Ross's office to the Office of General Counsel, to look into "how Commerce could add the question to the Census itself." Id.

13. The Secretary also involved David Langdon, a Senior Policy Advisor who reported to Comstock. See ECF No. 103-1, Joint Stips. ¶ 11; PX-150 (AR 12541). In late May 2017, Langdon had an unusually lengthy meeting with the Secretary about the Census. PX-150 at 2 (AR 12542). Langdon reported that the Secretary was interested in Census subjects and "puzzled why citizenship is not included in 2020." Id. at 1 (AR 12541). Following the meeting, Langdon requested further information from Census Bureau staff, including Census Bureau Chief of Decennial Communications and Stakeholder Relations Burton Reist, regarding "the criteria used to pick topics for 2020 versus ACS. Say, citizenship." Id. Langdon collected this and other citizenship-related materials from Reist, including a 1988 internal DOJ memorandum asserting that the Constitution does not require counting of undocumented residents in the Decennial Census. PX-618 (AR); PX-619 (AR).⁵ On the same day as Langdon's meeting with the Secretary and communications with Census Bureau staff, Langdon emailed Comstock a note with the subject line "Counting of illegal immigrants," in which he informed Comstock that the Census is required to count all persons: "Long story short," he wrote, "is that the counting of illegal immigrants (or of the larger group of noncitizens) has a solid and fairly long legal history." PX-565 (AR). He also wrote, "a second piece of interest in [sic] a Bush 41 era DOJ opinion that proposed legislation to exclude illegal aliens from the decennial census was illegal." Id. PX-565 (AR). Around 11:00 PM that same day, Langdon made an apparently urgent request to Acting Associate Director of the 2020 Census Lisa Blumerman to respond to his inquiry related to a citizenship question, asking for a response, "[i]deally this evening." PX-150 at 1 (AR 12541).

14. On July 14, 2017, Kobach emailed Secretary Ross to follow up on their prior telephone discussion about adding a citizenship question. PX-19 (AR 763). Kobach wrote that the lack of a citizenship question "impairs the federal government's ability to do a number of things accurately," and "also leads to the problem that *696aliens who do not actually 'reside' in the United States are still counted for congressional apportionment purposes." Id. at 2 (AR 764). Kobach urged that it was "essential" to add a citizenship question to the 2020 Census and offered "any assistance that I can provide to accomplish the addition of this question." Id. Kobach specifically suggested the addition of a "slight variant" of the citizenship question that appears on the ACS. Id. Kobach did not mention the VRA or any rationale tied to VRA enforcement. Id.

15. On July 21, 2017, Kobach called and then emailed Wendy Teramoto, re-copying his July 14 email and noting his previous discussion about the citizenship question with the Secretary was "at the direction of Steve Bannon." Id. at 1 (AR 763). On July 25, 2017, Secretary Ross had a further telephone conversation with Kobach concerning the addition of a citizenship question to the 2020 Decennial Census, in which Teramoto and Deputy Chief of Staff Israel Hernandez participated. Id.⁶

16. On August 8, 2017, Secretary Ross emailed Comstock to ask whether he was on a morning call about the Census, and to check in about whether DOJ had still "not come to a conclusion" about requesting the citizenship question. PX-97 (AR 4004). He asked Comstock to "please let me know your contact person and I will call the AG." Id. Comstock responded, "I have two attorneys in the DoC [ (i.e. Department of Commerce) ] General Counsel's office working on it." Id. The next day, Comstock advised the Secretary that "we are preparing a memo and full briefing for you on a citizenship question. The memo will be ready by Friday." PX-523 (AR 12476). Comstock cautioned: "Since this issue will go to the Supreme Court we need to be diligent in preparing the administrative record." Id. Secretary Ross responded on August 10, 2017 that the team "should be very careful, about everything, whether or not it is likely to end up in the SC." Id. The following day, Comstock and Uthmeier exchanged edits on briefing materials regarding the citizenship question and later sent the memorandum to the Secretary and Teramoto. See PX-50 (AR 2461); PX-146 (AR 11362).⁷ Uthmeier told Comstock that he looked forward to discussing his "new ideas/recommendations on execution," and observed that "[u]ltimately, we do not make decisions on how the [citizenship] data will be used for apportionment, that is for Congress (or possibly the President) to decide." PX-607 at 2 (AR). He added, "I think that's our hook here." Id.

17. On September 1, 2017, Secretary Ross wrote to Comstock and Teramoto that he had "received no update" on "the issue of the census question." PX-45 (AR 2424); PX-96 (AR 4002). Comstock responded, "Understood. Wendy and I are working on it." PX-96 at 1 (AR 4002). On September 6, 2017, Secretary Ross met with his senior staff including the new-Commerce General Counsel Peter Davidson, Under Secretary Karen Dunn Kelley, Teramoto, Comstock, and Uthmeier to discuss the citizenship question. PX-31 (AR 1411); PX-35 (AR 1996); PX-46 (AR 2426). The next day, Comstock emailed Uthmeier and Davidson stating that Secretary Ross "would like an update on progress since the discussion yesterday regarding the citizenship question." PX-37 (AR 2034); PX-49 (AR 2459). After receiving Uthmeier's response (withheld as privileged), Comstock *697reiterated "the Secretary is asking for progress on this." PX-49 at 1 (AR 2459). Davidson then wrote to Comstock, Uthmeier, and Teramoto expressing concern about directly contacting Kobach-whom Secretary Ross had mentioned in the September 6, 2017 meeting. PX-614 at 3 (AR). Davidson recommended instead contacting a "trusted" advisor, such as Neuman, before doing "anything externally." Id. Following Davidson's suggestion, on September 8, 2017, Uthmeier emailed Neuman on September 8, 2017, asking if Neuman had a "few minutes" that morning "to discuss" the Census. PX-38 (AR 2051_0001).

18. That same day, Comstock sent Secretary Ross an informal memorandum summarizing his so-far unsuccessful efforts to find an agency willing to request the addition of a citizenship question. See PX-537 (AR 12756). He reported on his discussions with McHenry at DOJ and Hamilton at DHS. Id. Around this time, impatient with the Department's failed efforts, Secretary Ross implemented his earlier promise to take the matter of inquiring whether DOJ would request inclusion of a citizenship question, PX-2 (AR 1321), into his own hands by involving Attorney General Jeff Sessions. PX-67 (AR 2653).

19. On September 13, 2017, John Gore, the then-Acting Assistant Attorney General for Civil Rights, emailed Teramoto to introduce himself and request a call "about a DOJ-DOC" issue. PX-68 (AR 2659); PX-59 (AR 2628); PX-60 (AR 2634).⁸ The Administrative Record does not specifically reveal what prompted Gore to make this contact, despite DOJ's prior reluctance to work on the citizenship question issue. However, Gore's outreach occurred about a month after Secretary Ross indicated he would "call the AG" to make progress with DOJ, PX-97 (AR 4004), and days before the Attorney General's assistant scheduled a call between Secretary Ross and the Attorney General, PX-63 (AR 2639); PX-67 (AR 2653).⁹ Regarding the substance of the call, the Attorney General's assistant told Teramoto: "From what John [Gore] told me, it sounds like we can do whatever you all need us to do and the delay was due to a miscommunication. The AG is eager to assist." PX-67 (AR 2653); PX-68 (AR 2659).

20. On September 18, 2017, Secretary Ross and Attorney General Sessions spoke about the citizenship question. PX-62 (AR 2637). It is not clear from the record whether this was their first communication about the Census. On December 12, 2017, DOJ submitted a letter signed by the General Counsel for the Justice Management Division, Arthur Gary, to the Census Bureau requesting that a citizenship question be asked on the 2020 Census. PX-32 at 3 (AR 1525). Gary had previously notified the Department of Commerce in November 2016 that DOJ had no need to amend the content of the ACS, except that DOJ wanted the Census Bureau to consider adding a new topic to the ACS relating to LGBT populations. PX-17 at 1 (AR 311). DOJ attached to this earlier request a spreadsheet reflecting "the legal authority supporting the necessity for the collection of this information." Id. at 2-5. Gary's *698November 2016 letter did not refer to citizenship and did not indicate any need by DOJ for additional data concerning citizenship for any purpose. See id. According to an earlier June 2014 letter from Gary to the Department of Commerce's General Counsel describing what Census data DOJ uses and confirming its continued use of that data, the "lowest geography" for which DOJ needs citizenship data is the "Census block group" level. PX-1 at 280 (AR 280).

21. The December 2017 DOJ letter, however, formally requested the Census Bureau "reinstate on the 2020 Census questionnaire a question regarding citizenship," asserting for the first time that citizenship data are "critical" to DOJ's enforcement of Section 2 of the Voting Rights Act (VRA), 42 U.S.C. § 1973. PX-32 at 1 (AR 1525). The DOJ letter noted that, from 1970 to 2000, the Census Bureau collected citizenship data through the "long form" questionnaire sent to a sample of the U.S. population in conjunction with the decennial census. Id. at 2. As of the 2010 Census, the "long form" was discontinued and the ACS replaced it as the source of citizenship data used by DOJ for VRA enforcement. Id. The DOJ Letter claimed that the ACS "does not yield the ideal data for such purposes," noting that unlike Decennial Census data, ACS data are not "reported to the Census block level." Id. at 2-3. These comments imply that DOJ believed the previously used long-form questionnaire was superior to the ACS in the quality of the citizenship data generated. The DOJ Letter did not acknowledge, however, that the long-form questionnaire was also a sample survey subject to margins of error and resulted in citizenship data reported only down to the block group level, as with the current ACS.

22. Though DOJ asserted that "decennial census questionnaire data regarding citizenship" would be "more appropriate for use" than ACS citizenship data, it did not claim that Census citizenship data were "necessary." Id. at 3. The Ross Memo would later characterize DOJ's request as a "need" for more accurate data, PX-26 at 5 (AR 1317), but nothing in the DOJ Letter or in the Administrative Record provides evidence that the DOJ or any other plaintiff lost or was unable to bring a VRA enforcement action due to the lack of block-level citizenship data from the Decennial Census or the insufficiency of CVAP data based on the ACS. See PX-32 (AR 1525).

3. Census Bureau Research, Analysis, and Recommendations

23. Upon receiving the DOJ letter, the Census Bureau became aware for the first time that such content was even under consideration. The Administrative Record contains no evidence that anyone at the Census Bureau knew of the steps taken by the Secretary and his staff to find a federal agency to request the addition of a citizenship question for the 2020 Census before the DOJ letter was sent. Nor is there any evidence in the Administrative Record indicating that any Census Bureau employee was informed by the Secretary or his staff about their direct involvement in arranging for and developing the DOJ Letter, at any time prior to the Secretary's March 28, 2018 Memorandum announcing the decision.

24. After receipt of the DOJ letter, Acting Director Jarmin and the Census Bureau's Chief Scientist, Dr. Abowd, assembled a team of Census Bureau experts (nicknamed the "SWAT team") to evaluate DOJ's request and formulate a response. PX-74 (AR 3354); see also PX-22 (AR 1277); PX-102 (AR). Between December 2017 and March 2018, the SWAT team conducted extensive research and statistical analysis to provide the Secretary with *699the Census Bureau's expert conclusions and recommendation as to how best to meet the DOJ's asserted need for block-level citizenship data. Based on this research and analysis, the Census Bureau repeatedly, consistently, and unanimously recommended against adding a citizenship question to the 2020 Decennial Census. PX-147 (AR 11634); PX-100 (AR 5473); PX-22 (AR 1277); PX-132 (AR 9812); PX-25 (AR 1308).

25. In particular, the Census Bureau's technical analysis concluded that DOJ's stated goals with respect to VRA enforcement could be accomplished more efficiently and effectively through the use of administrative records rather than the addition of a citizenship question to the 2020 Census. PX-147 (AR 11634); PX-100 (AR 5473); PX-22 (AR 1277); PX-132 (AR 9812). Moreover, the Census Bureau's expert team concluded that the addition of a citizenship question to the 2020 Census would objectively harm data quality while imposing greater expense and burden on the public than reliance on administrative records. PX-132 at 5 (AR 9816). The technical analysis compared "the self-response rates for the same household address" on the 2010 Census and the 2010 ACS and found that for the same universe of housing units, self response to the ACS, which contained a citizenship question, was lower than it was to the Decennial Census, which did not. PX-22 at 4-5 (AR 1280-81); PX-102 at 6-7 (AR 5505-06).

26. Led by Dr. Abowd, the Census Bureau SWAT team provided the Secretary with a memorandum dated December 22, 2017 (the "December 22 Memorandum"), analyzing "Alternative Sources of Citizenship Data for the 2020 Census." PX-147 (AR 11634). The December 22 Memorandum identified several sources of administrative records that the Census Bureau concluded could provide "a more accurate measure of citizenship" in a more "cost efficient" manner than adding a citizenship question to the Decennial Census. Id. at 11 (AR 11644). The December 22 Memorandum concluded that adding a citizenship question to the 2020 Decennial Census was likely to decrease self-response rates, particularly in households with noncitizens; increase costs; and produce lower quality citizenship data. Id. at 6-12 (AR 11639-11645. Accordingly, Dr. Abowd advised Dr. Jarmin: "[W]e recommend that the citizenship data for Department of Justice Voting Rights Act enforcement be obtained through the use of administrative records and not through the addition of a question to the decennial census instrument." PX-1130 at 1 (AR 11646).

27. Dr. Jarmin emailed Gary that day to convey the Census Bureau's determination that "the best way to" achieve DOJ's stated goal "would be through utilizing a linked file of administrative and survey data the Census Bureau already possesses." PX-71 (AR 3289). Dr. Jarmin further advised that this method of augmenting currently available ACS data "would result in higher quality data produced at lower cost" than the addition of a citizenship question to the 2020 Census. Id. The Census Bureau therefore recommended against adding a citizenship question to the 2020 Decennial Census. Id.

28. In the same email, Dr. Jarmin suggested that Census and DOJ technical experts meet to discuss the details of the Census Bureau's proposal. Id. Dr. Jarmin twice followed up with Gary to set up a meeting between Census Bureau and DOJ technical experts over the next few weeks. PX-101 (AR 5489). A meeting was eventually scheduled but was cancelled by Gary before it took place. PX-124 (AR 9193). As the deadline to report planned 2020 Decennial Census questions fast approached, Gary told Dr. Jarmin that he had spoken *700with DOJ Leadership and that they believed DOJ's letter fully described the department's request, and they did not want to meet. PX-3 (AR 3460).

29. In the meantime, after the December 22, 2017 initial recommendation, the Census Bureau experts continued to analyze both the impact of adding a citizenship question to the Decennial Census and potential alternative sources of citizenship data in addition to the ACS. In a January 3, 2018 memorandum to the Secretary, Dr. Abowd elaborated on the research and analysis discussed in the December 22 Memorandum. PX-100 (AR 5473). The January 3 Memorandum presented three alternatives, Alternatives A, B, and C, for meeting the DOJ request: (A) "[m]aintain the status quo for data collection, preparation and publication"; (B) "[a]dd a citizenship question to the 2020 Census questionnaire"; and (C) do not add a citizenship question to the Decennial Census, and provide DOJ with CVAP data using available administrative records. Id.

30. The January 3 Memorandum recommended adopting Alternative C because using administrative records rather than adding a citizenship question would provide higher quality citizenship data to DOJ and be less costly to implement. Id. at 3. In particular, the Census Bureau estimated that adding a citizenship question would cost an additional $ 27.5 million based on the need for increased NRFU operations and cause at least a 5.1 percentage point decline in self-response among noncitizen households, thereby leading to an estimated minimum of "154,000 fewer correct enumerations." Id. at 2 (emphasis in original).

31. On January 4, 2018, Dr. Abowd emailed multiple Census Bureau officials advising that Dr. Jarmin "reports that he has discussed this with the Under Secretary [Karen Dunn Kelley], and she agrees with the recommendation of Alternative C." PX-121 at 1 (AR 9008). While Dr. Abowd noted that "Alternative A [i.e., no change] remains a possibility as well," he made no mention of Alternative B [i.e., adding a citizenship question]. Id . A January 19 subsequent memorandum, prepared by Dr. Abowd for Secretary Ross and transmitted to him "through" acting Census Bureau Director Jarmin and Undersecretary Kelley, expanded on the Census Bureau's technical analysis of Alternatives A through C. PX-22 at 1 (AR 1277). The January 19 Memorandum's conclusion was the same as that stated in the January 3 Memorandum, and advised the Secretary that "Alternative C best meets DoJ's stated uses, is comparatively far less costly than Alternative B, does not increase response burden, and does not harm the quality of the census count." Id. Conversely, adding a citizenship question to the 2020 Census "is very costly, harms the quality of the census count, and would use substantially less accurate citizenship status data than are available from administrative sources." Id. The January 19 Memorandum concluded that, even if block-level CVAP data could be improved through a Census citizenship question, there would be "serious quality issues remaining" and the addition of a citizenship question would cause "[m]ajor potential quality and cost disruptions." Id. at 2.

32. The Census Bureau's technical analysis showed that a citizenship question on the 2020 Census would decrease self-response rates disproportionately among noncitizen households, causing a 5.1 percentage point decline in self-responses from households containing at least one noncitizen-an estimate that the Bureau characterized as conservative. Id. 4-6. The January 19 Memorandum noted that item nonresponse rates (i.e., nonresponse to a particular question on the questionnaire) for the citizenship question "are much *701greater than the comparable rates for other demographic variables like sex, birthdate/age, and race/ethnicity." Id. at 4. Between 2013 and 2016, item nonresponse rates of Hispanics were approximately double that of non-Hispanic whites. Id. Similarly, the "breakoff rate" for the citizenship question (i.e., the rate at which households stop answering the questionnaire at a particular question) is far higher for Hispanic respondents (36 percent) than for non-Hispanic whites (4 percent). Id. at 5. The analysis also revealed that a high percentage of noncitizens provide erroneous answers to the ACS citizenship question. Specifically, "[i]n 2010 and 2016, individuals for whom the administrative data indicate noncitizen respond citizen in 32.7% and 34.7% of the ACS questionnaires, respectively." Id. at 8. Further, the material decline in noncitizen household self-response rate would result in more such households becoming subject to NRFU procedures, and data obtained in NRFU have greater rates of erroneous enumeration and whole-person imputation. Id. at 5-6. Adoption of Alternative C would provide more accurate citizenship data than Alternative B, which would cause harm to the data quality of the Census count. Id.

33. On February 12, 2018, Census Bureau staff met with Secretary Ross to discuss the January 19 Memorandum. PX-127 (AR 9450). According to the Administrative Record, the February 12 meeting is the only in-person meeting between Secretary Ross and Census Bureau staff concerning the citizenship question. Despite the findings detailed in the December 22, January 3, and January 19 memos and communicated to Secretary Ross at the February 12 meeting, the Ross Memo would go on to repeatedly assert incorrectly that the Census Bureau and concerned stakeholders could not "document that the response rate would in fact decline materially" because of a citizenship question. PX-26 at 3 (AR 1315); see also id. at 4, 5, 6.

4. Secretary Ross and Staff Persist

34. Faced with the Census Bureau's analysis that DOJ's stated goals could be best achieved without adding a citizenship question and that a citizenship question would be costly and harmful to data quality, Secretary Ross nonetheless persisted in his efforts to add a citizenship question to the 2020 Census.

35. On February 12, 2018, the same day that the Secretary conferred with the Census Bureau about its analysis and recommendation, Kobach sent a letter to the Secretary, formally requesting that the Secretary add the citizenship question to the 2020 Census. PX-1 at 1141 (AR 1141). Unlike his earlier requests, see e.g. , PX-19 at 2 (AR 764), Kobach now endorsed the VRA enforcement rationale as a reason to add the citizenship question to the 2020 Census, and referenced concerns about voter fraud. PX-1 at 1141 (AR 1141). Kobach also repeated his interest in including the question for "election-related reasons" but he was not explicit about his earlier congressional-apportionment rationale. Id. Instead, he vaguely referred to the fact that lawful immigrants "are part of the population of continuous residents in a state, and are not temporarily or illegally present," implying that undocumented immigrants are not "residents" for purposes of apportionment. Id.

36. Meanwhile, apparently dissatisfied by the Census Bureau's consensus, the Secretary next directed the Census Bureau to analyze a fourth alternative, which he devised, and which is referred to in the Administrative Record as "Alternative D." PX-132 at 2 (AR 9813). Alternative D would "combin[e] Alternative B (asking the citizenship question of every household on the 2020 Census) with Alternative C (do *702not ask the question, link reliable administrative data on citizenship status instead)." Id.

37. At the Secretary's request, the Census Bureau conducted additional technical research and analysis on Alternative D and the weaknesses of Alternative C on its own, and Dr. Abowd sent an additional recommendation memorandum (the "March 1 Memorandum"), through Jarmin, Kelley, and Lamas to Secretary Ross. PX-25 (AR 1308); PX-132 (AR 9812). In the March 1 Memorandum, the Census Bureau continued to recommend against the addition of a citizenship question. Id. Specifically, the March 1 Memorandum concluded that Alternative D would increase the burden on "100 percent of respondents" while "result[ing] in poorer quality citizenship data than Alternative C" and having "all the negative cost and quality implications of Alternative B" described in the January 19 Memorandum. PX-25 at 5 (AR 1312).

38. The conclusion by the Census Bureau that adding a citizenship question would increase burden on "100 percent of respondents" was consistent with its earlier determination "that a question on citizenship would lead to some decline in overall self response because it would make the 2020 Census modestly more burdensome in the direct sense, and potentially much more burdensome in the indirect sense that it would lead to a larger decline in self-response for noncitizen households." PX-22 at 5 (AR 1281). Yet, the Ross Memo would ultimately state incorrectly that the addition of a citizenship question imposes no additional burden on citizens and there is "limited empirical evidence to support" the view that "recipients are generally less likely to respond to a survey that contained more questions than one that contained fewer." PX-26 at 6 (AR 1318); see also id. at 5 (AR 1317) (citizenship question would be "no additional imposition" for citizens and for the 70% of noncitizens who respond to the ACS citizenship question accurately); see also id. (stating that that the reinstatement of the citizenship question would impose a "limited burden"). To the contrary, the Census Bureau found that "inclusion of a citizenship question on the 2020 Census Questionnaire is very likely to reduce self-response rate, pushing more households into NRFU. Not only will this likely lead to more incorrect enumerations, but it is expected to increase the number of persons who cannot be linked to the administrative data because NRFU PII [i.e., personal identifying information] is lower quality than self-response data." PX-25 at 4 (AR 1311). The Census Bureau characterized the quality of any survey data collected through a citizenship question on the decennial Census as "suspect." Id. at 5 (AR 1312).

39. The Census Bureau elaborated on the differences between Alternative C and Alternative D in a further memorandum titled, "Summary of Analysis of the Key Differences Between Alternative C and Alternative D." PX-24 (AR 1304) (Key Differences Memo). This memorandum set forth the Census Bureau's additional conclusion that, in comparison to Alternative C, Alternative D will lead to a larger number of people for whom the Census cannot match an administrative record and whose answers must be produced by a model. PX-24 (AR 1304). Further, the memorandum explained:

Under Alternative C, there will be error in the administrative records, but we believe these to be relatively limited due[ ] to the procedure follow[ed] by SSA, USCIS and State. In both Alternative[s], the modeled cases will be subject to prediction error. ... Alternative D has an additional source or error, response error. This is where 2020 respondent give the incorrect status. Statisticians *703often hope these error[s] are random and cancel out. However, we know from prior research that citizenship status responses are systematically biased for a subset of noncitizens. Response error is only an issue in alternative D.

PX-24 at 2 (AR 1305). Because data quality would be better under Alternative C, a higher number of individuals could be linked to more reliable administrative records than under Alternative D. Id.

40. The Census Bureau anticipated needing to model the citizenship status of approximately 35 million people under Alternative C. Id. Under Alternative D, the Bureau expected 35.4 million people not to respond to a citizenship question. Id. Further, the Census Bureau asserted that of the approximately 295 million individuals for whom self responses to a citizenship question would likely exist, 263 million people would respond with citizenship data that matches administrative records. Id. Thus, taking these numbers together, Alternative D would provide no improvement to the citizenship data available under Alternative C for 298.4 million individuals or 90.4% of the population. See id.

41. The Census Bureau also concluded that, under Alternative D, a group of approximately 22 million people would respond to a citizenship question but would not be linked to existing administrative records because of self-response errors. Id. As a result, the citizenship data for this group would be less accurate under Alternative D than under Alternative C. Id. Thus, for 6.7% of the population, Alternative D will produce lower quality data than Alternative C because the Census Bureau would have to use survey responses that are generally less accurate than the imputation methods the Census Bureau would deploy under Alternative C. Id. And for the remaining 2.9% of the population, Alternative D creates a problem that does not exist under Alternative C, due to conflict between survey and administrative data as to citizenship. Id. In sum, the Census Bureau concluded that based on the relevant metrics, Alternative C was an objectively superior method than Alternative D to achieve DOJ's goal and limit harm to the overall Census. See id.

42. Despite the Census Bureau's finding that Alternative C was objectively superior to Alternative D in terms of data quality, cost, and respondent burden, Secretary Ross rejected Alternative C in favor of Alternative D. PX-26 at 4 (AR 1316). The Secretary supported his choice of Alternative D with the assertion that it could "eliminate the need for the Census Bureau to have to impute an answer for millions of people." Id. at 5. But the Census Bureau explained that while neither Alternative C nor Alternative D would yield perfect results, the errors caused by Alternative D would likely outweigh the margin of error caused by modeling under Alternative C. The Census Bureau's analysis thus does not support Secretary Ross's decision to prioritize reducing the need for the Census Bureau to impute answers. PX-24 at 2 (AR 1305).

43. The Ross Memo also asserts incorrectly that Alternative D "would maximize the Census Bureau's ability to match the decennial census responses with administrative records." PX-26 at 4 (AR 1316). This assertion is directly contradicted by the Census Bureau's analysis. PX-25 at 4 (AR 1311). The Census Bureau concluded that, under Alternative C, approximately 35 million people would not be linked to administrative citizenship data. PX-24 at 3 (AR 1036). Under Alternative D, 36 million people cannot be linked to administrative records. Id. at 4. There is no evidence in the Administrative Record contradicting the Census Bureau's conclusion.

*70444. The Secretary's staff also worked to downplay deviation from Census Bureau standard procedures and gin up support for the Secretary's proposal. First, the Ross Memo misleadingly states that the citizenship question has been "well tested." PX-26 at 2 (AR 1314). This statement suggests compliance with the Census Bureau's standard pretesting procedures, which apply unless the Census Bureau obtains a waiver or uses a question that has performed adequately in another survey. However, there is no evidence in the Administrative Record that the citizenship question underwent any testing required by the Census Bureau's Statistical Quality Standards. Although it is true that the citizenship question underwent testing before it was added to the ACS, id. at 7 (AR 1319), the Census Bureau reported to Secretary Ross that approximately a third of respondents identified as noncitizens by administrative records reported themselves as citizens on the 2016 ACS. PX-22 at 8 (AR 1284); PX-26 at 6. In light of this finding, the Census Bureau concluded that data on citizenship obtained from the citizenship question were of "suspect quality" and "may not be reliable," particularly for noncitizens. PX-25 at 4-5 (AR 1311-12). Similarly, the Ross Memo does not account for the significant differences between the Decennial Census questionnaire and the ACS. Id. at 1-8 (AR 1313-1320). By labeling the citizenship question "well tested" while failing to acknowledge the evidence that the question does not perform adequately on the ACS and ignoring the differences between the ACS and the Census questionnaire, the Ross Memo downplayed deviations from Census Bureau procedure.

45. Similarly, although the Ross Memo states that the "value" of generating the requested data for DOJ is of "greater importance" than any adverse effects on the Census, PX-26 at 7 (AR 1319), Secretary Ross did not require DOJ to "demonstrate[ ] a clear statutory or regulatory need for data at small geographies or for small populations" per the Census Bureau's well-established processes for adding new questions to a Decennial Census, PX-140 at 7 (AR 10901). There is nothing in the Administrative Record reflecting an assessment by Secretary Ross or any other Defendant of the "value" or "importance" of the DOJ's request. In fact, the Administrative Record undermines the assertion that the requested data were of great importance to DOJ. It shows that the Commerce Department originally approached DOJ to request the question, PX-2 (AR 1321); PX-51 (AR 2462); PX-84 (AR 3701); PX-537 (AR 12756); DOJ initially refused to request the question, PX-537 (AR 12756); DOJ ultimately requested it only after the Attorney General personally spoke with Secretary Ross, PX-57 (AR 2528); PX-62 (AR 2637); PX-97 (AR 4004); and DOJ refused to meet with the Census Bureau to discuss the request, PX-101 (AR 5489); PX-109 (AR 6659).

46. In another effort to downplay deviations from standard operations, the Commerce Department revised the Census Bureau's description of its "well-established process" for "adding or changing content on the Census." The Secretary's staff had prepared a set of 35 questions regarding the Census Bureau's January 2018 analysis and recommendation. PX-545 (AR 1976). Question 31, asked "[w]hat was the process that was used in the past to get questions added to the decennial Census or do we have something similar where a precedent was established?" Id. at 21. The Census Bureau's written response to Question 31 described the following well-established process:

The Census Bureau follows a well-established process when adding or changing content on the census or ACS to ensure the data fulfill legal and regulatory requirements *705established by Congress. Adding a question or making a change to the Decennial Census or the ACS involves extensive testing, review, and evaluation. This process ensures the change is necessary and will produce quality, useful information for the nation.

The Census Bureau and the Office of Management and Budget (OMB) have laid out a formal process for making content changes.

First, federal agencies evaluate their data needs and propose additions or changes to current questions through OMB.

In order to be included, proposals must demonstrate a clear statutory or regulatory need for data at small geographies or for small populations.

Final proposed questions result from extensive cognitive and field testing to ensure they result in proper data, with an integrity that meets the Census Bureau's high standards.

This process includes several opportunities for public comment.

The final decision is made in consultation with OMB.

Id. at 21-22. The procedure described above is consistent with other Census Bureau documents addressing the same subject. See, e.g. , PX-699 (AR 3890), PX-3 (AR 3560), PX-4 (AR 9867); PX-140 at 7. Defendants' initial, incomplete production of the Administrative Record included only a different, far shorter response to question 31:

Because no new questions have been added to the Decennial Census (for nearly 20 years), the Census Bureau did not feed [sic] bound by past precedent when considering the Department of Justices' [sic] request. Rather, the Census Bureau is working with all relevant stakeholders to ensure that legal and regulatory requirements are filled and that questions will produce quality, useful information for the nation. As you are aware, that process is ongoing at your direction.

PX-1 (AR 1296). Although there is evidence that one member of the Census Bureau team said she was "fine" with an interim iteration of the response that was drafted by the Secretary's staff, see PX-14 (AR 13023), there is no evidence in the Administrative Record indicating that anyone at the Census Bureau approved or even saw this altered version of the question 31 response.

47. The Secretary and his staff's dealings with stakeholders offer another example of Defendants' efforts to stack the deck in the citizenship question's favor. The Ross Memo refers to discussions with an executive at the Nielsen survey agency, during which the executive indicated that Nielsen had added certain sensitive questions to short survey forms "without any appreciable decrease in response rates." PX-26 at 3 (AR 1315); see also id. at 6 (describing "empirical evidence" from Nielsen). The Administrative Record does not contain any of the "empirical evidence" obtained from Nielsen that is referenced in the Ross Memo. See PX-1 to PX-14 (AR 1-13024). The only evidence in the Administrative Record of Secretary Ross's interactions with Nielsen-notes from a call between Secretary Ross and Christine Pierce, the Nielsen executive-contradict the Ross Memo's characterization of the "evidence" provided by Nielsen. These notes actually indicate that "Ms. Pierce stated that her biggest concerns [sic] was that the reinstatement of a citizenship question could lead to a lower response rate ...." and that Pierce "noted the importance of testing questions." PX-1 (AR 1276).

*70648. In early 2018, Secretary Ross conferred with multiple outside stakeholders and received numerous written submissions regarding the citizenship question. See e.g. , PX-1112-1129 (AR 787-AR 3608). However, the evidence in the Administrative Record shows that nearly all of the stakeholders who provided an opinion on the citizenship question to Secretary Ross opposed the addition of citizenship as a subject for the 2020 Census and overwhelmingly reinforced the Census Bureau's recommendation against the addition. See, e.g. , PX-1 (AR 780, AR 778, AR 787, AR 840, AR 1053, AR 1150, AR 1222); PX-116 (AR 8555); PX-1123 (AR 1122); PX-1128 (AR 3605).

49. The Administrative Record reflects that acting Director Jarmin attempted to identify stakeholders who would speak in favor of adding a citizenship question and had trouble finding supporters. PX-70 (AR 3275) at 2. On February 13, 2018, Dr. Jarmin wrote to Michael Strain at the American Enterprise Institute (AEI) asking if someone would "speak to the pros" of adding a citizenship question because "[m]ost stakeholders will speak against the proposal." Id. Strain declined, responding that "[n]one of my colleagues at AEI would speak favorably about the proposal." Id. Dr. Jarmin ultimately reported that the Census Bureau was only able to find two organizations, the Center for Immigration Studies and the Heritage Foundation, that supported the addition of the citizenship question. PX-1 (AR 1206, AR 1261); PX-4 (AR 4849), PX-113 (AR 8325). However, neither of these organizations submitted any empirical data or technical research addressing the likely impact of the addition on Census data quality or the feasibility of meeting the DOJ's request using administrative records rather than by altering the Census questionnaire. See id.

50. Finally, the Trump/Pence reelection campaign-an unusual stakeholder given the Census Bureau's status as a non-political statistical agency-communicated its position on the Secretary's proposal indirectly. A campaign email broadcasted that "[t]he President wants the 2020 United States Census to ask people whether or not they are citizens." PX-64 (AR 2643-44); PX-3 (AR 3424-25). The Administrative Record shows that before the Ross Memo issued, Secretary Ross's staff alerted the Secretary of media reports about the reelection campaign's support for adding a citizenship question to the 2020 Census. PX-78 (AR 3597).

51. Despite the Census Bureau unanimously and repeatedly recommending against adding a citizenship question because the objectively superior Alternative C would achieve DOJ's stated goal while avoiding the damage to the enumeration, the Secretary nonetheless announced the "reinstatement of a citizenship question on the 2020 decennial census" in the Ross Memo on March 26, 2018. PX-26 at 8 (AR 1320).

B. The Secretary's Decision (Extra-Record Evidence)

52. Evidence in the Trial Record provides additional context to the circumstances surrounding Secretary Ross's decision to add a citizenship question to the 2020 Decennial Census.

53. First, the evidence confirms the Court's finding that the Secretary decided in the Spring of 2017, months before receiving DOJ's request, that he wanted to add a citizenship question to the 2020 Census. ECF No. 103-8 (Comstock Dep.) at 146:1-15. Secretary Ross concealed this fact when he represented to Congress that the Department of Commerce analysis around the citizenship question was "solely" in response "to the Department of Justice's request," and not at the direction of *707President Trump or anyone at the White House, PX-491, and that DOJ "initiated the request for the inclusion of the citizenship question" to the 2020 Census, PX-480 at 51.

54. Extra-record evidence also provides more support for the finding that the Secretary's staff searched for a rationale and an agency to request the question. The Secretary's staff did not know why Secretary Ross wanted to add a citizenship question. ECF No. 103-5 (Teramoto Dep.) at 32:10-12; ECF No. 103-6 (Dunn Kelley Dep.) at 39:3-16; see Comstock Dep. 266:4-12. But they nonetheless set out to come up with a "legal rationale" to support the proposal. Comstock Dep. 266:4-12. Comstock, in particular, believed he needed to find another agency to request the addition of a citizenship question because OMB and the Paperwork Reduction Act required the Commerce Department to "justify" why a citizenship question was "need[ed]," and he understood that simply saying "the Secretary wanted it" would not "clear [the] legal thresholds." Id. 153:6-13, 154:6-11. Comstock believed that it was his job to "figure out how to carry out what my boss asks me to do" and "find a legal rationale" for the question. Id. at 266:4-12. According to Comstock, it did not "matter what [Secretary Ross's] particular personal perspective" was. Id.

55. When Comstock first contacted DOJ on May 4, 2017 to solicit DOJ's request for the addition of a citizenship question, he was not seeking to promote more effective enforcement of the VRA. See Comstock Dep. 167:5-172:15. Neither James McHenry nor DOJ's Executive Office of Immigration Review, which McHenry heads up, have responsibility for enforcing the VRA, yet McHenry was Comstock's first DOJ contact. Comstock Dep. at 270-71; ECF No. 103-10 (Gore Dep.) at 65:3-14; PX-485; PX-537. Similarly, Comstock contacted DHS, which has no responsibility for enforcement of the VRA. See Comstock Dep. at 276:19-277:3. As Comstock was reaching out to DOJ and DHS, Langdon, on behalf of Comstock, requested information from Census Bureau staff regarding whether and/or how the Census counts noncitizens and illegal immigrants for purposes of apportionment. ECF No. 103-12 (Langdon Dep.) 174:15-184:11.

56. Additionally, extra-record evidence confirms that around the beginning of September 2017, Secretary Ross initiated a discussion with Attorney General Sessions regarding the addition of a citizenship question to the Census, even though DOJ had never communicated that CVAP data obtained from the ACS were not meeting DOJ's needs for VRA enforcement purposes. Gore Dep. at 83:16-84:1; PX-302; PX-1194 at 996:19-23. The Trial Record shows that Attorney General Sessions made the decision to help Secretary Ross by authorizing DOJ to request that the Census Bureau ask a citizenship question on the Census. Gore Dep. at 442.

57. After Attorney General Sessions' call with Secretary Ross, the Attorney General spoke with Gore about requesting the addition of a citizenship question. Gore Dep. at 74:18-75:14, 83. Gore confirmed that any discussions between the Department of Commerce and DOJ prior to September 2017 regarding the citizenship question were initiated by the Department of Commerce. Id. at 67-68.

58. On September 11, 2017, Gary sent an email to Gore for more information about concerns "the Secretary of Commerce raised last week with the AG relating to the 2020 Census" and the citizenship question. PX-648 at 2. In this email, Gary noted that he previously worked on adding "potential questions to the ACS relating to LGBTQ status, and on behalf of [DOJ] ... signed correspondence with Census on *708that topic." Id. Gary explained that, as a part of his involvement in the 2020 Census, he had not heard about "citizenship issues." Id. After speaking with Gore by phone, Gary writes, "My contact at Census OGC (not at the Department level) has heard nothing, and is equally puzzled about the question." Id. In September and October 2017, Gore spoke with the Department of Commerce's Peter Davidson, James Uthmeier, and Wendy Teramoto regarding the citizenship question. Gore Dep. at 91:18-94:3, 118:15-17.

59. In one of Gore's several calls with the Secretary's advisors, id. at 102-03, 117-19, 120-24, Gore spoke with Uthmeier, who had no experience litigating Section 2 VRA redistricting cases involving the use of CVAP data or otherwise assessing the reliability of CVAP data used in VRA litigation. Id. at 117:14-18:5; 118:15-17. Following a call with Uthmeier, Gore received Uthmeier's August 11 memorandum regarding the addition of a citizenship question along with a handwritten note from Uthmeier. Id. at 18:18-19:4; 123:16-24:2.

60. The Trial Record reveals that Gore relied, at least in part, on Uthmeier's input in drafting the December 12 letter requesting the addition of the citizenship question. Gore Dep. at 123-24. Gore also communicated with Neuman, who was advising the Department of Commerce on the citizenship question. Id. at 437-38; Comstock Dep. 155-56. In mid-late November 2017, Gary conferred with Attorney General Sessions' advisors regarding the draft letter. Id. at 141:6-14; PX-710; PX-711. The Attorney General's office requested additional time to review the letter and made changes. PX-710; PX-711. On December 8, 2017, Gore emailed Gary a draft of the ghostwritten letter "with leadership's final changes," and stated "With these changes, we are authorized to send. Sending on Monday is fine." PX-711; Gore Dep. at 145:6-147:9. Final authorization to send the letter came from Attorney General Sessions' staff on his behalf on Tuesday, December 12, 2017. Gore Dep. at 158:7-160:4.

61. Beyond supplementing Administrative Record evidence that the Department of Commerce manufactured the VRA rationale, the Trial Record also supports the idea that DOJ did not need the data it requested. For example, Plaintiffs' expert David Ely testified, "there is no reasonable basis for the claim that citizenship data from the decennial Census would be 'more appropriate for use in redistricting and Section 2 litigation' than currently available data." ECF No. 99-3 ¶ 18 (Ely). Gore also testified that he did not believe it is necessary for DOJ's VRA enforcement efforts to collect CVAP data through the Decennial Census questionnaire. Gore Dep. at 300. Further, Gore did not know whether the data the Census Bureau would obtain through the addition of a citizenship question would have larger or smaller margins of error, or be more precise, than the existing data available to DOJ. Gore Dep. at 215:15-216:19, 226:1-227:19, 232:17-233:2.

62. As discussed above, Dr. Jarmin and the Census Bureau repeatedly sought to meet with DOJ to discuss the details of DOJ's request so the Census Bureau could best meet DOJ's stated needs. Extra-record evidence provides more context to DOJ's decision to avoid meeting with Census Bureau officials. When he received Dr. Jarmin's request for a meeting, Gary told Gore about the request and that the Census Bureau had proposed an alternative way to provide DOJ with block-level CVAP data other than through adding a citizenship question. Gore Dep. 262-63, 265-66. Gore told Gary that he "would think about the issue and discuss it further with others," but Gore did not ask Gary for more *709details on the Census Bureau's alternative proposal. Id. at 264, 268. Gore discussed what Gary had relayed with several DOJ officials, including Attorney General Sessions. Id. at 265, 268-69. The record confirms that Attorney General Sessions personally decided that DOJ would not pursue the Census Bureau's alternative proposal and directed DOJ not to meet with the Census Bureau to discuss the DOJ's request for a citizenship question. Id. at 271:10-272:19. On January 19, 2018, Gore informed Gary that DOJ technical staff would not meet with the Census Bureau to discuss the alternative proposal for producing higher quality CVAP data at a lower cost. Id. at 273:17-274:4.

63. According to Dr. Abowd, the Attorney General's decision to direct DOJ not to meet with the Census Bureau constituted "political influence," which is "very problematic" with respect to the OMB Statistical Directive that the Census Bureau maintain political independence. ECF No. 139, Trial Tr. (Jan. 31) at 28:18-23-29:3 (Abowd). Documents produced by DOJ also show that DOJ sought to keep the fact that Secretary Ross solicited the DOJ's request confidential. PX-718 (a draft briefing paper included the bullet "NOT PUBLIC: In 2017, Secretary of Commerce Wilbur Ross requested that the Justice Department send a letter requesting the addition of a citizenship question on the 2020 Census.").

64. The Trial Record also gives context to deviations from Census Bureau practice and the Secretary's efforts to limit the Census Bureau's involvement in the decision-making process. According to Dr. Jarmin, after receiving the December 12 DOJ letter, the Census Bureau SWAT team investigated using administrative records to provide DOJ with CVAP data because 13 U.S.C. § 6(a) directs the Census Bureau to use administrative records in lieu of direct collection from the American people whenever possible. ECF No. 103-4, Jarmin Dep. 59:9-60:7. Secretary Ross ultimately rejected this alternative, however. The record also confirms that the February 12, 2018 meeting with Secretary Ross was the only meeting the Census Bureau had with the Secretary to discuss the citizenship question prior to the Secretary's March 26, 2018 memorandum. PX-1194 at 883:84.¹⁰ At the time of the February 12, 2018 meeting, Dr. Abowd did not know that Secretary Ross had requested to add a citizenship question in early 2017, and Secretary Ross did not disclose this at the meeting. Id. at 1017:12-21; 1045:6-12. Dr. Abowd did not learn about this request until this litigation, and he was surprised by this fact. Id. at 1019:18-1020:5.

65. The Trial Record also fills in gaps regarding Comstock's 35 questions for the Census Bureau, PX-545 (AR 1976), completing the picture sketched by the Administrative Record that the Commerce Department deliberately worked to downplay the degree to which the decision to add a citizenship question deviated from Census Bureau standards. Specifically, Dr. Abowd testified that the Census Bureau's initial answer to Question 31, PX-132 at 21 (AR 9832), accurately summarizes the Census Bureau's formal process for adding questions to the Census. PX-1194 at 1007:19-1008:8;

*710see also ECF No. 103-4, Jarmin Dep. at 137-138. Commerce Department political appointees drafted the revised, misleading response to Question 31. PX-1194 at 1010-11; ECF No. 99-5 (Lowenthal Decl.) ¶ 90. The version that made its way into the initially disclosed Administrative Record states, "the Census Bureau did not [feel] bound by past precedent when considering the Department of Justices' [sic] request" because "no new questions have been added to the Decennial Census (for nearly 20 years)." PX-1 (AR 1296). In reality, although the Census Bureau has not recently added questions to the Decennial Census, it has considered adding questions. Lowenthal Decl. ¶ 66-81, 90; see also ECF No. 125, Trial Tr. (Jan. 22) at 58:7-13 (Thompson) (discussing proposal to add ancestry question, which was rejected after multi-year testing program showed the question would reduce data quality); PX-262. The Bureau chose not to add questions precisely because issues arose during the well-established process that is in place for assessing whether to add a question. Id.

66. While the Administrative Record includes details about the Census Bureau's typical pretesting practices, extra-record evidence further explains why pretesting is necessary and why the Census Bureau's pretesting requirements would apply to the addition of a citizenship question. Pretesting is necessary because "[s]eemingly minor changes in question wording or sequence sometimes can affect survey responses in important and unexpected ways." PX-364 at 3. The Ross Memo states that, to minimize any effects on response rates, the citizenship question is to be placed "last on the decennial census form." PX-26 at 8 (AR 1320). However, there is no evidence in the Administrative Record or Trial Record reflecting an assessment or testing of how the placement of the question will affect response rates. The Trial Record does reflect, though, that question ordering can affect responses, and sometimes these effects can occur in unexpected ways that cannot be identified without pretesting. ECF No. 138, Trial Tr. (Jan. 30) at 158:12-24 (Abowd). For example, when a question on race precedes a question on Hispanic origin, responses to the Hispanic origin question go down because "people make the assumption that I've already told you about my race." ECF No. 126, Trial Tr. (Jan. 23) at 6-24 (Mathiowetz). When Hispanic origin precedes the race question, however, item non-response to the Hispanic origin question drops significantly. Id.

67. Although the Ross Memo asserts that the citizenship question is "well tested," the record shows that the only testing relating to a citizenship question was done in connection with its placement on the ACS. ECF No. 103-4, Jarmin Dep. at 55:1-10, 56:2-5. The citizenship question on the ACS was last tested in 2006, meaning it has been over 14 years since the question has been tested. ECF No. 138, Trial Tr. (Jan. 30) at 156:18-24 (Abowd). This gap is significant because the macroenvironment in 2006 relating to issues of citizenship was different than it will be in 2020, and no testing has been done on the citizenship question in today's macroenvironment. Id. at 157:5-13. Further, while the Ross Memo does not acknowledge it, the ACS and the Decennial Census are significantly different instruments. ECF No. 125, Trial Tr. (Jan. 22) at 51:12-18 (Thompson); ECF No. 126, Trial Tr. (Jan. 23) at 181:3-7 (Mathiowetz); PX-945; PX-946. On the ACS, the citizenship question is one of more than 50 questions, whereas on the Decennial Census, the citizenship question is one of 11 questions. Id. The parties agree that, as a result of the significantly shorter length of the Decennial Census as compared to the ACS, the citizenship question *711will be much more prominent on the 2020 Census than it is on the ACS. PX-162 at 46; ECF No. 138, Trial Tr. (Jan. 30) at 157:18-22 (Abowd); ECF No. 125, Trial Tr. (Jan. 22) at 51:12-18 (Thompson); ECF No. 126, Trial Tr. (Jan. 23) at 172:1-11 (Mathiowetz). Moreover, the citizenship question on the ACS is prefaced by a nativity question, but the citizenship question on the Census will not be. PX-945; ECF No. 138, Trial Tr. (Jan. 30) at 159:9-21 (Abowd).

68. The Bureau's Statistical Quality Standards require pretesting unless the Census Bureau obtains a waiver or uses a question that has "performed adequately in another survey." PX-260 at 18; ECF No. 138, Trial Tr. (Jan. 30) at 162:22-163:2 (Abowd). The Census Bureau did not seek a waiver to avoid pretesting of the citizenship question on the 2020 Census. ECF No. 138, Trial Tr. (Jan. 30) at 162:10-16 (Abowd). And, as Dr. Abowd testified, the citizenship question has not "performed adequately" on the ACS. PX-1194 at 1282, 1287-88. This conclusion is consistent with evidence in the Administrative Record, PX-22 at 8 (AR 1284), with the opinions of Plaintiffs' experts, ECF No. 125, Trial Tr. (Jan. 22) at 91:2-10 (Thompson); ECF No. 126, Trial Tr. (Jan. 23) at 90:7-13 (Mathiowetz), and with the August 2018 report from the Census Bureau that between 30% and 37% of all people identified as noncitizens by administrative records reported themselves as citizens on the ACS, PX-162. Thus, neither of the exceptions to the pretesting requirement apply, and pretesting was required. See ECF No. 138, Trial Tr. (Jan. 30) at 162:22-163:2 (Abowd).

69. The Trial Record also casts further doubt on the truthfulness of Secretary Ross's statements about his conversation with Nielsen executive Christine Pierce, which the Ross Memo contended supported the view that adding a sensitive question to a survey would not reduce response rates. PX-26 at 4 (AR 1313). Pierce testified, inter alia , that the Ross Memo materially mischaracterizes her unrecorded telephone conversation with the Secretary, including by falsely attributing to her statements that she did not make. In particular, Pierce affirmed that she did not cite evidence that a citizenship question would not cause a decline in self-response rates, but rather expressed her "unequivocal[ ]" contrary concern "that a citizenship question would negatively impact self-response rates." ECF No. 95-1, Pierce Decl. at ¶¶ 9, 12-18.

70. Finally, evidence in the Trial Record demonstrates that persons around Secretary Ross had an interest in whether undocumented immigrants are counted in the Census for apportionment purposes and that the Secretary did look at that issue. In addition to the communications Ross had with Kobach, PX-19 (AR 763), Comstock emailed the Secretary an article entitled "The Pitfalls of Counting Illegal Immigrants" in response to the Secretary's inquiry into whether undocumented people were counted for apportionment purposes on March 10, 2017, shortly after the Secretary's confirmation. PX-55 (AR 2521); Comstock Dep. at 62:13-64:4, 65:5-8. "Potentially" that same day, Secretary Ross made what later in May 2017 he would term his "months old request" that a citizenship question be added to the 2020 Census. Comstock Dep. 146: 1-15; see also PX-88 (AR 3710). As previously noted, during that same time frame, the Secretary had conversations with other officials about the citizenship question, but the details of those conversations are unknown. PX-2 (AR 1321).

71. Additionally, in arguing that the Secretary acted with discriminatory intent, Plaintiffs place much significance on the fact that two days after Secretary Ross's *712March 26, 2018 memorandum, President Trump's re-election campaign sent an email to supporters, stating that President Trump "officially mandated" the addition of a citizenship question to the 2020 Census. PX-487; PX-1173. As discussed in closing argument, however, the Court is hesitant to attribute weight to a statement that may have reflected no more than a low-level campaign staffer seeking to assign direct credit to the President for an agency action assumed to be popular with the President's base.

72. The Trial Record also includes several statements by candidate, President-elect, and President Trump demonstrating his animus toward immigrants. See PX-1139; PX-1145; PX-1149; PX-1156; PX-1177. Similarly, the record includes statements in tweets and other mediums that show President Trump is concerned by the political power that undocumented immigrants may wield. See PX-1156; PX-1177. Although these public statements are relevant because Secretary Ross serves at the pleasure of the President, the Court ultimately is not persuaded by them given that Plaintiffs failed to firmly tether the statements to Secretary Ross's decision. Indeed, some of the President's statements were made after the Ross Memo issued. PX-1177 (statement made on June 19, 2018, two months after Ross Memo). Even if it is unlikely that President Trump did not hold this same view a few months earlier when the Secretary was making his decision, there is nothing in the record that shows the Secretary considered or adopted these statements before deciding to add a citizenship question to the 2020 Census.

73. Plaintiffs' have thus presented evidence that the President and Kobach harbored discriminatory animus towards non-citizens and evidence that the Secretary considered the impact of counting illegal immigrants in the Census, among other undisclosed issues.¹¹ They have also presented substantial evidence that the VRA rationale was not Secretary Ross's original or actual motivation. Ultimately though, the Court cannot, by a preponderance of the evidence, connect the dots between the President and Kobach's views, the Secretary's failure to disclose his real rationale, and the Secretary's final decision. Without more evidence demonstrating the Secretary was actually persuaded to make his decision based on discriminatory animus, a finding that, more likely than not, the Secretary's real motivation was to depress immigrant response rates cannot be made. Ultimately, Secretary Ross's original rationale remains, to some extent, a mystery.

C. Standing

74. Turning to the Court's findings of fact relevant to whether Plaintiffs have standing to assert their claims, the Court may look to evidence beyond the Administrative Record. See Lujan v. Defs. of Wildlife , 504 U.S. 555, 561, 112 S.Ct. 2130, 119 L.Ed.2d 351 (1992). Thus, in this section, the Court does not distinguish between the findings of fact derived from the Administrative Record and those derived from the Trial Record. Plaintiffs are individual residents of Texas, Arizona, Nevada, Florida, Washington, New Jersey and South Dakota, and civic engagement, community, and political organizations dedicated to serving Hispanic and immigrant communities in Texas, California, Georgia, Maryland, Arizona, New York, Washington, Massachusetts and across the country. If a citizenship question causes a greater decline in self-response rates among noncitizen and *713Hispanic households as compared to other households (i.e., a differential decline in participation) and that differential decline is not resolved through NRFU, a differential undercount of those groups will result, thus potentially causing injury to Plaintiffs for standing purposes.

1. Differential Decline in Census Participation

75. Overwhelming evidence supports the Court's finding that a citizenship question will cause a differential decline in Census participation among noncitizen and Hispanic households.

76. The starting point for this analysis is the issue of self response. The record supports the intuitive finding that question sensitivity impacts response rates, and that a citizenship question is particularly sensitive among noncitizen and Hispanic households. See e.g. , PX-22 at 4-5 (Census Bureau's January 19 Memo); PX-162 at 54; PX-696 at 591-92, 596 (Barreto). Questions are more sensitive if respondents do not trust that the information being collected is needed or that the information will be kept confidential. ECF No. 126, Trial Tr. (Jan. 23) at 98:7-23 (Mathiowetz). "[M]any in the Latino and immigrant community have trust issues with" questions "related to citizenship." PX-696 at 601 (Barreto). These communities also have concerns about the confidentiality of citizenship data even though responses are confidential by law. Id. at 602-04 (Barreto). The macroenvironment (i.e., the political and social climate) in which a survey is administered can increase concerns about trust and confidentiality. Id. at 610. The current political environment around immigration and immigration enforcement amplifies the sensitivity of a citizenship question. PX-1194 at 927:5-10; ECF No. 138, Trial Tr. (Jan. 30) at 125:17-24, 131:10-15 (Abowd); see also PX-22 at 6 (AR 1282); ECF No. 99-6, Massey Decl. ¶ 20-22; ECF No. 126, Trial Tr. (Jan. 23) at 62:10-63:2 (O'Hare).

77. Hispanic and noncitizen sensitivity to the topic of citizenship manifests itself in several ways. For example, from 2013 to 2016, Hispanic respondents were twice as likely not to answer the ACS citizenship question as non-Hispanic whites. PX-22 at 4; PX-162 at 8-10 (showing the item nonresponse rate for Hispanics increasing from 2013 to 2016, while the rate for non-Hispanic whites decreased); PX-1194 at 906:12-907:20; ECF No. 138, Trial Tr. (Jan. 30) at 130:3-12; Census Bureau 30(b)(6) Dep. Vol. II at 359:13-361:5. Also indicative of the citizenship question's sensitivity is the question's "breakoff rate"-Hispanics were over 8 times more likely to stop responding to the 2016 ACS online after they reached the citizenship question than non-Hispanic whites. PX-22 at 5; PX-69 at 1; PX-162 at 10-11; PX-1194 at 914:5-8; Census Bureau 30(b)(6) Dep. Vol. II at 361:6-363:4; ECF No. 138, Trial Tr. (Jan. 30) at 131:5-9 (Abowd). Further, qualitative evidence from focus groups show that although Hispanic and noncitizen respondents want to be counted in the Census, "fear of deportation outweighs any benefit." PX-152 at 22; PX-1194 at 930:9-24, 933:20-934:7, 938:22-939:17, 940:04-941:14; Census Bureau 30(b)(6) Dep. Vol. II at 449:5-13, 449:21-451:9. Similarly, a survey designed and conducted by the Census Bureau to provide the Bureau with reliable, actionable information about the macroenvironment's effect on survey response, ECF No. 138, Trial Tr. (Jan. 30) at 132:20-25, revealed that 32% of Hispanics and 34% of foreign born respondents believed their answers to a citizenship question would be used against them-a significantly higher percentage than the 22% of overall respondents who felt this way, ECF No. 126, Trial Tr. (Jan. 23) at 110:4-111:1 (Mathiowetz). See also id. at 106:10-107:11; PX-152; PX-163.

*71478. Community leaders representing the Organizational Plaintiffs also testified credibly that households with noncitizens fear answering the citizenship question. For example, Plaintiff Juanita Valdez-Cox, Executive Director and member of Plaintiff LUPE provided credible and persuasive testimony that because the social and political climate has been difficult for immigrant communities in Hidalgo County in Texas over the past two years, in particular for mixed-immigration status households, residents are "going to be so much more fearful of answering and [will] wonder about the consequences." ECF No. 125, Trial Tr. (Jan. 22) at 173:22-175:4. LUPE members in mixed status families are wondering "how will I hurt my mom or my dad or my grandmother who are undocumented if I answer?" and they are uncertain and "very fearful" about how responses to a citizenship question will be used. Id. at 179:19-180:22. Based on this fear, Ms. Valdez-Cox reported that in Hidalgo County, where she lives, families like her own that contain both native-born Hispanic citizens and naturalized citizens will not participate in the 2020 Census. Id. at 179:19-180:22, 182:6-21. Similarly, John Park, Executive Director of Plaintiff MinKwon, provided credible and persuasive testimony that "there is a real climate of fear, trauma" and "anxiety" in the Asian-American and immigrant communities that has been increased by the addition of the citizenship question and that makes MinKown's Census outreach efforts increasingly difficult. Id. at 214:5-218:3.

79. In this context, the Court finds that because of trust and confidentiality issues-heightened by the macroenvironment-questions about citizenship are particularly sensitive for Hispanics and noncitizens, meaning self-response rates among these groups will decline more than any decline in overall participation.

a. 5.8 Percentage Point Differential Decline in Noncitizen Participation

80. It is conservatively estimated that the differential decline in self response for households that contain a noncitizen is 5.8 percentage points. PX-162 at 38-39 (Tables 8 & 9); PX-1194 at 897:16-20; Census Bureau 30(b)(6) Dep. Vol. II at 372:2-12; see also ECF No. 126, Trial Tr. (Jan. 23) at 113:23-119:6, 125:12-21 (Mathiowetz). This estimate comes from the Census Bureau's own study (the Brown et al. Memo) published August 6, 2018, which updated and expanded upon the Census Bureau's earlier finding that at least a 5.1 percentage point differential decline in noncitizen self response would result from the addition of a citizenship question. PX-162.

81. In arriving at the 5.8 percentage point estimate, the Brown et al. Memo conducted a natural experiment comparing the decline in self-response rates between the 2016 ACS and the 2010 Decennial Census for households with no noncitizens and all other households (i.e., households that did contain or potentially contained noncitizens). PX-1194 at 898:2-899:6; Census Bureau 30(b)(6) Dep. Vol. II at 373:9-15; PX-162 at 38-39. The Brown et al. Memo then utilized statistical techniques to control for the burden of the longer ACS questionnaire and for other groups of questions on the ACS. ECF No. 138, Trial Tr. (Jan. 30) 68:17-71:3, 71:23-72:9 (Abowd). The 5.8 percentage point differential decline in self response for noncitizen households is therefore not explained by the burden caused by the greater length of the ACS questionnaire or by certain other groups of potentially sensitive questions on the ACS. PX-162 at 38-39; ECF No. 138, Trial Tr. (Jan. 30) at 69:14-70:5, 71:23-72:9 (Abowd).

82. The 5.8 percentage point estimate is "conservative," PX-162 at 39, because it is based on an analysis of ACS data, and the *7152020 Census questionnaire is much shorter, which means a citizenship question would be significantly more prominent. PX-162 at 39; PX-1194 at 901:22-902:4; Trial Tr. (Jan. 23) 172:1-11 (Mathiowetz). The 5.8 percentage point estimate is also conservative because it is based on 2016 ACS data, which do not account for post-2016 salient developments in the macroenvironment that might impact response rates. PX-162 at 39; PX-1194 at 902:11-23, 944-46; ECF No. 138, Trial Tr. (Jan. 30) at 124:20-125:24 (Abowd); ECF No. 126, Trial Tr. (Jan. 23) at 119:16-120:5, 125:22-126:12 (Mathiowetz).

83. Judge Furman accurately described why the Defendants' criticisms of the Census Bureau's own research as uncertain and inadequate-criticisms repeated verbatim by Defendants here, ECF No. 150 ¶¶ 59, 60 -are unpersuasive. New York v. Dep't of Commerce , 351 F.Supp.3d 502, 582 ¶¶ 202-04 (S.D.N.Y. 2019). First, although it "goes without saying that the magnitude of any decline in self response to the 2020 census questionnaire is, today, 'unknown' (since the census has not yet occurred)," the weight of the evidence definitively shows "a probable range of 'magnitudes,' all illustrating the overwhelming likelihood that the 'magnitude' of the differential decline in self responses among noncitizens" will be quite high, likely higher than 5.8 percentage points. Id. ¶ 202. Further, Defendants' "misguided criticisms of the Census Bureau's own research" as inadequate for failing to conduct a randomized control study ignores that the Brown et al. Memo "did conceive of, and control for, numerous potentially confounding variables." Id. ¶ 203 (emphasis in original).

b. 8.7 Percentage Point Differential Decline in Hispanic Participation

84. As for the Hispanic differential decline, Plaintiffs have shown by a preponderance of the evidence that Hispanic self response will decrease by a magnitude of approximately 8.7 percentage points. Although the Census Bureau did not conduct an analysis quantifying the impact of a citizenship question on Hispanic self response, ECF No. 138, Trial Tr. (Jan. 30) at 148:5-149:2 (Abowd), Plaintiffs' expert Dr. Mathiowetz did, ECF No. 126, Trial Tr. (Jan. 23) at 120:12-122:5. PX-938. Dr. Mathiowetz quantified the impact of the citizenship question on self-response rates for Hispanics by comparing self-response rates of the 2010 ACS and 2010 Decennial Census forms for Hispanic and non-Hispanic households. Id. This analysis mirrors the Census Bureau's methodologically-sound natural experiment comparing citizen and noncitizen responses rates. Id. ; see also PX-162; PX-22.

85. To be sure, Dr. Mathiowetz's analysis could not control for all the same potential explanations that the Brown et al. Memo considered, like the differential effect a household size has on the number of questions a respondent has to answer on the ACS versus the Census or how other sensitive questions that appear on the ACS might account for the difference between the decline in self-response rates by Hispanic and non-Hispanic households. ECF No. 126, Trial Tr. (Jan. 23) at 191:21-25, 194:19-195:4, 195:16-196:7. After all, because the analyses contained in the Brown et al. Memo relied on non-public Census Bureau data, Dr. Mathiowetz's analysis could not be identical to those in the Brown et al. Memo. Nevertheless, Dr. Mathiowetz's study is analogous to the Census Bureau's own research, and, as Dr. Abowd acknowledged, it comes as close to the kind of natural experiment contained in the Brown et al. Memo as is possible based on publicly-available data. ECF No. 138, Jan. 30 Trial Tr. at 67:22-24 (Abowd).

*71686. Further, although Dr. Mathiowetz's analysis could not control for the length of the ACS or for other potentially sensitive ACS questions that do not appear on the Census, Plaintiffs provided substantial evidence to support Dr. Mathiowetz's opinion that the citizenship question was responsible for the difference in self-response rates between Hispanic and non-Hispanic households. See e.g. , PX-696 at 596, 602-04 (Barreto); PX-162 at 8-10; PX-1194 at 906:12-15, 906:25-907:20; ECF No. 138, Trial Tr. (Jan. 30) at 130:3-12; Census Bureau 30(b)(6) Dep. Vol. II at 359:13-361.¹² Additionally, while it makes sense that other potential factors like the length of the ACS could be responsible for driving down overall self-response rates to the ACS, Defendants point to no record evidence explaining why the length of the ACS would differentially drive down Hispanic participation while not impacting non-Hispanic self response. Similarly, while Defendants speculate that other potentially sensitive ACS questions could be responsible for driving down Hispanic ACS participation, there is no evidence in the record that such a causal relationship exists for any question other than citizenship. Further, the overlap between noncitizen households and Hispanic households-57.5% of the households in which at least one person is a noncitizen are also Hispanic households, Trial Tr. (Jan. 23) 54:4-56:14 (O'Hare)-suggests that the same trust and confidentiality concerns impacting noncitizen self-response rates can explain depressed Hispanic self-response rates.

87. Ultimately, the weight of the evidence shows the overwhelming likelihood that the "magnitude" of the differential decline in self response among Hispanics will be 5.8 percentage points or higher for noncitizen households and 8.7 percentages points or higher for Hispanic households.

2. Differential Undercount of at Least 2 Percentage Points

88. A net undercount occurs for a particular demographic when the Bureau's independent estimate, derived from a census coverage measurement or post-enumeration survey meant to determine the accuracy of the Census, is larger than the Decennial Census count for that group. As defined by the Census Bureau, the term "differential undercount rate" means the difference between the net undercount rate for a particular demographic or geographic domain and the net undercount rate either for another domain or for the nation. ECF No. 103-1, Joint Stips. ¶ 55.

89. The Court next finds that demographic groups with lower self-response rates are more likely to be undercounted. This straight line between lower participation prior to NRFU and an ultimate net undercount is supported by common sense and a preponderance of the evidence. ECF No. 126, Trial Tr. (Jan. 23) 40:7-51:24 (O'Hare); id. at 133:1-137:13, 139:2-145:23 (Mathiowetz); PX-890; PX-891; PX-1083-1085; PX-1087; PX-1088; PX-1089; PX-1090; PX-1091; PX-1092; PX-1214. Logic dictates, as Plaintiffs' counsel explained, "[i]f you self-respond to the Census, you are 100 percent guaranteed not to be undercounted," but "if you do not self-respond to the Census, you are less than 100 percent guaranteed not to be undercounted." ECF No. 153, Trial Tr. (Feb. 27) at 241:22-25. The record shows specifically that lower self response causes higher net *717undercounts because lower participation results in more enumerations through NRFU, which generates poorer quality data and undercounts. ECF No. 126, Trial Tr. (Jan. 23) at 50:25-51:7, 53:20-54:32 (O'Hare), 145:13-20 (Mathiowetz); see also PX-22 at 1, 5-6; PX-162 at 42, 43 n. 60 (noting that citizenship question will increase "enumeration errors" that may be unavoidable through NRFU fieldwork). Because, as explained above, the citizenship question will cause lower participation by noncitizens and Hispanics, these groups are also more likely to be undercounted if a citizenship question is included on the 2020 Census.

90. Plaintiffs' experts, Dr. O'Hare and Dr. Mathiowetz, opined that the relationship between a decline in self response and an increase in net undercount is causal. ECF No. 126, Trial Tr. (Jan. 23) at 50:25-51:7, 53:20-54:3 (O'Hare); id. at 143:13-148:8 (Mathiowetz). Defendants argue to the contrary that correlation does not imply causation. Social scientists typically look for four factors to show causation: (1) that the causal agent occurs prior in time to the thing that it is causing; (2) that there is an association or correlation between the causal agent and the thing being caused; (3) that a reasonable explanation for the relationship between the causal agent and the thing being caused can be specified; and (4) that other potential explanations have been controlled. ECF No. 126, Trial Tr. (Jan. 23) at 52:2-25 (O'Hare); see also id. at 145:21-146:12 (Mathiowetz). While correlation alone does not usually imply causation, correlation combined with other causation factors may, and not all four factors need to be satisfied to prove causation. ECF No. 126, Trial Tr. (Jan. 23) at 52:2-3 (O'Hare) ("Causation usually involves more than correlation"); id. at 52:4-25. Three factors of causation are satisfied here: self response occurs prior in time to net undercounting, the decline in self response is moderately to highly correlated with net undercounting, and the explanation for the causal relationship is clear: "lower self-response rates of a group mean that more of that group will have to be counted in the NRFU operation of the Census, and there is lots of good evidence saying the accuracy of the NRFU operation is not as good as the accuracy of the self-response rate." ECF No. 126, Trial Tr. (Jan. 23) at 52:16-25 (O'Hare); see also id. at 145:8-147:22 (Mathiowetz) (noting the factors that support a causal inference and describing the causal mechanism linking self response and undercount: "once you don't have self response, the quality of the data that you get ... is nowhere near as good as the quality of the data that you get from self-response, and so you can see the mechanism by which a lower self-response leads to poorer quality data and a higher undercount."). Thus, the Court is comfortable finding that Plaintiffs have demonstrated a causal relationship between decreased Census participation and an increased likelihood of net undercounting by a preponderance of the evidence.

91. Further, the differential decline in self-response rates caused by a citizenship question is especially likely to lead to differential undercounts of Hispanics and noncitizens because at every step in the NRFU and imputation process, these remedial efforts will be less effective at mitigating the decline in these groups' participation rates.

92. The Census always fails to count some people. Census Bureau 30(b)(6) Dep. Vol. I at 253:20-254:4; see PX-267 at 18, 20 (Tables 7 & 9). In particular, some demographic groups have proven more difficult to count in past decennial censuses. These groups are referred to as "hard to count." ECF No. 103-1, Joint Stips. ¶ 47. Racial *718and ethnic minorities, immigrant populations, and non-English speakers have historically been some of the hardest groups to count accurately in the Decennial Census. Id. ¶ 48. Hispanics and noncitizens are considered hard-to-count. ECF No. 138, Trial Tr. (Jan. 30) at 191:10-15 (Abowd); ECF No. 126, Trial Tr. (Jan. 23) at 146:18-147:7 (Mathiowetz); PX-267 at 18 (Table 7). Even with NRFU efforts, the Census Bureau was not completely successful in remedying omissions for Hispanics in the 1990, 2000, and 2010 Censuses. ECF No. 126, Trial Tr. (Jan. 23) at 146:18-147:7 (Mathiowetz); PX-267 at 18 (Table 7).

93. The current macro-environment will only make it more difficult than in prior Decennial Census years for NRFU operations to mitigate the decline in noncitizen and Hispanic self response. As Dr. Mathiowetz credibly and persuasively testified, in contrast to prior censuses where most people went into NRFU because they forgot to self-respond or did not realize there was a deadline, the inclusion of a citizenship question in the 2020 Census will "actually motivate nonresponse" and this "intentional concealment" makes it unlikely that households that do not self-respond will later respond in NRFU. ECF No. 126, Trial Tr. (Jan. 23) at 147:8-148:8 (Mathiowetz); see also id. 149:20-150:20 (Mathiowetz); PX-162 at 42-43 nn. 59-60; PX-1214 at 11. The record supports the idea that individuals choosing not to respond to the Census out of fear are not likely to be reassured by an enumerator knocking on their door. See id.¹³ There is no evidence in the record to the contrary, and Dr. Abowd is unaware of any empirical evidence that someone who chooses not to respond to the 2020 Census because of the citizenship question would respond to a face-to-face enumerator. Census Bureau 30(b)(6) Dep. Vol. I at 251:15-21; ECF No. 138, Trial Tr. (Jan. 30) at 179:14-19, 181:6-12 (Abowd).

94. While the Census Bureau plans to use high quality administrative records as part of NRFU efforts to enumerate individuals who do not self respond, ECF No. 138, Trial Tr. (Jan. 30) at 182:13-15 (Abowd), these records are unlikely to correct enumeration errors for Hispanic and noncitizen households, id. at 182:22-183:21. See also Census Bureau 30(b)(6) Dep. Vol. I at 252:21-253:6, Vol II at 389:12-392:4. Although administrative records are a superior source for obtaining the citizenship data of already-enumerated individuals (as opposed to asking a citizenship question), see PX-22 at 1-2, PX-25 at 5, administrative records are less effective at enumerating Hispanic and noncitizen households as compared to other populations. ECF No. 138, Trial Tr. (Jan. 30) at 182-83 (Abowd); Census Bureau 30(b)(6) Dep. Vol. I at 252:21-253:6, Vol. II at 389:12-392:4; id. at 182:22-183:1, 183:2-5 (Abowd); see also ECF No. 103-4, Jarmin Dep. at 285-86; Census Bureau 30(b)(6) Dep. Vol. II at 389:12-390:5; ECF No. 126, Trial Tr. (Jan. 23) at 148:14-25 (Mathiowetz); Census Bureau 30(b)(6) Dep. Vol. I at 252:16-253:6; ECF No. 138, Trial Tr. (Jan. 30) at 183:14-17 (Abowd) (Administrative records are more likely to exist for citizens than noncitizens). Because noncitizen and Hispanic households are less likely to have high quality administrative records as compared to other groups, the use of administrative *719records to enumerate households may actually exacerbate the differential undercount of these groups. PX-696 at 638:16-640:15.

95. The next step in the NRFU process-the use of proxies-is also unlikely to cure the decline in self response by noncitizen and Hispanic households. Census Bureau 30(b)(6) Dep. Vol. II at 382:9-383:5, 386:2-387:10; PX-696 at 640:16-643:21; ECF Nos. 126, Trial Tr. (Jan. 23) at 149:1-151:2 (Mathiowetz); 138 Trial Tr. (Jan. 30) at 183:22-186:3 (Abowd). Proxy responses are generally more likely to result in omissions of household members. Census Bureau 30(b)(6) Dep. Vol. II at 382:22-383:5; Trial Tr. (Jan. 30) at 185:3-24 (Abowd); PX-339 at 22-23 ("[T]he stark difference between nonmatch rates for household respondents ... versus proxy respondents ... suggest that unknowledgeable or unwilling proxy respondents may be a key factor in the undercount of young children."). There are currently no protocols for the Bureau to enumerate omitted household members when a proxy enumerates a household but omits one or more household members. ECF No. 138, Trial Tr. (Jan. 30) at 185:25-186:3 (Abowd). The record also demonstrates that proxy responses are likely to be less accurate because the citizenship question will also make proxies less willing to provide information that could be used against their neighbors. ECF No. 126, Trial Tr. (Jan. 23) at 149:1-8, 150:21-151:5 (Mathiowetz); PX-158 (AR); PX-162 at 43; PX-163; see also Census Bureau 30(b)(6) Vol. II at 386:16-387:10.

96. Imputation, the final step in the enumeration process, will also not offset the decline in self-response rates among noncitizens and Hispanics. As previously described, after three failed proxy attempts, the Census Bureau will impute the number of persons living in a household and their characteristics. ECF No. 138, Trial Tr. (Jan. 30) 111:5-11 (Abowd). Because imputation involves using data from households that have responded to the Census to model and assign a count to non-responding households, ECF No. 138, Trial Tr. (Jan. 30) at 187:11-15 (Abowd), imputation will reinforce or exacerbate a differential undercount of noncitizens and Hispanics, ECF No. 138, Trial Tr. (Jan. 30) at 190:21-191:6 (Abowd); PX-696 at 727:3-734:25 (Barreto); PX-682; PX-683. As Dr. Barreto testified, "for multiple different demographic indicators, responding units are not statistically the same as non-responding units." PX-696 at 711:4-711:8. Specifically, households that will respond to a Census with a citizenship question typically have smaller household sizes than those that will choose not to respond. PX-680. Accordingly, the Census Bureau's imputation model incorrectly assumes that responding-unit data are representative of non-responding units when in reality the data that are used for imputation represents groups with smaller household sizes than the non-responding demographics. PX-696 at 711:18-712:16.

As Dr. Barreto explained,

[T]he decision not to respond appears to be correlated with household size, that is, people who are the most anxious and nervous and not willing to respond have larger household sizes that cannot be accounted for by other demographic differences. This is consistent with the literature that suggested that people would be more fearful if they had other relatives who were noncitizens and others living in the house. So when the imputation model is applied at the very end of the process, there will be more Latino and immigrant households in need of imputation, first of all, because of the lower self-response and because of the lesser success of NRFU. So when *720we get to the imputation component, this model suggests that there will be a larger miss, disproportionately larger miss of Latino household sizes leading to a net undercount.

PX-696 at 733:7-733:24 (Barreto).

97. The Court gives Dr. Barreto's opinion on this point only limited weight because, as Dr. Abowd testified, Dr. Barreto's survey "doesn't appear to have controlled his weights so that they give an accurate estimate of the household size and the population as a whole," meaning the study's "estimate for the size of the households is too big." ECF No. 138, Trial Tr. (Jan. 30) at 76:24-77:2 (Abowd). However, other evidence also supports the finding that Hispanic and noncitizen households are likely to be larger than non-Hispanic white households or citizen households. PX-388 at 21, 23 (Figures 13 & 14); PX-389 at 12-13; PX-430 at 5-6.

98. Even assuming NRFU and imputation are more successful at accurately enumerating noncitizen and Hispanic households than the record suggests they will be, the problem of "rostering omissions" remains. Rostering omissions occur when households leave certain individual members off their questionnaire. ECF No. 126, Trial Tr. (Jan. 23) at 70:19-23 (O'Hare). Neither NRFU nor imputation can cure undercounts caused by rostering omissions. Census Bureau 30(b)(6) Dep. Vol. II at 396:2-399:2, 459:21-460:15; ECF No. 126, Trial Tr. (Jan. 23) at 71:9-24 (O'Hare); id. at 160:12-20 (Mathiowetz); ECF No. 138, Trial Tr. (Jan. 30) at 194:19-195:25 (Abowd); PX-410 at 13. And substantial evidence indicates that a citizenship question will cause a differential increase in rostering omissions among Hispanic and noncitizen households. ECF No. 126, Trial Tr. (Jan. 23) 156-60 (Mathiowetz); ECF No. 138, Trial Tr. (Jan. 30) 193-94 (Abowd); PX-1216 at 63; Census Bureau 30(b)(6) Dep. Vol. II at 394:7-20.¹⁴

99. Although quantifying the precise size of the differential undercount is difficult, Plaintiffs have proven that, at the very least, the citizenship question will result in a 2 percentage point increase in the differential undercount of noncitizens and a 2 percentage point increase in the differential undercount of Hispanics. ECF No. 126, Trial Tr. (Jan. 23) 161:17-170:14; PX-1214 at 13-14.

100. First, the Court credits Dr. Mathiowetz's conservative calculation for a differential undercount of noncitizen households, which she arrived at using the following steps. She began by estimating the undercount of noncitizen households based on the relationship between the decline in self response and increased net undercount observed in recent decennial censuses-i.e., that a 10 percentage point drop in self response was associated with a 2 percentage point increased undercount-to calculate that the 5.8 percentage point drop in self response calculated in the Brown et al Memo would translate to a 1.2 percentage point increase in undercount due to unit non-response. ECF No. 126, Trial Tr. (Jan. 23) 164:15-21 (Mathiowetz); PX-1214 at 13. Next, Dr. Mathiowetz estimated the undercount of noncitizen households resulting from rostering omissions. Based on the Census Bureau's estimate of a 60.5% self-response rate for the general population and a 5.8 percentage point decline in self response for noncitizen households, she estimated that the self-response rate for noncitizens will be 54.7%. ECF

*721No. 126, Trial Tr. (Jan. 23) 163:3-23. For the 54.7% of noncitizen households (19.7 million households) that are expected to respond to the 2020 Census with a citizenship question, she further focused on households with at least one adult citizen living with a noncitizen, a universe of 10.7 million households. Id. at 165:11-23. For these households, Dr. Mathiowetz applied a 5% rostering error rate, which she extrapolated from analogous literature and other Census Bureau data, id. at 152:18-160:2 and determined that roughly 2% of the noncitizen population would be undercounted through rostering omissions as a result of the citizenship question. Id. at 163:9-164:13, 165:11-166:19 (Mathiowetz). Putting these two figures together (1.2 + 2), yielded an increase in the differential undercount of 3.2 percentage points for noncitizen households as a result of the citizenship question. Id. at 166:25-167:1. However, to be conservative, Dr. Mathiowetz estimated that the increase in the differential undercount of noncitizens will be at least 2 percentage points. ECF No. 126, Trial Tr. (Jan. 23) at 166:25-167:11 (Mathiowetz).

101. The Court notes that the weakest part of Dr. Mathieowetz's analysis, which the Court gives the least amount of weight, is her determination that rostering omissions will occur at a rate of 5%. Dr. Mathiowetz based the 5% rostering omission estimate on four sources: (a) the Brown et al. Memo (PX-162); (b) a journal article by Roger Tourangeau et al. (PX-923); (c) a Census Bureau memo about respondent confidentiality concerns (PX-158); and (d) a journal article by David J. Fein (PX-394). See PX-1214 at 12. However, while these sources all support the finding that rostering omissions by Hispanic and noncitizen households will occur, they do not demonstrate the magnitude of the phenomenon. The Tourangeau et al. article examined the effect that de-anonymizing a survey (and thus making it more sensitive) had on responses to the survey, particularly among African American respondents. PX-923. The study found that when a survey was anonymous (and thus less sensitive), there was approximately a 5% increase in the number of usual residents enumerated for a household. PX-923 at 8-10. The study's authors concluded that concealment, rather than confusion about the survey, likely explained the omission of household members from the de-anonymized survey. PX-923 at 15; Trial Tr. (Jan. 23) at 156:19-159:21 (Mathiowetz). The Tourangeau et al. study demonstrated how increasing the sensitivity of a survey could increase rostering omissions and provides a parallel to what the addition of a sensitive citizenship question would do to responses on the 2020 Census. ECF No. 126, Trial Tr. (Jan. 23) at 159:1-21 (Mathiowetz). However, while a qualitative analogy can certainly be drawn from the study, the Court is not convinced that Dr. Mathiowetz can credibly extrapolate the quantitative finding that rostering omission errors will occur at a rate of 5%.

102. In any case, the Court ultimately credits Dr. Mathiowetz's conclusion that the differential undercount of noncitizens will increase by at least 2 percentage points because the other assumptions upon which she relies are extremely conservative. For example, the 2 percentage point differential undercount estimate assumes that NRFU and imputation remedies will be as effective as they have been in the past by using the relationship observed historically between lower self response and increased undercount, rather than a measure that reflects the increased sensitivity of the citizenship question in 2020. In reality, the evidence shows that NRFU operations will be subject to all the same negative effects on response rates that will *722depress initial participation. And, for the reasons already described, significant imputation errors will occur because the data used to model noncitizen households will be representative of smaller, less-complex all-citizen households. Dr. Mathiowetz also assumed, conservatively, that where rostering omissions occur, only one individual is omitted per household. Finally, to be "ultra conservative," Dr. Mathiowetz reported to the Court that the differential undercount for noncitizens would increase by at least 2 percentage points even though she initially calculated, using conservative parameters, an increase in the differential undercount of 3.2 percentage points.

103. The Court also credits Dr. Mathiowetz's conservative 2 percentage point estimated increase in the differential undercount of Hispanics. ECF No. 126, Trial Tr. (Jan. 23) at 167:20-170:14 (Mathiowetz); PX-1214 at 14. Based on an 8.7 percentage point decline in self response that she calculated from 2010 Census and ACS data, and the fact that a 10 percentage point decline in self response has historically been associated with a 2 percentage point increase in net undercount, Dr. Mathiowetz calculated that the differential undercount of Hispanics would increase by approximately 1.7 percentage points, attributable to a decline in self response. ECF No. 126, Trial Tr. (Jan. 23) at 169:7-14 (Mathiowetz); PX-1214 at 14. Next, Dr. Mathiowetz calculated the differential undercount of Hispanics attributable to rostering omissions. Drawing on the Census Bureau's expected national self-response rate (60.5%) and her calculation that the decline in self response among Hispanics will be 8.7 percentage points, Dr. Mathiowetz calculated that the Hispanic self-response rate will be 51.8%. ECF No. 126, Trial Tr. (Jan. 23) at 169:1-3. From this 51.8%, Dr. Mathiowetz focused only on those households with a Hispanic head of household and calculated an increase in the differential undercount of Hispanics of approximately 0.8 percentage points resulting from rostering omissions.¹⁵ ECF No. 126, Trial Tr. (Jan. 23) at 168:1-170:7 (Mathiowetz); PX-1214 at 14. Putting these two figures together (1.7 + 0.8), Dr. Mathiowetz estimated the citizenship question would result in an increase of the differential undercount of Hispanics by 2.5 percentage points. However, to be conservative, she estimated that the magnitude of the increase in the differential undercount would be at least 2 percentage points. ECF No. 126, Trial Tr. (Jan. 23) at 170:8-14.

104. Dr. Mathiowetz's estimate for Hispanic households is conservative for the same reasons her estimate for noncitizen households is conservative (i.e., that she used historical assumptions about the effectiveness of NRFU and imputation and the relationship between self response and net undercounts). And it is also conservative because she considered only those households with a Hispanic head of household rather than using a broader definition of Hispanic household.

105. Based on the foregoing, the Court finds that a citizenship question will increase the differential undercount of noncitizens and of Hispanics by at least 2 percentage points.

3. Effect of a Differential Undercount

106. A differential undercount of Hispanics and noncitizens of any magnitude will injure the individual Plaintiffs and members of the Organizational Plaintiffs. First, it will cause vote dilution due to intrastate congressional and state legislative redistricting. Further, a differential undercount as low as 1.56 percentage *723points will cause the malapportionment of congressional districts. Finally, a differential undercount of any magnitude will cause Plaintiffs' communities to lose out on federal funding.

a. Intrastate Vote Dilution

107. The following states in which Plaintiffs reside expressly use Decennial Census counts to draw equal-population congressional and state legislative districts: Arizona, California, Florida, Maryland, Nevada, and Texas. See e.g. , Ariz. Const. art. 4, Pt. 2, § 1 (3).¹⁶ Many of these states consider county boundaries in drawing congressional and state legislative district boundaries. See, e.g. , Ariz. Const. art. 4, Pt. 2 § 1 (14); Cal. Const. art. 21, § 2 (d)(4); N.J. Const., art. 4, § 2, ¶ 3 ; Tex. Const. art. 3 § 26. A PUMA is a Public Use Microdata Area originally defined for the dissemination of the Decennial Census and American Community Survey Public Use Microdata Sample (PUMS) research files, which provide respondent-level characteristics. PX-972. PUMAs nest within states, contain at least 100,000 people when drawn, and are built on census tracts and counties. Id. The Census Bureau disseminates period estimates from the ACS for PUMAs as well. Id.

108. If there is a differential undercount of Hispanics and/or noncitizens, a county or PUMA that has a higher percentage of Hispanics and noncitizens relative to the rest of the state will experience a differential undercount relative to the rest of the state and will suffer a reduction in its statewide share of the Decennial Census population count. See ECF No. 99-1, Brace Decl. ¶¶ 11, 30-31, 35; ECF No. 126, Trial Tr. (Jan. 23) at 236:22-239:19 (Brace).

109. Plaintiffs' expert Kimball Brace provided reasonable, reliable, and credible calculations of the likely 2020 population and demographic composition in the counties, PUMAs, and states in which individual Plaintiffs and members of organizational Plaintiffs reside. See ECF No. 99-1 Brace Decl. ¶¶ 14-16, 28, 31, 34; PX-955; PX-965; PX-968; PX-974; PX-975; ECF No. 126, Trial Tr. (Jan. 23) at 242:10-246:23 (Brace). Defendants nitpick at Brace's population and demographic estimates, ECF No. 150 ¶¶ 146-151, but did not provide any evidence disputing Brace's projections. ECF No. 137, Trial Tr. (Jan. 29) at 116:15-117:8 (Gurrea). While Defendants succeeded in convincing the Court that no population projection can perfectly account for potentially catastrophic weather or economic events, ECF No. 150 ¶¶ 146-150, they failed to persuade the Court that Brace's estimate is unreliable because of this imperfection. Further, even if the Court were to assume the worst (i.e., that a weather event on the scale of Hurricane Katrina or an economic event as significant as the Great Recession is set to occur), Defendants did not show how these events would change the ultimate outcome for several of the geographic areas in which Plaintiffs reside. It would be difficult to assume, for example, that a hurricane or a recession could shift the Hispanic or noncitizen population in counties like Hidalgo County in an outcome determinative way. ECF No. 126, Trial Tr. (Jan. 23) at 243:7-244:9 (Brace). That is because in Hidalgo county, 92.7% of residents are Hispanic *724and/or noncitizens as compared to the statewide demographics in Texas where 35% of residents are Hispanic and/or noncitizens. Id. It is therefore extremely unlikely that even a catastrophic event would cause the percentage of these populations in Hidalgo county to drop below the statewide percentage of Hispanic/noncitizen residents. Thus, even accepting some uncertainty, Brace's population estimates provide a reliable baseline for the geographic areas that are relevant to Plaintiffs' potential to experience intrastate vote dilution.

110. Accepting Brace's estimates, the Court finds that in several of Plaintiffs' counties and PUMAs, which will have higher percentages of Hispanic and noncitizen residents compared to the rest of the state as of April 2020, a differential undercount of Hispanics and/or noncitizens of any magnitude will dilute Plaintiffs' vote. Brace Decl. ¶¶ 11, 30-31, 35; ECF No. 126, Trial Tr. (Jan. 23) at 236:22-239:19 (Brace). That is because a differential undercount would cause Plaintiffs who reside in the relevant counties and PUMAs to be drawn into congressional and state legislative districts that are overpopulated relative to other districts in the state. Defendants criticize Brace's use of PUMA data because in some states, such as Texas, laws dictate that counties are the appropriate boundary used for redistricting, rather than a smaller geographic unit. The Court finds this unpersuasive. As Brace explained, some Plaintiffs live in "large counties that contain multiple legislative districts, such as Maricopa County and Clark County," making it "more appropriate to look at smaller geographies" like PUMAs to evaluate vote dilution. Compare ECF No. 150 ¶ 164with ECF No. 99-1, Brace Decl. ¶ 32. However, because Texas uses counties rather than PUMAs to make redistricting determinations, the Court does not rely on Brace's findings related to PUMAs in Texas.

111. According to Brace, the most appropriate measure for assessing intrastate vote dilution is determining whether there is a negative percentage change in the statewide share of the population for a particular county or PUMA. ECF No. 127, Trial Tr. (Jan. 24) 18:7-21:7 (Brace). If a county or a PUMA has a small share of the overall statewide population, the absolute change in the statewide population share could be very small even with a large differential undercount in that area that dilutes the votes of residents of that area. Id. An increase of 2 percentage points in the differential undercount of Hispanics and noncitizens will result in a negative shift in each of the following relevant counties' and PUMAs' share of their statewide Decennial Census count relative to their actual statewide share of the population: Maricopa County, AZ; Santa Cruz County, AZ; Yuma County, AZ; Los Angeles County, CA; Miami-Dade County, FL; Prince George's County, MD; Clark County, NV; Hudson County, NJ; Todd County, SD; Dallas County, TX; El Paso, TX; Harris County, TX; Hidalgo County, TX; Webb County, TX; King County, WA; PUMAs 121, 900, and 700 in Arizona; PUMAs 3735 and 3744 in California; PUMA 8619 in Florida; PUMAs 1101 and 1103 in Maryland; PUMA 406 in Nevada; PUMA 601 in New Jersey. ECF No. 99-1, Brace Decl. ¶ 30; PX-955; PX-965. As a result, the following individual Plaintiffs who reside in these counties and PUMAs will be drawn into congressional and state legislative districts that are overpopulated relative to the other districts in the state, thereby causing them to suffer vote dilution: Alejandro Chavez, Jose Moreno, Richard McCune, Elizabeth Buchanan, Jacob Cunningham, Maegan Ortiz, Lazara Yoelvis Magadan, Catherin Nwosu, Nnabugwu *725Nwosu, T. Carter Ross, Michael Kagan, Raj Mukherji.

112. Even an implausibly conservative increase of .08 percentage points in the differential undercount of individuals living in noncitizen households will cause several Plaintiffs-Alejandro Chavez, Jose Moreno, Richard McCune, Elizabeth Buchanan, Jacob Cunningham, Maegan Ortiz, Lazara Yoelvis Magadan, Catherin Nwosu, Nnabugwu Nwosu, T. Carter Ross, Joanne Wilson, Michael Kagan, Raj Mukherji-to lose political power because the depressed enumeration in their PUMAs will leave them living in overpopulated congressional and state legislative districts. ECF No. 99-1, Brace Decl. ¶ 39; PX-974. This undercount scenario assumes that the citizenship question would cause only a 5.8 percentage point differential drop in self-response among noncitizen households and that 98.58% of this differential drop would be mitigated through NRFU operations. Yet several Individual Plaintiffs would still experience vote dilution.

113. Similarly, just a 0.5 percentage point differential undercount of noncitizens due to rostering omissions would lead to vote dilution according to Brace's credible testimony and calculations. ECF No. 99-1, Brace Decl. ¶ 40; PX-975. This undercount scenario assumes that NRFU operations would be entirely successful in mitigating any differential drop in self response, but that noncitizens would still experience a small differential increase in rostering omissions. ECF No. 99-1, Brace Decl. ¶ 40.

114. Defendants' expert, Dr. Gurrea, performed a "sensitivity analysis" testing Plaintiffs' impact evaluation models with alternative undercount assumptions. ECF No. 137, Trial Tr. (Jan. 29) at 39:13-21. Dr. Gurrea's alternative scenario assumed a NRFU success rate equal to that in 2010 before imputation and assumed no errors based on rostering omissions. Id. Yet Dr. Gurrea did not offer the opinion that there would be no rostering omissions caused by a citizenship question. Id. at 101:13-15. However, even using Dr. Gurrea's alternative scenario, assuming a 2010 NRFU success before imputation and assuming no rostering errors, an impact on vote dilution remains. Id. at 39:13-21.

b. Apportionment

115. Next, Plaintiffs have also proved that the differential undercount caused by a citizenship question will injure Plaintiffs by leading certain jurisdictions to lose seats in the next congressional apportionment. This finding is based on Brace's calculations of the likely population and demographic composition of each state in 2020 when the Decennial Census is conducted. ECF No. 99-1, Brace Decl. ¶¶ 14-16, 21; PX-951-PX-954; ECF No. 126, Trial Tr. (Jan. 23) at 241:22-242:9 (Brace); ECF No. 127, Trial Tr. (Jan. 24) at 8:17-15:12 (Brace). Defendants provided no evidence disputing those population projections, ECF No. 137, Trial Tr. (Jan. 29) at 116:15-117:8 (Gurrea), and for the reasons described above, the Court is not moved by Defendants' criticisms of Brace's calculations as uncertain.

116. The 435 seats in the House of Representatives are apportioned through the Method of Equal Proportions, which Congress adopted in 1941. ECF No. 99-1, Brace Decl. ¶ 18; see also 2 U.S.C. § 2a(a). Using this apportionment method and Brace's population projections, an increase of 2 percentage points in the differential undercount of Hispanics and noncitizens will cause California to lose a congressional seat in the reapportionment process that follows the 2020 Census. Id. ¶ 22; PX-952. In fact, an increase in the differential undercount of Hispanics and noncitizens as low as 1.56 percentage points will cause California to lose a congressional seat in the reapportionment process that follows *726the 2020 Census. ECF No. 99-1, Brace Decl. ¶ 22. Plaintiffs Elizabeth Buchanan, Jacob Cunningham, and Maegan Ortiz reside in California and would be impacted by the loss. ECF Nos. 94-4, 94-6, 94-16.

c. Loss of Federal Funding

117. The Court also finds that a differential undercount of Hispanics and noncitizens will cause states, localities, and their residents to lose access to federal funding from domestic financial assistance programs that allocate funding based on Census-tied geographic formulas. A large number of federal domestic financial assistance programs-including the Surface Transportation Block Grant (STBG) Program, Medicaid, Title I of the Elementary and Secondary Education Act of 1965 (Title I), State Children's Health Insurance Program (CHIP), Supplemental Nutrition Program for Women, Infant, and Children (WIC), and Social Services Block Grants (SSBG)-rely on Decennial Census data to allocate money. ECF No. 99-8, Reamer Decl. ¶¶ 10, 17-18; ECF. No. 99-7, Mingo Decl. ¶ 13; ECF No. 99-4, Gordon Decl. ¶ 14; PX-329. In fiscal year 2017, for example, at least 320 federal programs used census-derived data to distribute approximately $ 900 billion. ECF No. 99-8, Reamer Decl. ¶ 10.

118. Federal programs that allocate funds based on census-derived data are highly sensitive to inaccuracies in those data. Id. ¶ 12. These programs include formulas based on state per capita income relative to U.S. per capita income and those programs that rely, in whole or part, on a state's share of the total U.S. population (state-share programs). Id. ¶ 11. Even modest geographic differences in census accuracy will lead to measurable changes in fund distribution. Id. Of the 24 large federal financial assistance programs Plaintiffs' expert Dr. Andrew Reamer identified, not one has, even temporarily, ever halted the use of state-share and per-capita-income-based funding formulas. Id. ¶ 20.

119. Beginning with the effect of a differential undercount on transportation funding, the STBG Program distributes funds to states and localities to assist with the planning and construction of transportation projects and transportation alternatives. ECF No. 99-7, Mingo Decl. ¶ 10. The intrastate distribution of those federal funds depends on Decennial Census data. Id. ¶ 13. Each state must distribute approximately half of the Surface Transportation Block Grant (STBG) funds it receives to different areas of the state based on population. Id. This distribution of funds is referred to as a suballocation. Id. This intrastate suballocation of the STBG Program has been population-based for 45 years and is not expected to change in the foreseeable future. Id.

120. Every urbanized area with a population over 200,000 receives federal STBG suballocation funds (for transportation projects) and federal transportation alternative set-aside (TA set-aside) suballocation funds (for transportation alternative projects) that must be used in that urbanized area. Id. ¶ 15. In both the STBG suballocation and the TA set-aside suballocation, the proportion of the state's suballocation funds distributed to an urbanized area with a population over 200,000 is equal to (1) that urbanized area's total population, divided by (2) the state's total population, as determined by the most recent Census. Id. A differential undercount of the population of an urbanized area with a population over 200,000 relative to the state as a whole will result in reduced federal transportation funding for the urbanized area under the STBG Program, including the TA set-aside, for the entire decade following the Census. Id. ¶¶ 7, 20. It is very unlikely that a state would provide an urbanized *727area with extra funding out of the state's general STBG allocation funds to make up for a funding shortfall. ECF No. 127, Trial Tr. (Jan. 24) at 45:14-46:13 (Mingo).

121. Plaintiffs' expert Kimball Brace calculated the likely population and demographic composition in 2020 of the urbanized areas and states in which Plaintiffs reside, as well as the 2020 Census counts for each of these areas with a 2 percentage point differential undercount of Hispanics and noncitizens. ECF No. 99-1, Brace Decl. ¶¶ 44-46; PX-961; PX-962; PX-991; PX-992. These urbanized areas are all projected to have total populations over 200,000. See PX-961. Brace's population projections for the 2020 population, as reflected in PX-961, PX-962, PX-991, and PX-992, are reasonable, reliable, and credible, particularly with respect to whether an urbanized area will have higher percentages of Hispanic and noncitizen residents relative to the rest of the state. ECF No. 99-1, Brace Decl. ¶¶ 14-16, 44-46; ECF No. 126, Trial Tr. (Jan. 23) at 242:10-246:23 (Brace).

122. Defendants note that Brace's re-calculation, using a range of undercount assumptions consistent with the Government's own statements and/or Dr. Gurrea's report, does not account for the effects of imputation. ECF No. 150 ¶ 168 (citing ECF No. 126, Trial Tr. (Jan. 23) at 235:11-236:10 (Brace) ). However, this complaint does not persuade the Court to find Brace's calculation unreliable for at least two reasons. First, for the reasons already described above (see § I.C.1-2), the Court is comfortable relying on Plaintiffs' experts' conservative undercount assumptions rather than Dr. Gurrea's assumptions, rendering Defendants' criticism of Brace's recalculations in response to Dr. Gurrea irrelevant. Second, for the reasons described earlier, imputation is unlikely to cure depressed participation at the self-response or NRFU stage because the data that it will be relying on to model Hispanic and noncitizen households will be inaccurate. Otherwise, Defendants provided no evidence disputing Brace's projections. ECF No. 137, Trial Tr. (Jan. 29) at 116:15-117:8 (Gurrea).

123. Based on Brace's calculations, the following urbanized areas in which individual Plaintiffs reside will have higher percentages of Hispanic and noncitizen residents relative to the rest of the state as of April 1, 2020: Los Angeles-Long Beach-Anaheim, California; Miami, Florida; Atlanta, Georgia; the Maryland portion of Washington, DC-VA-MD; Las Vegas-Henderson, Nevada; the New Jersey portion of New York-Newark, NY-NJ-CT; Houston, Texas; Laredo, Texas; and McAllen, Texas. PX-961; PX-962; PX-991; PX-992. A differential undercount of Hispanics and/or noncitizens of any magnitude will cause these urbanized areas to suffer a differential undercount relative to the rest of their respective states, which will result in reduced federal transportation funding for the urbanized area under the STBG Program, including the TA set-aside, for the entire decade following the 2020 Census. ECF No. 99-7, Mingo Decl. ¶¶ 7, 20. The following Individual Plaintiffs and members of Organizational Plaintiffs reside in these urbanized areas: Elizabeth Buchanan, Jacob Cunningham, Maegan Ortiz, Lazara Yoelvis Magadan, Catherine Nwosu, Nnabugwu Nwosu, T. Carter Ross, Michael Kagan, Michael Kravitz, Robyn Kravitz, Raj Mukherji, Diana Alexander, Sarah Bryan, Martha Sanchez, Sonia Casarez Shafer, Virginia Garcia, Juanita Valdez-Cox, Joanne Wilson. ECF Nos. 94-1, 94-3, 94-4, 94-6-94-11, 94-14-94-20, 133-6; ECF No. 125, Trial Tr. (Jan. 22) at 146:1 (Valdez-Cox).

*728124. Using Brace's population projections, Plaintiffs' expert Roger Mingo calculated the approximate loss in STBG suballocation funding and TA set-aside suballocation funding that an increase of 2 percentage points in the differential undercount of Hispanics and noncitizens will cause for the following urbanized areas following the 2020 Census: Los Angeles-Long Beach-Anaheim, California; Miami, Florida; Atlanta, Georgia; the Maryland portion of Washington, DC-VA-MD; Las Vegas-Henderson, Nevada; the New Jersey portion of New York-Newark, NY-NJ-CT; Houston, Texas; Laredo, Texas; and McAllen, Texas. ECF No. 99-7, Mingo Decl. at 2-3 ¶ 8, 10-11 Tables 2 and 3. There is a substantial risk that such a reduction in STBG funding will negatively affect one or more of the roads, highways, bridges, sidewalks, trails, or other structures in that urbanized area. Id. ¶¶ 8-9, 25-28. Such a reduction is almost certain to delay, eliminate, or change the scope of a transportation or transportation-alternative project of the type covered by the STBG Program in that urbanized area. Id.

125. Next, the Court turns to the effects of a differential undercount on Medicaid funding. The allocation of federal Medicaid funds depends on Decennial Census data. ECF No. 99-2, Carruth Decl. ¶ 17. The federal government reimburses state Medicaid expenditures, on a quarterly basis, based on the state's Federal Medical Assistance Percentage (FMAP). Id. ¶ 13. The statutory formula to calculate a state's FMAP compares (1) the state's per capita income over the three most recent years for which data are available to (2) the U.S. per capita income over the three most recent years for which data are available. Id. ¶ 14. The higher the state's per capita income is relative to the national average, the lower its FMAP and, therefore, the lower its federal reimbursement for Medicaid spending. Id. This statutory formula has been in place for more than 50 years and is not expected to change. Id.

126. A state's per capita income is calculated as (1) the state's personal income, divided by (2) the state's population. Id. ¶ 17. A state's population is derived from the Census Bureau's annual midyear (July 1) Population Estimates, which are based on the most recent Census. Id. A state's population will be artificially low if state residents are not counted by the Census. Id. Similarly, the U.S. per capita income is calculated as (1) the U.S. personal income, divided by (2) the U.S. population. Id. ¶ 18. The sources used to determine state personal income and state population are also used to determine the U.S. personal income and U.S. population. Id. Thus, an undercount on the Decennial Census in a state relative to the U.S. as a whole will falsely inflate the state's per capita income relative to the U.S. per capita income. Id. ¶ 19. This will result in a lower FMAP and, therefore, reduced federal Medicaid funding for the state until the next Decennial Census is fully phased into the FMAP calculations, provided that the state's FMAP is above the statutory minimum of 50. Id.

127. Even a 0.01% reduction in a state's FMAP poses a substantial risk that Medicaid beneficiaries' health care benefits and/or medical costs will be negatively affected, provided that the state's FMAP is above the statutory minimum of 50. Id. The FMAPs for Arizona, Florida, Nevada, New Mexico, and Texas are well above 50. See PX-803 at 3. If a state's FMAP is reduced, it is unlikely that a state like Texas would use state funds to make up for the lost federal funds and continue the status quo of benefits and services offered through its Medicaid program. ECF

*729No. 127, Trial Tr. (Jan. 24) at 105:17-23 (Carruth).

128. Brace calculated each state's likely population and demographic composition in July 1 of 2020-2022, as well as the Census's Population Estimates for each state for July 1 of each year assuming a 2 percentage point differential undercount of Hispanics and noncitizens. ECF No. 99-1, Brace Decl. ¶¶ 14-16, 41-43; PX-958 to PX-960. Defendants object to these calculations for the same reasons previously addressed by the Court, ECF No. 150 ¶ 169, and the Court continues to find Brace's projections reasonable, reliable, and credible, particularly with respect to whether a state will have higher percentages of Hispanic and noncitizen residents relative to the rest of the U.S. ECF No. 99-1, Brace Decl. ¶¶ 14-16, 44-46; ECF No. 126, Trial Tr. (Jan. 23) at 242:10-246:23 (Brace).

129. Using these population projections, Plaintiffs' expert Lisa Carruth calculated that a 2 percentage point differential undercount of Hispanics and noncitizens will cause the following states to lose federal Medicaid funding for several years following the 2020 Census: Arizona, Florida, Nevada, New Mexico, and Texas. ECF No. 99-2, Carruth Decl. ¶¶ 8, 21. Texas would lose more than $ 150 million in just a single year. Id. ¶ 23. And there is a substantial risk that a reduction in federal Medicaid funds will negatively affect Medicaid beneficiaries' health care benefits and/or medical costs. Id. ¶¶ 9, 24-32. Plaintiffs Sonia Casarez Shafer and Sarah Bryan reside in Texas and have children who receive health insurance under the Texas Medicaid program. ECF Nos. 94-3 & 94-19. Members of Plaintiffs LUPE and PAZ also reside in Arizona and Texas and receive health insurance under the Arizona and Texas Medicaid programs. ECF No. 133-4, Navarrete Decl. ¶¶ 5-6; ECF No. 125, Trial Tr. (Jan. 22) at 146:1, 158:21-24 (Valdez-Cox).

130. Ms. Carruth's calculations are reasonable and reliable, and the Court credits them. Defendants argue that Ms. Carruth's analysis should not be relied upon because it is based on Mr. Brace's calculations, but as previously discussed, the Court accepts Brace's population estimates as reliable. Defendants' only other response to Ms. Carruth's analysis is Dr. Gurrea's sensitivity analysis. But even using Dr. Gurrea's implausibly conservative undercount assumptions, some estimated loss of federal Medicaid funds for the State of Texas will result. ECF No. 137, Trial Tr. (Jan. 29) at 66:17-67:13 (Gurrea).

131. The Court also finds that Plaintiffs' communities will receive reduced Title I education funding. Title I is the single largest federal funding stream for compensatory elementary and secondary education. ECF No. 99-4, Gordon Decl. ¶ 10. School districts use Title I funds to support educationally disadvantaged students. Id. ¶ 11. Title I directs funds to school districts through four separate grant programs: basic grants, concentration grants, targeted grants, and education finance incentive grants. Id. ¶ 10. Though these grants each have their own formula determining district-level allocations, a school district treats the sum of funds distributed through the four formulas as a single funding source. Id.

132. Annual estimates of the number of school-aged children in poverty in each school district are provided by the Census Bureau's Small Area Income Poverty Estimates (SAIPE). Id. ¶ 14. These estimates are derived by applying the poverty rate for the school district (or its component pieces) to the estimate of all school-aged children in the district (or its component pieces) provided by the Census Bureau's annual Population Estimates, which are *730based on Decennial Census counts of the number of school-aged children. Id. An undercount on the Decennial Census of the school-aged children residing in a school district relative to other school districts nationwide will erroneously deflate the school district's "eligibility count" relative to the "eligibility count" of other districts. Id.

133. Using Title I's four distinct formulas, Plaintiffs' expert Dr. Nora Gordon estimated an effect on district-level Title I allocations of a 2 percentage point differential undercount of Hispanics relative to the rest of the population under a range of undercount scenarios for the rest of the population. Id. ¶ 8. For each of the undercount scenarios examined, Dr. Gordon's analysis shows that many of the Plaintiffs' school districts would experience a decline in Title I funding dollars as a consequence of including the citizenship question on the 2020 Census questionnaire. Id. ¶ 8.

134. A 2 percentage point differential undercount of Hispanics will cause several school districts to lose Title I funds. ECF No. 99-4, Gordon Decl. ¶ 33 (Table 1). The precise number of school districts that would lose funding depends on the absolute undercount of the Hispanic and non-Hispanic population that would be caused by the citizenship question. Id. Even under the assumptions generating the most limited impact (a 3.5% undercount for Hispanics and a 1.5% undercount for the rest of the population resulting in a 2 percentage point differential undercount for Hispanics), the following districts in Arizona, Nevada, and California where Plaintiffs reside and/or their children attend school would experience a loss in funding: Alhambra Elementary District, Balsz Elementary District, Isaac Elementary District, Roosevelt Elementary District, Somerton Elementary District, and Union Elementary District in Arizona; Clark County School District in Nevada; Edinburg Consolidated Independent School District, Hidalgo Independent School District, Lyford Consolidated Independent School District, Pharr-San Juan-Alamo Independent School District, and Rio Grande City Consolidated Independent School District in Texas. Id ; PX-844. In the case of the Clark County School District, the loss in funding would exceed $ 500,000 for a single year. PX-844. According to Dr. Gordon, it is highly likely that the immediate effect of a reduction in a school district's Title I funding would be a reduction in educational services available to students in the school district. ECF No. 99-4, Gordon Decl. ¶ 9.

135. Defendants ask the Court to disregard Dr. Gordon's calculations because the net undercount scenarios she used were provided by Plaintiffs' counsel and were not generated by Plaintiffs' experts. ECF No. 150 ¶ 233. Unlike Plaintiffs' other experts who modeled the impact of various differential undercount scenarios on federal funding, Dr. Gordon's calculations used inputs for various net undercount scenarios of Hispanics versus the rest of population, which would yield a 2 percentage point differential undercount of Hispanics. ECF No. 99-4, Gordon Decl. ¶ 33 (table 1). For example, a 2% net undercount of Hispanics combined with no net undercount for the rest of the population would yield a 2 percentage point differential undercount of Hispanics. Id. Similarly, a 3.5% net undercount of Hispanics combined with a 1.5% net undercount of the rest of the population would lead to a 2 percentage point differential undercount of Hispanics. The Court, however, credits Dr. Gordon's testimony regarding the most limited likely effect of a 2 percentage point differential undercount of Hispanics. For all the reasons already described, the weight of the evidence shows that a 2 percentage point differential undercount is a conservative estimate, and the Court relies only on *731Dr. Gordon's calculations based on the assumptions generating the most limited impact (a 3.5% undercount for Hispanics and a 1.5% undercount for the rest of the population). The Court does not rely on Dr. Gordon's analysis to make findings about the net undercount rates from which a 2 percentage point differential undercount will ultimately be derived, and instead understands that Dr. Gordon used various net undercount scenarios to illustrate the range of possible effects on Title I.

136. Further, according to Dr. Gordon's analysis, as the net undercount rate for the non-Hispanic population gets closer to zero, more school districts will lose Title I funding. See ECF No. 99-4, Gordon Decl. ¶ 33 (Table 1). And although none of Plaintiffs' experts offered predictions about the net undercount rates from which an increase of 2 percentage points differential undercount will be derived, the record does include evidence that Dr. Gordon's most limited-impact assumptions (a 3.5% undercount for Hispanics and a 1.5% undercount for the rest of the population) are conservative. For example, the record reflects that in 2010, net undercounts of Hispanics of about 1.5% were actually offset by overcounting non-Hispanic whites by approximately 0.8%. PX-267 at Table 7. Thus, assuming that the non-Hispanic population will be undercounted by 1.5%, when in 2010, non-Hispanic whites were in fact overcounted, likely downplays the impact of an increase of 2 percentage points in the differential undercount among Hispanics on Title I funding. Taken together, the Court credits Dr. Gordon's finding that at least 36 school districts will lose Title I funding based on an increase of 2 percentage points in the differential undercount of Hispanics.

137. Finally, the record reflects that for programs with allocation formulas based on a state's population or per capita income relative to the U.S., a differential undercount of noncitizens-even as low as 0.1 percentage points-would lead to measurable fiscal loss for states with percentages of noncitizens above the nationwide average. ECF No. 99-8, Reamer Decl. ¶¶ 19-20. To illustrate his opinions, Dr. Reamer analyzed the estimated effect of (1) a 5.8 percentage point differential undercount of noncitizens, and (2) a 5.8 percentage point differential non-response rate for noncitizens, mitigated by an 86.63 percent NRFU success rate, on four funding formulas: Medicaid, CHIP, WIC, and SSBG. Id. ¶¶ 36-38. The Court credits Dr. Reamer's opinion despite Defendants' argument that the Court should disregard it because Dr. Reamer's undercount assumptions were calculated by an expert who did not testify at trial. Experts may render opinions even if based on inadmissible evidence, so long as, as here, the evidence is of the type reasonably relied upon by experts in that field, Fed. R. Evid. 703 ; Daubert v. Merrell Dow Pharms., Inc. , 509 U.S. 579, 595, 113 S.Ct. 2786, 125 L.Ed.2d 469 (1993).

138. Dr. Reamer's core opinion that a differential undercount will cause some loss in funding does not rely on any assumptions about the magnitude of such an undercount. Id. ("Even a 0.1% percent differential undercount, for example, would cause losses to state share programs and FMAP programs for states with large noncitizen populations relative to the national average."). Further, the record shows that the undercount assumptions Dr. Reamer relied on, while less conservative than those calculated by Dr. Mathiowetz, are nonetheless reasonable. As previously discussed, it is not unreasonable to assume that NRFU will not be effective at remedying reduced self response by noncitizens because noncitizens are unlikely to respond to enumerators or be accounted for *732by proxies for the same reasons that they tend to conceal themselves at the self-response stage. Although Defendants have repeatedly flagged that the illustrations Dr. Reamer relies upon use undercount estimates calculated by an outside expert, they did not present any evidence disputing Dr. Reamer's calculations. They did not, for example, use Dr. Reamer's methods with more conservative inputs to demonstrate that the Court should not rely on Dr. Reamer's examples. In this context, the Court finds Dr. Reamer's calculations reasonable, reliable, and credible.

139. Under the less conservative undercount scenarios Dr. Reamer analyzed, Texas, Florida, Arizona, Nevada, Hawaii, Washington, and Illinois would have lost Medicaid funds. Id. ¶¶ 50-51. Each of the following states would have lost CHIP funds (money for health insurance to certain low-income children) under both undercount scenarios Dr. Reamer analyzed: Texas, Florida, Nevada, Hawaii, Arizona, and Washington. Id. ¶¶ 61-62. Additionally, California, Texas, New York, Florida, New Jersey, Nevada, Arizona, and Hawaii would have lost WIC funds (money for certain low-income pregnant, breastfeeding, or postpartum women, infants, and young children) based on the inputs used by Dr. Reamer. Id. ¶¶ 69-70. Finally, each of the following states would have lost SSBG funds under both undercount scenarios Dr. Reamer analyzed: California, Texas, New York, Florida, New Jersey, Nevada, Arizona, Hawaii, Washington, Maryland, Illinois, and Massachusetts. Id. ¶¶ 74-75.

140. Notably, even were the Court to rely only on Defendants' expert's extremely conservative undercount assumptions, funding losses of some non-nominal magnitude will occur in areas that certain Plaintiffs reside. ECF No. 137, Trial Tr. (Jan. 29) at 125:21-126:20 (Gurrea).

4. Additional Harms

141. The Court also finds that inclusion of a citizenship question on the 2020 Census will harm the quality of the Census data regardless of whether it also causes a differential undercount of Hispanics and/or noncitizens. Early and often, the Census Bureau has concluded that a citizenship question will harm data quality because it will lower self-response rates and increase the NRFU workload, and "data obtained from NRFU have greater erroneous enumeration and whole-person imputation rates." PX-22 at 5; see also PX-25 at 4-5; PX-162 at 41; ECF No. 138, Trial Tr. (Jan. 31) at 19:9-18 (Abowd); PX-1194 at 881:19-882:5, 885:17-21, 952:23-953:14. One reason for this is that data for households that choose not to self respond are more likely to be collected from a proxy who "has less accurate information than self-responders." PX-22 at 6 (AR 1282); see also PX-162 at 41. Further, the Census Bureau concedes that whole-person imputation is not very accurate. Census Bureau 30(b)(6) Dep. Vol. I at 253:7-15; see also ECF No. 138, Trial Tr. (Jan. 31) at 92:21-93:3 (Abowd) (Census Bureau treats whole-person imputations as an error). Finally, the degradation of Census data quality caused by the citizenship question will not be mitigated. ECF No. 138, Trial Tr. (Jan. 31) at 180:17-20 (Abowd).

142. Several Organizational Plaintiffs rely on the accuracy of Census data for purposes of strategic planning and communication, resource allocation, and advocacy. See, e.g. , ECF Nos. 94-21, Garcia Decl. ¶ 4; 94-23, Gonzalez Decl. ¶ 4; 94-22, Tso Decl. ¶ 4; ECF No. 125, Trial Tr. (Jan. 22) at 198:16-199:5 & 204:12-25 (Park). For example, MinKwon heavily relies on data from the Census Bureau for every single report that they produce and "all of the community organizations that [MinKwon] work[s] with, including the MinKwon Center, *733[ ] couldn't do what [they] do without all of the [Census] data." ECF No. 125, Trial Tr. (Jan. 22) at 198:16-199:5 (Park). The degradation of Census data quality would impact these organizations' ability to accurately allocate their limited resources. ECF No. 139, Trial Tr. (Jan. 31) at 20:22-21:1 (Abowd).

143. In addition to potentially throwing sand in the gears of the Organizational Plaintiffs' missions by reducing data quality, adding a citizenship question to the 2020 Census will also force the Organizational Plaintiffs to divert more of their limited resources to Census outreach. See e.g. , ECF No. 125, Trial Tr. (Jan. 22) at 175:5-176:3 (Valdez-Cox); ECF No. 133-3, Salas Decl. ¶¶ 12-14; ECF No. 133-1, Garcia Decl. ¶ 6.

144. For example, LUPE, a membership organization and community union that seeks to build strong communities in the Rio Grande Valley, has already begun and will continue to increase its Census outreach and advocacy. ECF No. 125, Trial Tr. (Jan. 22) at 175:5-176:3 (Valdez-Cox). Since the decision to add the citizenship question was announced, LUPE began receiving calls and inquiries about the citizenship question. Id. at 172:21-173:5. Responding to community fear about answering a citizenship question and a lack of understanding about the consequences, id . at 173:6-175:4, LUPE has diverted staff time away from advocacy and education around Texas Senate Bill 4 to community education on the importance of the Census and the repercussions of not responding to the survey. See, e.g., id. at 175:14-176:3. The addition of the citizenship question has also impacted all of LUPE's core programs in that the organization must divert staff resources away from other core programs-such as income tax filing, assistance with social service benefit applications, and community organizing-to respond to concerns about the citizenship question and educate members on the importance of responding to the Census. Id. at 176:17-178:7. LUPE has diverted social media resources to mitigating the impact of the citizenship question. Id. Additionally, instead of having candidate forums or forums on other issues, LUPE has organized community forums on the Census and the impact of an undercount. Id. at 168:4-8.

145. Defendants question whether the Organizational Plaintiffs will in fact divert resources in response to a citizenship question. ECF No. 150 ¶ 248-253. However, as Defendants' expert Dr. Abowd testified, as a result of the citizenship question, the Census Bureau's "trusted voices"-community organizations that the Bureau partners with for outreach to hard-to-count groups-will need to expend additional resources getting out the message that participating in the Census is safe and important. ECF No. 138, Trial Tr. (Jan. 30) at 193:5-9 (Abowd); see also PX-696 at 699:3-699:10 (Barreto). Several of the Organizational Plaintiffs that Defendants question are part of the Census Bureau's "trusted voices" program and will need to devote additional resources to convincing the immigrant community to participate in the 2020 Census in light of the citizenship question. See, e.g. , ECF No. 133-1, Garcia Decl. ¶¶ 6-11; ECF No. 133-2, Gonzalez Decl. ¶¶ 9-13; ECF No. 133-3, Salas Decl. ¶¶ 11-15; ECF No. 133-4, Navarrete Decl. ¶¶ 7-11; ECF No. 133-5, Tso Decl. ¶¶ 9-12; ECF No. 125, Trial Tr. (Jan. 22) at 173:6-178:7 (Valdez-Cox); id . at 214:5-218:11 (Park).

II. Conclusions of Law

The United States Constitution mandates that every ten years, Congress take on the herculean task of counting "the whole number of persons in each state."

*734U.S. Const., Art. I, § 2, cl. 3 and Am. XIV § 2 (The Enumeration Clause). This "actual Enumeration," also known as the Decennial Census, is used to apportion representatives in the U.S. House of Representatives, draw intra-state congressional and state legislative districts of equal proportion, and allocate billions in federal funding for numerous federal programs. See Baldrige v. Shapiro , 455 U.S. 345, 353 & n. 9, 102 S.Ct. 1103, 71 L.Ed.2d 199 (1982). Congress has delegated the Decennial Census to the Secretary of Commerce, allowing the Secretary to conduct the survey "in such form and content as he may determine." 13 U.S.C. § 141(a) (The Census Act). While fulfilling the constitutional requirement that the Census provide "an actual enumeration" of the United States' population, U.S. Const., Art. I, § 2, cl. 3 and Am. XIV § 2, the Secretary is also "authorized to obtain" other "necessary" information. 13 U.S.C. § 141(a). Collected data must be kept confidential and used only for the purpose for which it is supplied. 13 U.S.C. §§ 8(b), 9(a). And "[t]o the maximum extent possible and consistent with the kind, timeliness, quality and scope of the statistics required," the Secretary must "acquire and use information available from" existing administrative records "instead of conducting direct inquiries" through questionnaires. 13 U.S.C. § 6.¹⁷

Section 141(f) of the Census Act details the timeline that the Secretary must follow as the Department of Commerce prepares for the Decennial Census:

[T]he Secretary shall submit to the committees of Congress having legislative jurisdiction over the census-

(1) not later than 3 years before the appropriate census date, a report containing the Secretary's determination of the subjects proposed to be included, and the types of information to be compiled, in such census;

(2) not later than 2 years before the appropriate census date, a report containing the Secretary's determination of the questions proposed to be included in such census; and

(3) after submission of a report under paragraph (1) or (2) of this subsection and before the appropriate census date, if the Secretary finds new circumstances exist which necessitate that the subjects, types of information, or questions contained in reports so submitted be modified, a report containing the Secretary's determination of the subjects, types of information, or questions as proposed to be modified.

13 U.S.C. § 141(f).

The APA also cabins the Secretary's otherwise broad discretion by providing "the procedures by which federal agencies are held accountable to the public." Franklin v. Massachusetts , 505 U.S. 788, 796, 112 S.Ct. 2767, 120 L.Ed.2d 636 (1992). In an APA challenge, "[t]he reviewing court shall ... hold unlawful and set aside agency action, findings, and conclusions found to be," among other things, "arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law; contrary to constitutional right, power, *735privilege, or immunity; in excess of statutory jurisdiction, authority, or limitations, or short of statutory right; [or] without observance of procedure required by law." 5 U.S.C. §§ 706(2)(A)-(D).

The Census Bureau is a principal statistical agency within the federal statistical system. 79 Fed. Reg. 71,610, at 71,610. As a statistical agency, the Census Bureau is subject to the standards and directives of the Office of Management and Budget (OMB) under the Paperwork Reduction Act (PRA), 44 U.S.C. §§ 3501 - 21 ; 5 C.F.R. § 1320.18(c). The Census Bureau has also published its own Statistical Quality Standards, which the Bureau follows in conducting surveys. PX-260 at 5. OMB's Statistical Policy Directive No. 2, entitled "Standards and Guidelines for Statistical Surveys," requires the Census Bureau to pretest survey components if they have not been successfully used before to "ensure that all components of a survey function as intended when implemented in the full-scale survey" and that "measurement error is controlled." Office of Mgmt. and Budget, Statistical Policy Directive No. 2, Standards and Guidelines for Statistical Surveys at §§ 1.3, 1.4, 2.3 (2006) (PX-359); see also 71 Fed. Reg. 55, 522 (Sept. 22, 2006).

The Census Bureau's Statistical Quality Standards, which implement the OMB's Statistical Policy Directives, PX-260 at 5, similarly require that "[d]ata collection instruments and supporting materials must be pretested with respondents to identify problems (e.g., problems related to content, order/context effects, skip instructions, formatting, navigation, and edits) and then refined, prior to implementation, based on pretesting results." Id. at 18. Pretesting is required unless the Census Bureau obtains a waiver or uses a question that has "performed adequately in another survey." Id. OMB's Statistical Policy Directive No. 2 requires the Census Bureau to (i) design its surveys "to achieve the highest practical rates of response, commensurate with the importance of survey uses," and (ii) administer the survey in a way that "maximiz[es] data quality" while "minimizing respondent burden and cost." Office of Mgmt. and Budget, Statistical Policy Directive No. 2, Standards 1.3, 2.3 (PX-359).

OMB's Statistical Policy Directive No. 1 requires the Census Bureau to "apply sound statistical methods to ensure statistical products are accurate," and to "produce data that are impartial, clear, and complete and are readily perceived as such by the public." Office of Mgmt. and Budget, Statistical Policy Directive No. 1., Fundamental Responsibilities of Federal Statistical Agencies and Recognized Statistical Units , 79 Fed. Reg. 71,610, 71,615 (Dec. 2, 2014) (PX-354 at 7). To guarantee impartiality, statistical agencies, including the Census Bureau, "must function in an environment that is clearly separate and autonomous from the other administrative, regulatory, law enforcement, or policy-making activities within their respective Departments" and "must be able to conduct statistical activities autonomously when determining what information to collect and process." Id. ; see also id. at 4 (citing PX-355 at 72, National Research Council of the National Academy of Sciences, Principles and Practices for a Federal Statistical Agency , Principle 4 ("A Federal statistical agency must be independent from political and other undue external influence in developing, producing, and disseminating statistics.") )

With this general legal framework in mind, the Court turns to the issue of standing before addressing the merits of Plaintiffs' APA and constitutional claims.

*736A. Standing

Defendants challenge the standing of both the Individual and Organizational Plaintiffs in this action. The Article III standing doctrine requires a plaintiff, as the party invoking the Court's jurisdiction, to establish three elements at trial: (1) a concrete and particularized injury-in-fact, either actual or imminent; (2) a causal connection between the injury and defendants' challenged conduct, such that the injury is "fairly trace[able] to the challenged action of the defendant"; and (3) a likelihood that the injury suffered will be redressed by a favorable decision. Lujan v. Defenders of Wildlife , 504 U.S. 555, 560-61, 112 S.Ct. 2130, 119 L.Ed.2d 351 (1992) ; see also Spokeo, Inc. v. Robins , --- U.S. ----, 136 S.Ct. 1540, 1547, 194 L.Ed.2d 635 (2016). The "presence of one party with standing is sufficient to satisfy Article III's case-or-controversy requirement." Rumsfeld v. Forum for Acad. and Institutional Rights, Inc. , 547 U.S. 47, 53 n.2, 126 S.Ct. 1297, 164 L.Ed.2d 156 (2006) ; see Carey v. Population Servs. Int'l , 431 U.S. 678, 682, 97 S.Ct. 2010, 52 L.Ed.2d 675 (1977). An organizational plaintiff's standing can be based on: (1) representational standing, where the injury is to the organization's members, Hunt v. Wash. State Apple Advert. Comm'n , 432 U.S. 333, 343, 97 S.Ct. 2434, 53 L.Ed.2d 383 (1977), or (2) direct organizational standing, where the injury is to the organization itself, Havens Realty Corp. v. Coleman , 455 U.S. 363, 379, 102 S.Ct. 1114, 71 L.Ed.2d 214 (1982).

1. Injury to Individual Plaintiffs and Members of Organizational Plaintiffs

To satisfy standing's injury-in-fact requirement, proof of a "future injury may suffice if the threatened injury is 'certainly impending,' or there is a 'substantial risk that the harm will occur.' " Susan B. Anthony List v. Driehaus , 573 U.S. 149, 158, 134 S.Ct. 2334, 189 L.Ed.2d 246 (2014) (quoting Clapper v. Amnesty Int'l USA , 568 U.S. 398, 409, 133 S.Ct. 1138, 185 L.Ed.2d 264 (2013) ); see also Clapper , 568 U.S. at 414 n.5, 133 S.Ct. 1138 (recognizing that standing does not require plaintiffs to "demonstrate that it is literally certain that the harms they identify will come about."). Although Defendants repeatedly claim that Plaintiffs have provided no "credible quantitative evidence" to support various concrete, non-speculative injuries, e.g. , ECF No. 150 ¶¶ 281, 282, 285, for standing purposes, a plaintiff may rely on qualitative evidence to establish a substantial risk of future injury. See e.g., Monsanto Co. v. Geertson Seed Farms , 561 U.S. 139, 153 n.3, 130 S.Ct. 2743, 177 L.Ed.2d 461 (2010) (holding that plaintiffs had standing based on their declarations' unquantified assertions of risk of potential future harm); Air All. Houston v. EPA , 906 F.3d 1049, 1058-59 (D.C. Cir. 2018) (holding that plaintiffs had standing based on their declarations' statements that they had previously been put at risk by chemical disasters and that the same risks that caused the previous disasters still existed).

In this context, Plaintiffs have demonstrated that, at minimum, there is a substantial risk that the addition of a citizenship question to the 2020 Census will lead to a differential undercount of Hispanics and/or noncitizens that will cause Individual Plaintiffs and certain members of Organizational Plaintiffs to suffer three kinds of injury: (1) vote dilution from intra-state redistricting; (2) vote dilution from malapportionment of congressional seats; and (3) loss of federal funding to their states and localities.

Turning to the first injury Plaintiffs allege, the Supreme Court has found that plaintiffs challenging a proposed Census *737procedure can establish standing "on the basis of the expected effects" challenged Census Bureau conduct will have "on intrastate redistricting"-in particular, based on a factual finding that certain jurisdictions are "substantially likely" to "suffer vote dilution in state and local elections." Dep't of Commerce v. U.S. House of Representatives , 525 U.S. 316, 332-33, 119 S.Ct. 765, 142 L.Ed.2d 797 (1999) (internal quotation marks omitted). In Department of Commerce v. U.S. House of Representatives , residents of counties that "were substantially likely to lose population" (i.e., be undercounted) under a proposed Census procedure satisfied Article III's injury-in-fact requirement. Id. at 334, 119 S.Ct. 765. To arrive at this conclusion, the Court required only proof that (1) certain states relied on federal Decennial Census data for intrastate redistricting, (2) voters in certain counties in those states were "substantially likely ... to suffer vote dilution as a result of the [Census Bureau's] plan," and (3) plaintiffs were among the voters who lived in those counties in those states. Id. at 332-33, 119 S.Ct. 765 (internal citation omitted). The Court, therefore, found that expert testimony that plaintiffs were substantially likely to suffer vote dilution in state and local elections was sufficient to establish injury-in-fact. Id. at 333-34, 119 S.Ct. 765.

So too here. Through Mr. Brace's expert testimony, Plaintiffs have proven that a differential undercount of Hispanics and/or noncitizens of any magnitude will result in a negative shift in the statewide share of the population for the counties and PUMAs in states in which the following Plaintiffs reside: Alejandro Chavez, Jose Moreno, Elizabeth Buchanan, Jacob Cunningham, Maegan Ortiz, Lazara Yoelvis Magadan, Catherine Nwosu, Nnabugwu Nwosu, T. Carter Ross, Michael Kagan, Raj Mukherji. See Findings of Fact ¶¶ 111-13. Each of the states in which these Plaintiffs reside uses Decennial Census population counts to draw Congressional and state legislative districts of equal population. See id. ¶ 107. See also Ariz. Const. art. 4, Pt. 2, § 1 (3); Cal. Const. art. 21, § 1 ; Nev. Const. art. 15, § 13 ; Fla. Stat. § 11.031(1) ; Md. Code Ann., Elec. Law § 8-701 ; N.J. Const. art. 4, § 2 ¶¶ 1, 3 ; Tex. Const. art. 3, § 26. Thus, because Plaintiffs have shown that a differential undercount of Hispanics and/or noncitizens of some magnitude is virtually certain to occur, they have also demonstrated that the addition of a citizenship question creates the substantial risk that certain Plaintiffs will suffer vote dilution, thereby satisfying standing's injury-in-fact requirement.

Next, Plaintiffs have shown that certain Individual Plaintiffs and members of certain Organizational Plaintiffs will suffer vote dilution because a 2 percentage point differential undercount of Hispanics and noncitizens will lead California to lose a congressional seat. Where a plaintiff challenges a proposed Census procedure that would lead to an "expected loss of a Representative to the United States Congress," this "undoubtedly satisfies the injury-in-fact requirement of Article III standing." Dep't of Commerce , 525 U.S. at 331-32, 119 S.Ct. 765. The loss of a Representative means that Plaintiffs' "votes will be diluted," and faced with the loss of a Representative, the harm of vote dilution "is 'concrete' and 'actual or imminent, not 'conjectural' or 'hypothetical.' " Id. (quoting Whitmore v. Arkansas , 495 U.S. 149, 155, 110 S.Ct. 1717, 109 L.Ed.2d 135 (1990) ) (quotation missing in original); accord Carey v. Klutznick , 637 F.2d 834, 836-38 (2d Cir. 1980) ; cf. Utah v. Evans , 536 U.S. 452, 458, 122 S.Ct. 2191, 153 L.Ed.2d 453 (2002).

*738Because it is substantially likely that the citizenship question will lead to a differential undercount of Hispanics and noncitizens of at least 2 percentage points, see Findings of Fact § I.C.2, and because a differential undercount of Hispanics and noncitizens as low as 1.56 percentage points will cause California to lose a congressional seat, see id. ¶¶ 107, 116, it is substantially likely that eligible voters residing in California-including Plaintiffs Elizabeth Buchanan, Jacob Cunningham, and Maegan Ortiz, as well as members of Organizational Plaintiffs who reside in California, see, e.g. , ECF No. 133-3, Salas Decl. at ¶¶ 4-5,-will suffer vote dilution as a result of the addition of the citizenship question to the 2020 Census. Defendants' position that this injury is not sufficiently imminent because the Census is still months away, ECF No. 50 ¶ 293, has been rejected by the Supreme Court. See Dep't of Commerce , 525 U.S. at 327, 332, 119 S.Ct. 765 (finding vote dilution caused by Census modifications sufficiently "imminent" in two cases brought twenty-five months before the Census date).

Additionally, given the Court's factual findings, many of the Individual Plaintiffs and certain members of the Organizational Plaintiffs have met standing's injury-in-fact requirement by showing they will suffer a loss of funding from federal programs that distribute money using Census data. "[C]itizens who challenge a census undercount on the basis ... that improper enumeration will result in loss of funds" to their state or locality have established a concrete and particularized injury. Carey , 637 F.2d at 838 ; Glavin v. Clinton , 19 F.Supp.2d 543, 550 (E.D. Va. 1998) (three-judge panel) ("As a matter of law, allegations of decreased federal and state funding is fairly traceable to population counts reported in the decennial census."), aff'd sub nom. Dep't of Commerce v. U.S. House of Representatives , 525 U.S. 316, 119 S.Ct. 765, 142 L.Ed.2d 797 (1999).

A differential undercount of Hispanics and/or noncitizens of any magnitude will cause the following urbanized areas to lose Surface Transportation Block Grant (STBG) suballocation funding and federal transportation alternative set-aside (TA set-aside) suballocation funding following the 2020 Census: Los Angeles-Long Beach-Anaheim, California; Miami, Florida; Atlanta, Georgia; the Maryland portion of Washington, DC-VA-MD; Las Vegas-Henderson, Nevada; the New Jersey portion of New York-Newark, NY-NJ-CT; Houston, Texas; Laredo, Texas; and McAllen, Texas. See Findings of Fact ¶¶ 119-20. Thus, the following Individual Plaintiffs and members of Organizational Plaintiffs who reside in these urbanized areas will suffer an injury-in-fact if a citizenship question appears on the 2020 Census: Elizabeth Buchanan, Jacob Cunningham, Maegan Ortiz, Lazara Yoelvis Magadan, Catherine Nwosu, Nnabugwu Nwosu, T. Carter Ross, Michael Kagan, Michael Kravitz, Robyn Kravitz, Raj Mukherji, Diana Alexander, Sarah Bryan, Martha Sanchez, Sonia Casarez Shafer, Virginia Garcia, Juanita Valdez-Cox, Joanne Wilson. See id. ¶ 124.

There is also a substantial risk that a loss in funding to these urbanized areas caused by a differential undercount of at least 2 percentage points will negatively affect one or more of the roads, highways, bridges, sidewalks, trails, or other structures in that urbanized area. Such a reduction is almost certain to delay, eliminate, or change the scope of transportation or transportation-alternative projects of the type covered by the STBG Program in that urbanized area. See id. ¶ 124. Furthermore, the Plaintiffs listed above reside in the urbanized areas identified above.

*739These Plaintiffs regularly drive on highways and roads in and around their respective urbanized areas. Thus, there is a substantial risk that reduced funding for the STBG projects in their urbanized areas will cause these Plaintiffs to suffer an injury from the delay, elimination, or change of the projects covered by the STBG funding.

Additionally, a differential undercount of Hispanics and noncitizens of at least 2 percentage points will cause the following states to lose federal Medicaid funding following the 2020 Census: Arizona, Florida, Nevada, New Mexico, and Texas. See Findings of Fact ¶¶ 125-30. There is a substantial risk or strong likelihood that a reduction in federal Medicaid funds will negatively affect Medicaid beneficiaries' health care benefits and/or medical costs, such as reduction of benefits, services, or access to care, as well as increased costs, for those beneficiaries. Id. ¶ 129.

Because it is highly likely that the citizenship question will cause a differential undercount of Hispanics and noncitizens of at least 2 percentage points, Plaintiffs Sonia Casarez Shafer and Sarah Bryan, who reside in Texas and have children who receive health insurance under the Texas Medicaid program, will suffer an injury-in-fact due to the addition of a citizenship question to the 2020 Census. See id. In addition, members of LUPE and PAZ who reside in Arizona and Texas and receive health insurance under the Arizona and Texas Medicaid programs will suffer an injury-in-fact due to the addition of a citizenship question to the 2020 Census. See id.

It is also substantially likely that a differential undercount of Hispanics and noncitizens of at least 2 percentage points will cause a loss in Title I education funding for numerous school districts in which individual Plaintiffs reside and/or in which their children attend school, including the following districts in Arizona, Nevada, and Texas: Alhambra Elementary District, Balsz Elementary District, Isaac Elementary District, Roosevelt Elementary District, Somerton Elementary District, l Union Elementary District, Clark County School District, Edinburg Consolidated Independent School District, Hidalgo Independent School District, Lyford Consolidated Independent School District, Pharr-San Juan-Alamo Independent School District, and Rio Grande City Consolidated Independent School District. See Findings of Fact ¶ 134. Plaintiffs residing in these school districts will thus suffer an injury-in-fact. Moreover, there is a substantial risk that a loss in Title I education funding to a school district will reduce educational services available to students in the school district, including the children of individual Plaintiffs.

For other programs with allocation formulas based on a state's population or per capita income relative to the United States, there is also a substantial risk that a differential undercount of noncitizens would lead to measurable fiscal loss for states with percentages of noncitizens above the nationwide average. See id. ¶ 138. Plaintiff Juanita Valdez Cox and members of Plaintiff LUPE reside in Texas-a state with a percentage of Hispanics and noncitizens above the national average-and rely on federally funded programs that use Census data, such as Medicaid, the State Children's Health Insurance Program (CHIP), Supplemental Nutrition Program for Women, Infants, and Children (WIC), and Social Services Block Grants (SSBG). See id. ¶¶ 137, 139. Thus, these Plaintiffs will suffer injury from a differential undercount of noncitizens.

*740Notably, even were the Court to rely only on Defendants' expert's extremely conservative undercount assumptions, funding losses would occur in some areas in which certain Plaintiffs reside. Findings of Fact ¶ 140. And a "loss of even a small amount of money is ordinarily" considered an injury for standing purposes. Czyzewski v. Jevic Holding Corp. , --- U.S. ----, 137 S.Ct. 973, 983, 197 L.Ed.2d 398 (2017).

2. Causation and Redressability of Individual Plaintiffs' and Member Plaintiffs' Injuries

To prove a causal connection between threatened injuries and the conduct challenged in a lawsuit, "what matters is not the length of the chain of causation, but rather the plausibility of the links that comprise the chain." See e.g., Mendia v. Garcia , 768 F.3d 1009, 1012-13 (9th Cir. 2014). For standing purposes, Plaintiffs may prove causation using qualitative data. Cf. Whitfield v. S. Maryland Hosp. Inc. , No. DKC 12-2749, 2014 WL 923255, at *7 (D. Md. Mar. 7, 2014) ("A literature review can be an appropriate part of a method of determining causation.").

Plaintiffs have proven a "causal connection" between their injuries and the conduct they challenge in this lawsuit. Lujan , 504 U.S. at 560, 112 S.Ct. 2130. As explained in detail in the Findings of Fact, the evidence establishes that the addition of a citizenship question will cause a differential undercount of Hispanics and noncitizens of at least 2 percentage points that will, in turn, lead to the individual injuries set forth above; these injuries are fairly traceable to the challenged conduct. Plaintiffs have therefore proven each step in the causal chain between their injury and Defendants' conduct.

Defendants rehash various arguments about how third-party decisions and the macro-environment will contribute to or prevent Plaintiffs' injuries, see e.g. , ECF No. 150 ¶¶ 307, 309, 320, 323 -arguments that this Court rejected at the motion to dismiss and summary judgment stages. For example, Defendants argue that Plaintiffs' vote dilution injury may be prevented if states that ordinarily rely on Census data to draw congressional and legislative districts choose to use some other source of population data. Id. ¶ 320. But there is no evidence in the record that the relevant states will stray from their use of Census data to draw voting districts. To the contrary, Plaintiffs' experts consistently opined that it was highly unlikely that Plaintiffs' injuries would be remedied by third-party conduct, even if that conduct was in the realm of lawful possibilities. E.g. , Findings of Fact ¶¶ 119-20, 125, 127.

Defendants' particular argument that states may choose to use other population data for redistricting and apportionment fails for another reason: the Supreme Court has directly contradicted Defendants' position in Department of Commerce v. U.S. House of Representatives. Defendants even acknowledge "that the Supreme Court has found intrastate vote dilution to be fairly traceable to Census Bureau decisions about how to conduct the decennial census by virtue of the fact that some states require the use of federal decennial census population numbers for their state legislative redistricting." Id. ¶ 320 n. 15 (citing Dep't of Commerce , 525 U.S. at 332-34, 119 S.Ct. 765, 142 L.Ed.2d 797 (1999) ). There, the Supreme Court was not concerned with the fact that nothing compels states to continue using Decennial Census data for redistricting and did not find that states' ability to use other data had an effect on the plaintiffs' ability to prove standing's traceability prong. While Defendants note that this finding in Department of Commerce may be dicta, id. , they provide the Court with no good *741cause to stray from the Supreme Court's reasoning.

Ultimately, Defendants' hypothetical that the law will change or that a third-party will swoop in to mitigate Plaintiffs' injuries fails because it simply proves too much. The fact that a third party may prevent or compensate a plaintiff for his injury is true in every situation where there is an allegation of future injury. But the Supreme Court has made clear that Article III is concerned with the risk of future injury, rather than its ultimate realization, and that the risk of future injury may satisfy Article III's injury and causation requirements even if steps on the causal chain still stand between a defendant's conduct and the plaintiff's injury when the case is filed. Davis v. Fed. Election Comm'n , 554 U.S. 724, 734, 128 S.Ct. 2759, 171 L.Ed.2d 737 (2008).¹⁸

3. Organizational Plaintiffs' Standing

A plaintiff organization has representational standing and can sue on behalf of its members if it shows that: "(a) its members would otherwise have standing to sue in their own right; (b) the interests it seeks to protect are germane to the organization's purpose; and (c) neither the claim asserted nor the relief requested requires the participation of individual members in the lawsuit." Hunt v. Wash. State Apple Advert. Comm'n , 432 U.S. 333, 343, 97 S.Ct. 2434, 53 L.Ed.2d 383 (1977). Plaintiff LUPE is among the Organizational Plaintiffs with members that satisfies this standard. First, for the reasons discussed above, LUPE's members, including Juanita Valdez-Cox who testified at trial, would otherwise have standing to sue in their own right, meaning Hunt's first prong is met.

Defendants do not even dispute Hunt's second prong. In any case, an interest is "germane" to an organization's purpose if the lawsuit would "reasonably tend to further the general interests that individual members sought to vindicate in joining the association and ... bears a reasonable connection to the association's knowledge and experience." Bldg. & Constr. Trades Council of Buffalo, N.Y. & Vicinity v. Downtown Dev., Inc. , 448 F.3d 138, 149 (2d Cir. 2006). Here, LUPE's mission involves obtaining government benefits for its community, including ensuring a fair and accurate Census count of that community. See, e.g. , Trial Tr. (Jan. 22) at 150 (Valdez-Cox).

As for the third prong, which Defendants also do not contest, concerns of "administrative convenience and efficiency" favor associational standing, as neither the claims asserted nor the relief requested in this litigation calls for significant participation by individual members; at most, the claims call for proof of their residence, but that can be established without direct participation. United Food and Commercial Workers Union Local 751 v. Brown Grp., Inc. , 517 U.S. 544, 556-57, 116 S.Ct. 1529, 134 L.Ed.2d 758 (1996).

Organizational Plaintiffs can also establish standing if they can demonstrate that the addition of a citizenship question will cause "(1) frustration of [their] organizational mission[s]; and (2) diversion of [their] resources" to mitigate the effects of the challenged action. Smith v. Pac. Props. & Dev. Corp. , 358 F.3d 1097, 1105 (9th Cir. 2004) ; see Havens Realty Corp. , 455 U.S. at 379, 102 S.Ct. 1114 ;

*742White Tail Park, Inc. v. Stroube , 413 F.3d 451, 458 (4th Cir. 2005) (citing Warth v. Seldin , 422 U.S. 490, 511, 95 S.Ct. 2197, 45 L.Ed.2d 343 (1975) ). The Organizational Plaintiffs have succeeded on this basis as well.

Several Organizational Plaintiffs have already begun or will imminently be forced to divert resources to help mitigate the impact of the citizenship question by encouraging the communities that they serve to participate and making sure they understand that being included in the final Census count is critical. See, e.g. , ECF No. 125, Trial Tr. (Jan. 22) at 150-152, 168, 173-179; 220-221 (Valdez-Cox & Park). For example, Ms. Valdez-Cox testified that due to the citizenship question, LUPE has begun its Census outreach and advocacy work earlier than it would otherwise and has already diverted resources from its other core programs. Findings of Fact ¶¶ 143-44. Likewise, the Executive Director of MinKwon, John Park, also testified that the addition of the citizenship question has already resulted in a diversion of resources. Id. ¶ 146. Moreover, the Trial Record shows that the Census Bureau will rely more heavily on its "trusted voices" program, which includes some of the Organizational Plaintiffs, because a citizenship question will make traditionally hard-to-count populations even more reluctant to respond to the Census. Id. Given that the Census Bureau expressly intends to rely on such efforts by community organizations as a strategy for ameliorating the reduced willingness to participate in the Census that the citizenship question will cause, this diversion of resources is fairly traceable to a citizenship question. See id .

Further, several Organizational Plaintiffs have proven that they will suffer an injury-in-fact from the degradation of data quality that would occur if the citizenship question appears on the 2020 Census. Regardless of how successful NRFU operations are in remedying a differential undercount due to a differential decline in self-response rates, the addition of the citizenship question will result in harm to the quality of Census data. Id. ¶ 142. Several Organizational Plaintiffs-including MinKwon-rely on Decennial Census data for the purposes of strategic planning and communication, resource allocation, and advocacy. Id. The degradation in quality of the data would harm these Organizational Plaintiffs' ability to effectively conduct their operations. This injury is fairly traceable to a citizenship question because even if the Census Bureau's NRFU operations are able to prevent any differential undercount, it is undisputed that the data would be of poorer quality and less accurate overall.

Finally, these injuries are "likely" to "be redressed by a favorable decision." Lujan , 504 U.S. at 561, 112 S.Ct. 2130 (internal quotation marks omitted). A favorable decision here would reduce the funds that Plaintiffs would need to expend on Census outreach and avoid the data quality problems that would undermine Plaintiffs' reliance on Census data.

For the foregoing reasons, Plaintiffs have established that they have standing to assert their claims.

B. APA Claims

The APA "sets forth the procedures by which federal agencies are accountable to the public and their actions subject to review by the courts." Franklin v. Massachusetts , 505 U.S. 788, 796, 112 S.Ct. 2767, 120 L.Ed.2d 636 (1992). "A person suffering legal wrong because of agency action" is thus "entitled to judicial review thereof." 5 U.S.C. § 702. Section 706(2) provides that, in a suit challenging agency action, "[t]he reviewing court shall ... hold unlawful and set aside agency action, findings, and conclusions found to *743be," among other things, "arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law; contrary to constitutional right, power, privilege, or immunity; in excess of statutory jurisdiction, authority, or limitations, or short of statutory right; [or] without observance of procedure required by law." 5 U.S.C. §§ 706(2)(A)-(D). Section 706(2) review is to be "thorough, probing, [and] in-depth." Citizens to Preserve Overton Park, Inc. v. Volpe , 401 U.S. 402, 415, 91 S.Ct. 814, 28 L.Ed.2d 136 (1971) ; see id. at 416, 91 S.Ct. 814 ("searching and careful" review).

In evaluating agency action for compliance with the APA, "the focal point for judicial review should be the Administrative Record already in existence, not some new record made initially in the reviewing court." Camp v. Pitts , 411 U.S. 138, 142, 93 S.Ct. 1241, 36 L.Ed.2d 106 (1973) (per curiam). Ordinarily, therefore, courts reviewing agency action for compliance with § 706(2)(A)"confine their review to" the Administrative Record. James Madison Ltd. v. Ludwig , 82 F.3d 1085, 1095 (D.C. Cir. 1996) (internal citation omitted); accord Fort Sumter Tours, Inc. v. Babbitt , 66 F.3d 1324, 1335 (4th Cir. 1995).

Courts have held that the agency's designation of an Administrative Record, "like any established administrative procedure, is entitled to a presumption of administrative regularity." Bar MK Ranches v. Yuetter , 994 F.2d 735, 740 (10th Cir. 1993) ; see also Cent. Elec. Power Co-op, Inc. v. Southeastern Power Admin. , 338 F.3d 333, 337 (4th Cir. 2003). However, "[w]hile review of an agency decision is usually confined to [the Administrative] Record, 'there may be circumstances to justify expanding the record or permitting discovery.' " Nat'l Audubon Soc'y v. Dep't of Navy , 422 F.3d 174, 188 n.4 (4th Cir. 2005) (quoting Fort Sumter Tours, Inc. , 66 F.3d at 1336 ). Relevant here, these circumstances include where there has been a "strong showing" that the agency has acted in "bad faith" in constructing the record or making the agency decision. Nat'l Audubon Soc'y v. Hoffman , 132 F.3d 7, 14 (2d Cir. 1997) (citing Overton Park , 401 U.S. at 420, 91 S.Ct. 814 ).

This Court previously concluded that Plaintiffs had made a strong preliminary showing of bad faith or pretext sufficient to warrant discovery beyond the Administrative Record initially produced-and thereafter materially expanded at the direction of the court in the New York Cases-by Defendants. ECF No. 48 at 33. The Court is now able to conclude that its previous threshold finding of bad-faith has matured into a factual finding of bad faith or pretext. See § II.A-B. However, for the sake of allowing efficient resolution of Plaintiffs claims, the Court will focus on those conclusions of law that are based solely on the Administrative Record and will only point to supplemental Trial Record evidence in this section's footnotes.

"[T]he focal point" of the Court's review, therefore remains, "the [Administrative Record] already in existence, not some new record made initially in the reviewing court." Camp , 411 U.S. at 142, 93 S.Ct. 1241. Indeed, this case presents the unusual instance in which the Administrative Record alone provides more than sufficient evidence to demonstrate not only the invalidity of the Secretary's announced decision on the conventional grounds set forth in APA § 706(2) but also the pretextual nature of the Secretary's stated reasons in his March 26 Memorandum announcing the decision.

1. The Decision to Add the Citizenship Question Was Arbitrary and Capricious and Must Be Set Aside Under APA § 706(2)(A)-(D).

Agency action is arbitrary and capricious "if the agency has relied on factors *744which Congress has not intended it to consider, entirely failed to consider an important aspect of the problem, offered an explanation for its decision that runs counter to the evidence before the agency, or is so implausible that it could not be ascribed to a difference in view or the product of agency expertise." Motor Vehicle Manufacturers Association v. State Farm, 463 U.S. 29, 43, 103 S.Ct. 2856, 77 L.Ed.2d 443 (1983).

The decision to add a citizenship question to the 2020 Census ran "counter to the evidence before the agency," and was not based on facts, a "difference in view," or "agency expertise." State Farm, 463 U.S. at 43, 103 S.Ct. 2856.

a. Factual Assertions in the Ross Memo are Contradicted by the Administrative Record

Here, critical factual assertions in the Ross Memo directly contradicted or ignored the evidence in the Administrative Record, including the Census Bureau's expert memoranda prepared specifically for the Secretary's review, and were not supported by any evidence to the contrary found in the Administrative Record. For instance:

• The Ross Memo claimed that "no one provided evidence that reinstating a citizenship question on the decennial census would materially decrease response rates." PX-26 at 5 (AR 1317).

This assertion is counter to the uncontroverted evidence in the Administrative Record. See e.g. , PX-22 at 1, 2 (AR 1277, 1278); PX-102 at 6-7 (AR 5505-06); PX-136 (AR 10386); PX-147 (AR 11634). In reality, the Administrative Record contains substantial evidence detailing how the addition of a citizenship question to the Decennial Census would reduce response rates among noncitizen households. The Census Bureau conducted several natural experiments to assess the effect of a citizenship question on response rates. See PX-22 at 4 (AR 1280).¹⁹ The Census Bureau examined differential declines in self-response rates, and differential sensitivity to citizenship inquiries as shown in item non-response rates and breakoff rates. See PX-22 at 4 (AR 1280); PX-1194 at 888:7-12; PX-102 at 6-7 (AR 5505-06). On this basis, the Census Bureau calculated that adding a citizenship question to the 2020 census would lead to a 5.1 percentage point decrease in self-response rates. PX-22 at 4-5 (AR 1280-81); PX-100 at 2 (AR 5473); PX-147 (AR 11634).²⁰

The scientific analyses done by the Census Bureau are the only quantitative evidence in the Administrative Record that address a citizenship question's effect on self response. The Administrative Record contains only evidence that the citizenship *745question will harm response rates and no evidence to the contrary. The qualitative evidence in the Administrative Record also reinforces the Census Bureau experts' well-researched finding that adding a citizenship question to the 2020 Census will cause a material decline in self-response rates, particularly among noncitizen households. PX-147 (AR 11634); PX-100 (AR 5473); PX-22 at 4, 5 (AR 1280, 1281); Findings of Fact § I.A.3-4.²¹

Although the Ross Memo refers to studies that purport to contradict the evidence recommending against the addition of a citizenship question, no such evidence appears anywhere in the Administrative Record. See PX-1 to PX-14 (AR). For example, the Ross Memo claimed that an unnamed "SVP of Data Science at Nielsen"-a private-sector media rating company-had advised that, "[w]hen Nielsen added questions on place of birth and time or arrival in the United States (both of which were taken from the ACS) to a short survey, the response rate was not materially different." PX-26 at 6 (AR 1318). There is no documentary support in the Administrative Record for this supposed survey result-one that, even if true, would have no more than anecdotal interest as it lacks any colorable empirical application to the Census.²²

• The Ross Memo contends that Alternative D²³ "would maximize the Census Bureau's ability to match the decennial census responses with [administrative records]," PX-26 at 4 (AR 1316), so as to allow for "more complete" citizenship data.

To the contrary, every scientific analysis in the Administrative Record confirms that the addition of the citizenship question will result in less accurate and less complete citizenship data. Findings of Fact § I.A.3-4. The Administrative Record reflects that because adding a citizenship question would drive down the self-response rate and put more households into NRFU operations, Alternative D actually reduces the Census Bureau's ability to match survey responses with administrative records. PX-25 *746at 4 (AR 1311). As the Census Bureau itself made clear, when more households are pushed into NRFU operations, as will happen if a citizenship question is included, NRFU in turn produces "lower quality" personal identifying information. See id. With reduced quality data, "the number of persons who cannot be linked to" administrative data increases. Id. There is no evidence in the record to the contrary.

In their proposed Findings of Fact and Conclusions of Law, Defendants offer a post-hoc explanation for what the Secretary supposedly meant when he said that a citizenship question would result in more accurate and complete citizenship data. ECF No. 150 ¶ 393-403. Specifically, they claim that the reason the Secretary suggested that a citizenship question would lead to more complete and accurate data-when all the evidence points to the contrary-is that the Secretary was more concerned with the prediction error associated with modelling responses under Alternative C than the combination of lower self response, response errors, and prediction errors that would occur under Alternative D. As a threshold matter, it is a "foundational principle of administrative law that a court may uphold agency action only on the grounds that the agency invoked when it took the action," Michigan v. EPA , --- U.S. ----, 135 S.Ct. 2699, 2710, 192 L.Ed.2d 674 (2015), meaning the post-hoc explanations offered by Defendants and not described in the Ross Memo "cannot serve as a sufficient predicate" for the Secretary's decision, Am. Textile Mfrs. Inst., Inc. v. Donovan , 452 U.S. 490, 539, 101 S.Ct. 2478, 69 L.Ed.2d 185 (1981).

In any case, Defendants' post-hoc explanation is still contradicted by the Administrative Record and lacking a foundation in the facts. After all, the Ross Memo itself acknowledges that survey data regarding citizenship are inaccurate; noncitizens respond to inquiries into their citizenship status by responding that they are citizens approximately 30% of the time. PX-26 at 4 (AR 1316); PX-22 at 7 (AR 1283); PX-25 at 1 (AR 1311). While the Ross Memo correctly asserts that Alternative C-using administrative records "alone"-would require the Census Bureau to "impute" the citizenship of a portion of the population, see PX-26 at 4 (AR 1316), Alternative D would rely on imputation as well, see PX-24 at 2, 4 (AR 1305, 1307). The difference is that in Alternative C, the missing citizenship data would be imputed from a more accurate source (administrative records) than in Alternative D (proxy responses). Id. To the extent that the Secretary meant that a citizenship question would yield more complete data because it would give "100 percent of the population ... the opportunity to provide an answer," this conclusion includes a significant logical leap. What matters is not whether people have the opportunity to respond to a citizenship question, but whether they take it. And the Administrative Record shows that all the experts believed a citizenship question would drive down participation. Findings of Fact ¶¶ 39-43.²⁴ Moreover, any superficial gain in "completeness" derived from potentially "eliminat[ing] the need for the Census Bureau to have to impute an answer for millions of people," PX-26 at 5 (AR 1317), would come at the expense of accuracy, for the reasons explained at Findings of Fact ¶¶ 27, 30, 41, 62, 95, 141.

The Ross Memo contains other flawed assertions that either have no factual basis or are directly contradicted by the Administrative Record, and that, taken together *747with the unsupported assertions discussed above, render the Secretary's decision arbitrary and capricious. For example,

• The Ross Memo asserted that it is "difficult to assess" whether increased NRFU "resulting from inclusion of the citizenship question would lead to increased costs." PX-26 at 7 (AR 1319).

But the Administrative Record reflects that the Census Bureau estimated that the expected self-response decline from adding a citizenship question would increase NRFU costs by at least $ 27.5 million. PX-22 at 6 (AR 1282).

• The Ross Memo also asserts that placing the citizenship question "last" on the Census will "minimize any impact." PX-26 at 8 (AR 1320).

Neither the Ross Memo nor the Administrative Record provides any evidence to support this hypothesis. In fact, that record indicates that the Secretary gave no consideration whatsoever to the specific wording of the proposed citizenship question or the impact of overall questionnaire design, including the placement of the citizenship question within the Census questionnaire. Indeed, since the Census questionnaire requires the respondent to provide information sequentially concerning more than one member of a household, it is not clear what placing the question "last" means, or how the Secretary may have understood this unexplained instruction. In any event, the supposition that this directive would minimize adverse impact is based on rank speculation unsupported by any testing or other expert analysis.

Ultimately, this is not a case where the Administrative Record shows that "specialists express conflicting views," and the agency decision-maker made a rational choice between those views. See Nat'l Audubon Soc'y, Inc. v. U.S. Fish & Wildlife Serv. , 55 F.Supp.3d 316, 351 (E.D.N.Y. 2014) (quoting Marsh v. Oregon Natural Res. Council, 490 U.S. 360, 378, 109 S.Ct. 1851, 104 L.Ed.2d 377 (1989) ). The Secretary's decision did not simply "overrule[ ] the views of some of his subordinates" Wisconsin v. City of New York, 517 U.S. 1, 23, 116 S.Ct. 1091, 134 L.Ed.2d 167 (1996) (emphasis added). To the contrary, the evidence in the Administrative Record uniformly confirms that adding a citizenship question will result in lower quality, more costly data. Yet, counter to the evidence in the record and without explanation for the departure from the facts, the Secretary decided to add a citizenship question to the 2020 Census.

b. The Ross Memo Fails to Follow Applicable Guidance and Regulations or to Consider Appropriate Factors

• The Ross Memo asserts that the citizenship question has been "well tested." PX-26 at 2 (AR 1314).

OMB's Statistical Policy Directive No. 2 ("Standards and Guidelines for Statistical Surveys") requires the Census Bureau, inter alia , to pretest the survey components, if they have not been successfully used before, to "ensure that all components of a survey function as intended when implemented in the full-scale survey" and that "measurement error is controlled." PX-359; see also 71 Fed. Reg. 55,522 (Sept. 22, 2006). The Census Bureau's own Statistical Quality Standards similarly require that "[d]ata collection instruments and supporting materials must be pretested with respondents to identify problems (e.g. , problems related to content, order/context effects, skip instructions, formatting, navigation, and edits) and then refined, prior to implementation, based on pretesting results." PX-260 at 8. The Census Bureau must pretest questions and questionnaires prior to administering them *748unless the Census Bureau obtains a waiver or uses a question that has "performed adequately in another survey." Id.

Although the Ross Memo states that the citizenship question has been "well tested" because it appears on the ACS, PX-26 at 2 (AR 1314), the Administrative Record shows that in fact the citizenship question does not perform adequately on the ACS. As the Census Bureau reported to Secretary Ross, and as the Ross Memo itself acknowledges, approximately one third of respondents identified as noncitizens by administrative records reported themselves as citizens on the 2016 ACS. PX-22 at 8 (AR 1284); PX-26 at 6 (AR 1318). In light of this finding, the Census Bureau concluded that data on citizenship obtained from the citizenship question was of "suspect quality" and "may not be reliable," particularly for noncitizens. PX-25 at 4-5 (AR 1311-12).²⁵ In his cursory wave-of-the-hand treatment of the obligation to pretest survey materials, the Secretary failed to identify any plausible rationale for deviating from the Bureau's rigorous pretesting standards.

• The Ross Memo also failed to even acknowledge, let alone follow, § 6(c)'s mandate.

Section 6 of the Census Act authorizes the Secretary, "whenever he considers it advisable, [to] call upon any other department, agency, or establishment of the Federal Government ... for information pertinent to the work provided for in [Title 13]." 13 U.S.C. § 6(a). In doing so, however, "[t]o the maximum extent possible and consistent with the kind, timeliness, quality and scope of the statistics required, the Secretary shall acquire and use information available from any source referred to in [ § 6(a) ] instead of conducting direct inquiries ." 13 U.S.C. § 6(c) (emphasis added).²⁶

The Administrative Record here demonstrates that, consistent with § 6(c), it was "possible" to "acquire and use" administrative records from other government agencies that would produce data "consistent with the kind, timeliness, quality and scope of the statistics required," or in this case requested by DOJ, although not "required." 13 U.S.C. § 6(c). The Ross Memo does not discuss whether the alternative method for block level data proposed by the Census Bureau (relying on administrative data from other federal agencies) would not have timely or adequately met DOJ's request. And it does not meaningfully grapple with the Census Bureau's repeated advice that the quality of the citizenship data would be higher from administrative records rather than a citizenship question, due to noncitizens'

*749propensity to report as citizens. See Findings of Fact § I.A.3. Because the Ross Memo fails to consider § 6(c), it is not clear to the Court whether the Secretary was unaware of the requirement or chose to ignore it. It is clear though that he treated the choice between Alternatives C (using administrative records) and D (conducting direct inquires) as entirely within his discretion and acted "as though the choice between them were a matter of complete indifference from the statutory point of view ...." Ohio v. U.S. Dep't of the Interior , 880 F.2d 432, 444 (D.C. Cir. 1989). Moreover, he sought to rebrand the imposition of an additional question on 100 percent of the U.S. population as an "opportunity" rather than a burden. PX-26 at 5 (AR 1317). But the congressional preference for use of administrative records in lieu of "direct inquiries" "precludes" a decision "which totally ignores that preference." 880 F.2d at 444. The failure to acknowledge § 6(c)'s unequivocal requirement is particularly egregious given that " 'the Constitution vests Congress ,' not the Executive, 'with wide discretion over ... the conduct of the census,' and it is only because of Congress's statutory delegations that the Secretary of Commerce has any authority to design and conduct the decennial census at all." New York , 351 F.Supp.3d at 641 (emphasis in original) (quoting Wisconsin , 517 U.S. at 15, 116 S.Ct. 1091 ).

Defendants' post-hoc argument that § 6(c) could not bar the Secretary from asking a citizenship question because such a question has appeared on past Census questionnaires does not persuade the Court. As previously described, it is a "foundational principle of administrative law that a court may uphold agency action only on the grounds that the agency invoked when it took the action." Michigan v. EPA , --- U.S. ----, 135 S.Ct. 2699, 2710, 192 L.Ed.2d 674 (2015). Accordingly, "an agency's action must be upheld, if at all, on the basis articulated by the agency itself." State Farm , 463 U.S. at 50, 103 S.Ct. 2856. Given the Secretary's failure to acknowledge or comply with § 6(c), the Court cannot presume that the Secretary in fact considered § 6(c) but determined that it was not applicable.

Taken together, the Secretary's decision ran counter to the evidence before him and the Secretary also failed to consider appropriate factors in the decision. The Secretary's decision was thus arbitrary and capricious and must be set aside under the APA.

2. Secretary Ross's Stated Rationale was Pretextual in Violation of the APA.

The APA requires a decisionmaker to "disclose the basis of its" decision to "give clear indication that it has exercised the discretion with which Congress has empowered it." Burlington Truck Lines, Inc. v. United States , 371 U.S. 156, 168, 83 S.Ct. 239, 9 L.Ed.2d 207 (1962) (internal citation omitted); accord Fed. Power Comm'n v. Texaco Inc. , 417 U.S. 380, 396, 94 S.Ct. 2315, 41 L.Ed.2d 141 (1974). Absent compliance with this basic requirement, a reviewing court would be unable to measure agency action against the relevant governing standard. Cf. e.g., U.S. Lines, Inc. v. Fed. Mar. Comm'n , 584 F.2d 519, 533 (D.C. Cir. 1978) (observing that the court could not "determine whether the final agency decision reflect[ed] the rational outcome of the agency's consideration of all relevant factors," as required by the APA, because it "ha[d] no idea what factors ... were in fact considered by the agency").

As the Findings of Facts describe in detail, the Administrative Record establishes that the Secretary's articulated reason for adding a citizenship question to the *7502020 Census-responding to DOJ's request-was not his real reason. Findings of Fact § II.A-B. Because the VRA enforcement rationale did not actually motivate the Secretary's decision, the Secretary has failed to "disclose the basis of" his decision in violation of the APA.

Defendants' argument that the Secretary simply had an informal, pre-decisional policy preference for adding a citizenship question but ultimately relied on the VRA rationale is meritless. The Secretary's own statements, along with the emails and documents contained in the Administrative Record, establish that the Secretary was pursuing a citizenship question with urgency long before he had any awareness of the purported VRA-enforcement rationale, which the record shows was manufactured by his staff. PX-55 (AR 2521); PX-19 (AR 763); PX-150 (AR 12541); PX-607 (AR).²⁷ Given the flaws in the VRA-enforcement rationale and the consistent recommendation from the Census Bureau that a citizenship question would not be the best way to achieve DOJ's purported goal, the logical conclusion is that some other goal continued to motivate the Secretary.

There is also no merit to Defendants' argument that as long as the Secretary "actually believed" the stated VRA rationale for the decision, it does not matter if he had undisclosed "additional reasons" for the decision. ECF No. 150 ¶ 456. First, the case upon which Defendant relies for this proposition is easily distinguishable. In Jagers v. Federal Crop Insurance Corporation, plaintiffs had "provide[d] neither evidence nor argument to call the validity of the [agency's] scientific evidence into question," nor any concrete evidence of bad faith or improper deviation from normal procedure in developing that evidence. 758 F.3d 1179, 1185 (10th Cir. 2014). Instead, the plaintiffs had merely alleged, without evidence, that agency personnel allegedly had a "subjective desire" to reach the agency's result. Id. at 1185-86. To the contrary, here the Plaintiffs have substantiated their argument that the VRA rationale was pretextual with evidence, and the Secretary's decision flew in the face of the agency's "scientific evidence."

Mississippi Commission on Environmental Quality v. EPA , 790 F.3d 138 (D.C. Cir. 2015) also does not support Defendants' contention that the Secretary's action must be upheld unless he had "an unalterably closed mind" or was "unwilling or unable to rationally consider arguments." ECF No. 150 ¶ 457. That case addressed the legal standard for disqualifying an agency decision-maker from participation in an agency rulemaking, see 790 F.3d at 183, and Defendants have cited no case and offered no reason for why this Court should apply that standard here. In any event, the facts of this case might actually meet the Mississippi Commission's "unalterably closed mind" standard because the Administrative Record ultimately shows that the citizenship question was the Secretary's predetermined answer to a question that he and his staff solicited.

Defendants' position that this Court should only evaluate whether the evidence assembled in the Administrative Record shows the Secretary "actually believed" the VRA rationale also conveniently ignores the realities of how the record in a case like this one is compiled. As Judge Furman described,

[T]his is not a case in which, because of the nature of the administrative proceedings below (such as agency adjudication or notice-and-comment rulemaking), *751either Secretary Ross or the Department of Commerce compiled an "administrative record" in the course of making his decision. Instead, as is often the case with "informal" agency actions, Secretary Ross amassed some information, consulted it, and made his decision on that basis. Then, only after these lawsuits were filed, the Department of Commerce conducted a search for materials that were "before" the Secretary at the time he made that decision, compiled those materials, and submitted them to the Court as the "Administrative Record."

New York , 351 F.Supp.3d at 631 (emphasis in original). Thus, were the Court to adopt Defendants' approach here, it would be encouraging defendants in a case like this one to sanitize the record so as to make it appear that the decisionmaker "actually believed" the articulated rationale for a particular agency action. The incentive to scrub the record would be particularly strong given that, as Defendants have repeatedly emphasized throughout this litigation, courts have held that the agency's designation of an Administrative Record, "like any established administrative procedure, is entitled to a presumption of administrative regularity." Bar MK Ranches v. Yuetter , 994 F.2d 735, 740 (10th Cir. 1993) ; see also Cent. Elec. Power Co-op, Inc. v. Southeastern Power Admin. , 338 F.3d 333, 337 (4th Cir. 2003).

Because the full Administrative Record demonstrates that the VRA-enforcement rationale-the only reason provided by the Ross Memo-was manufactured to "get [ ] in place" Secretary Ross's "months old request" that a citizenship question be included on the 2020 Census, PX-88 (AR 3710), and the Secretary failed to disclose the true basis for his decision, the decision must be set aside.

C. Enumeration Clause Claims

The Court next moves to Plaintiffs' claim that the addition of a citizenship question violates the Enumeration Clause.²⁸ The U.S. Constitution provides for an "actual Enumeration" of the population once every decade to count "the whole number of persons" in each state. U.S. Const. art. I, § 2, cl. 3, and amen. XIV, § 2. The U.S. Constitution recognizes no exception based on citizenship status; it is long settled that all persons residing in the United States-citizens and noncitizens alike-must be counted to fulfill the U.S. Constitution's "actual Enumeration" mandate. Id. ; Fed'n for Am. Immigration Reform v. Klutznick , 486 F.Supp. 564, 576 (D.D.C. 1980).

The Enumeration Clause requires the Secretary's conduct of the Census to bear "a reasonable relationship to the accomplishment of an actual enumeration of the population, keeping in mind the constitutional purpose of the census." Wisconsin v. City of New York , 517 U.S. 1, 20, 116 S.Ct. 1091, 134 L.Ed.2d 167 (1996) (quoting Franklin v. Massachusetts , 505 U.S. 788, 804, 112 S.Ct. 2767, 120 L.Ed.2d 636 (1992) ). While Defendants are not required to achieve a perfect count, they must not act to unreasonably compromise the distributive accuracy of the Census. Id.

Plaintiffs have established that the addition of a citizenship question to the Census does not bear "a reasonable relationship to the accomplishment of an *752actual enumeration of the population." Wisconsin , 517 U.S. at 19, 116 S.Ct. 1091. Plaintiffs have marshaled substantial evidence, based on the Census Bureau's own data and analyses-including information that was available to the Secretary at the time of the Ross Memo-showing that the citizenship question will lead to a differential undercount that will dilute Plaintiffs' votes. See Findings of Fact §§ I.A.3-4, I.C.3.a. The unreasonableness of Defendants' addition of a citizenship question to the Census is underscored by the lack of any genuine need for the citizenship question, the woefully deficient process that led to it, the mysterious and potentially improper political considerations that motivated the decision and the clear pretext offered to the public. Findings of Fact § I.A.

Specifically, by March 2018, there was compelling evidence presented to Secretary Ross and available to the Defendants showing that the citizenship question would produce a differential undercount that would undermine the constitutional purpose of the Census. Findings of Fact § I.A.3. At the same time, the Bureau's analyses made clear to the Defendants that there was no genuine need for the citizenship data that could justify or outweigh the constitutional harm of undermining the distributive accuracy of the Census count. Id. The Census Bureau had determined that DOJ's request for block-level citizenship data would have been better satisfied through the use of Administrative Records. See id. ¶¶ 30, 39.

Additionally, Administrative Record evidence establishes that the purported rationale of supporting VRA enforcement was a mere pretext to rationalize a decision that had already been made for other reasons. See Findings of Fact § I.A. At best, the Secretary ignored clear evidence that the citizenship question would harm the distributive accuracy of the Census for some mysterious reason known only to him. At worst, the Secretary intended to negatively affect the distributive accuracy of the Census by reducing immigrant response rates to the Census. Id. ¶¶ 6, 14, 16. Both possibilities disregard the need to accomplish an actual enumeration of the population-the constitutional purpose of the Census.

Defendants point out that a citizenship question and other sensitive demographic questions have historically appeared on the Decennial Census. ECF No. 150 ¶ 494-496. However, a citizenship question has not appeared on the short form since 1950, and, as discussed previously, it has not performed well even on the much longer ACS survey where it is significantly less noticeable. Regarding other sensitive demographic questions, there has been no evidence showing that any other demographic question has the potential to disrupt the distributive accuracy of the Census in the way the citizenship question will. For example, as noted in the Court's previous opinion, ECF No. 48 at 25-26, a male in New York is no more or less likely to respond to the Census questionnaire because of a question about gender than a woman in Maryland. Any burden caused by such a question is therefore evenly distributed and does not disturb distributive accuracy as the citizenship question would. Because the Secretary ignored evidence regarding the impact of the question and provided no legitimate rationale to support it, the addition of the citizenship question would unreasonably compromise the distributive accuracy of the Census, and the addition violates the Enumeration Clause.

D. Equal Protection and 42 U.S.C. § 1985 Claims

To establish a violation of the equal protection component of the Due *753Process Clause, Plaintiffs bear the burden of demonstrating that an "invidious discriminatory purpose was a motivating factor" behind the Secretary's decision. Vill. of Arlington Heights v. Metro. Hous. Dev. Corp. (Arlington Heights) , 429 U.S. 252, 266, 97 S.Ct. 555, 50 L.Ed.2d 450 (1977). To make such a showing, "proof of racially discriminatory intent or purpose is required." Id. at 265, 97 S.Ct. 555. Racially discriminatory purpose "may often be inferred from the totality of the relevant facts, including the fact" that the effect of the addition of the citizenship question "bears more heavily on one race than another." Washington v. Davis , 426 U.S. 229, 242, 96 S.Ct. 2040, 48 L.Ed.2d 597 (1976) ; N.C. State Conference of NAACP v. McCrory , 831 F.3d 204, 220 (4th Cir. 2016). However, although a showing of disparate impact on a protected group and the foreseeability of this impact is relevant to prove that the decisionmaker acted with a forbidden purpose, "impact alone is not determinative, and the Court must look to other evidence." Id. at 266, 97 S.Ct. 555. Other evidence demonstrating a defendant's racially discriminatory purpose may include:

(1) evidence of a "consistent pattern" of actions by the decisionmaking body disparately impacting members of a particular class of persons; (2) historical background of the decision, which may take into account any history of discrimination by the decisionmaking body or the jurisdiction it represents; (3) the specific sequence of events leading up to the particular decision being challenged, including any significant departures from normal procedures; and (4) contemporary statements by decisionmakers on the record or in minutes of their meetings.

Sylvia Development v. Calvert County, 48 F.3d 810, 819 (4th Cir.1995). In addition to alleging an equal protection violation, the LUPE Plaintiffs claim that Defendants violated 42 U.S.C. § 1985(3), which provides:

If two or more persons ... conspire ... for the purpose of depriving, either directly or indirectly, any person or class of persons of the equal protection of the laws, or of equal privileges and immunities under the laws ... whereby another is injured in his person or property, or deprived of having and exercising any right or privilege of a citizen of the United States, the party so injured or deprived may have an action for the recovery of damages, occasioned by such injury or deprivation, against any one or more of the conspirators.

42 U.S.C. § 1985(3). To establish a claim under § 1985(3) for conspiracy to deny equal protection of the law, a plaintiff must show: "(1) a conspiracy of two or more persons, (2) who are motivated by a specific class-based, invidiously discriminatory animus to (3) deprive the plaintiff of the equal enjoyment of rights secured by the law to all, (4) and which results in injury to the plaintiff as (5) a consequence of an overt act committed by the defendants in connection with the conspiracy." Simmons v. Poe , 47 F.3d 1370, 1376 (4th Cir. 1995). As the LUPE Plaintiffs acknowledge, their § 1985 claim rises and falls with their Equal Protection claim. ECF No. 151-2 ¶ 288.

Before addressing whether Defendants acted with discriminatory purpose to harm a protected class, the Court pauses to consider the proper scope of its review. Defendants contest the Court's authority to decide Plaintiffs' equal protection and § 1985 claims based on the full Trial Record, which includes evidence outside the Administrative Record. ECF No. 150 ¶¶ 508-509. However, "it would be nearly impossible to smoke out discriminatory purpose if litigants and courts evaluating *754whether government actors have engaged in invidious discrimination cannot look beyond the record that those very decisionmakers may have carefully curated to exclude evidence of their true intent and purpose." New York , 351 F.Supp.3d at 668 (internal quotation marks and citation omitted).

To be sure, courts routinely decide constitutional claims based on the Administrative Record. See ECF No. 150 ¶ 508 (collecting cases). However, in the cases cited by Defendants, plaintiffs had not met their burden to prove a preliminary finding of bad faith entitling them to discovery outside the Administrative Record. See e.g., Chang v. U.S. Citizenship & Immigration Servs. , 254 F.Supp.3d 160, 161 (D.D.C. 2017) ("Plaintiffs do not attempt to argue that they have satisfied the high standard required to supplement the administrative record"); Evans v. Salazar , Case No. C08-0372-JCC, 2010 WL 11565108, at *1 (W.D. Wash. July 7, 2010) (assuming the record was properly designated because plaintiffs had not submitted any evidence to the contrary). Not so here.

In any case, even looking to evidence beyond the Administrative Record, when the Court considers the background of the decision, the process that led to it and relevant contemporary statements, Plaintiffs have not met their burden to prove by a preponderance of the evidence that Defendants acted with racially-motivated discriminatory intent. Outside of demonstrating that a citizenship question will disparately impact Hispanics, Plaintiffs have offered little, if any evidence, showing Secretary Ross harbors animus towards Hispanics or that such animus impacted his decision. E.g. PX-1145. On the other hand, Plaintiffs' have put forward some evidence that certain other individuals, including the President and Secretary Kobach, did harbor discriminatory animus towards noncitizens more broadly and that their desire for a citizenship question may have been motivated by that animus; but Plaintiffs have not sufficiently tied those views to Secretary Ross's decision. In the absence of any other non-pretextual rationale, discriminatory animus may well be the most likely explanation for Secretary Ross's addition of the citizenship question but that is not the same as saying it has been proven by a preponderance of the evidence. All that is clearly proven regarding Secretary Ross is that, seemingly from the beginning of his tenure, he had a strong interest in the citizenship question, and his discussions on the issue included information regarding the impact of counting undocumented immigrants on the Census, as well as other discussions that have not been explored on the record. This is not sufficient for the Court to find by a preponderance of the evidence that the Secretary's decision was made for the purpose of depressing immigrant response and motivated by discriminatory animus. Thus, Plaintiffs' Equal Protection Claim fails. Because Plaintiffs failed to prove that Secretary Ross was motivated by invidious class-based discriminatory animus, they have also inevitably failed to prove their § 1985 claim.²⁹

*755E. Remedies

Based on the conclusion that the Secretary's decision to add a citizenship question to the 2020 Census violated the APA, the Court is required to "hold unlawful and set aside" that decision. 5 U.S.C. § 706(2). Thus, under the APA's plain terms, vacatur and remand is appropriate.

However, Plaintiffs also seek a permeant, nationwide injunction preventing Defendants from placing the citizenship question on the 2020 Census questionnaire regardless of action taken by Secretary Ross on remand to comply with the APA. A plaintiff seeking a permanent injunction must show: "(1) that it has suffered an irreparable injury; (2) that remedies available at law, such as monetary damages, are inadequate to compensate for that injury; (3) that, considering the balance of hardships between the plaintiff and defendant, a remedy in equity is warranted; and (4) that the public interest would not be disserved by a permanent injunction." Monsanto Co. v. Geertson Seed Farms , 561 U.S. 139, 156-57, 130 S.Ct. 2743, 177 L.Ed.2d 461 (2010). Because the government is a party, and "the government's interest is the public interest," the last two factors merge. Pursuing Am. Greatness v. Fed. Election Comm'n , 831 F.3d 500, 511 (D.C. Cir. 2016) ; accord Nken v. Holder , 556 U.S. 418, 435, 129 S.Ct. 1749, 173 L.Ed.2d 550 (2009). These elements are satisfied based on the myriad harms Plaintiffs proved they would suffer if the citizenship question is included on the 2020 Census-harms like loss of federal funding and vote dilution that cannot be remedied through monetary damages but can be avoided through injunctive relief. For the reasons described by Judge Seeborg, an injunction broader than one that solely makes vacatur effective is appropriate given the Enumeration Clause violation. State v. Ross , 358 F.Supp.3d 965, 1050-51 (N.D. Cal. 2019). Specifically, although Secretary Ross may be able to cure the plethora of APA violations, the record reflects that the inclusion of a citizenship question would still undermine the Census's primary constitutional purpose by unreasonably distorting the accuracy of "an actual Enumeration." Finally, as the New York and California Courts concluded, the injunctive relief requested here could not practically be limited to only one geographic area or certain litigants-the citizenship question will either be included or barred from the 2020 Census on a nationwide basis. Ross , 358 F.Supp.3d at 1050-51 ; New York , 351 F.Supp.3d at 678. It will be barred.

*756Although Plaintiffs also request declaratory relief, the vacatur of Secretary Ross's decision and the injunction barring Defendants from including the citizenship question on the 2020 Census, render declaratory relief unnecessary.

III. Conclusion

Based on the foregoing, the Court will enter judgment for the Plaintiffs, enjoining Defendants from including the citizenship question on the 2020 Census, regardless of any technical compliance with the APA. A separate order follows.

2 The Kravitz Individual named Plaintiffs are: Diana Alexander (Houston, Texas); Lauren Rachel Berman (Dallas, Texas); Sarah Bryan (Edinburg, Texas); Alejandro Chavez (Phoenix Arizona); Virginia Garcia (Laredo, Texas); Michael Kagan (Las Vegas, Nevada); Robyn Kravitz (District Heights, Maryland); Michael Kravitz (District Heights, Maryland); Yamile Labori (West Palm Beach, Florida); Lazara Yoelvis Magadan (Miami, Florida); Richard McCune (Nogales, Arizona); Jose Moreno (Somerton, Arizona); Catherine Nwosu (Langley Park, Maryland); Nnabugwu Nwosu (Langley Park, Maryland); Linda Rivas (El Paso, Texas); T. Carter Ross (Hyattsville, Maryland); Martha Sanchez (McAllen, Texas); Sonia Casarez Shafer (Pharr, Texas); and Joanne Wilson (Bowie, Maryland). The LUPE Individual named Plaintiffs are: Juanita Valdez-Cox (Texas), Gene Wu (Texas), Mia Gregerson (Washington), Cindy Ryu (Washington), Sharon Tomiko Santos (Washington), Raj Mukherji (New Jersey), OJ Semans (South Dakota). The LUPE Organizational named Plaintiffs are: La Unión del Pueblo Entero (LUPE), Texas Senate Hispanic Caucus, Texas House of Representatives Mexican American Legislative Caucus, Southwest Voter Registration Education Project, California Latino Legislative Caucus, Coalition for Humane Immigrant Rights, Dolores Huerta Foundation, Mi Familia Vota Education Fund, Somos Un Pueblo Unido, Georgia Association of Latino Elected Officials, Labor Council for Latin American Advancement, Promise Arizona, El Pueblo, Inc., Maryland Legislative Latino Caucus, Asian Americans Advancing Justice-Chicago, Asia Services in Action, Inc., Minkwon Center for Community Action Inc., Chelsea Collaborative, Chicanos por La Causa, Latino Community Fund of Washington, Arizona Legislative Caucus, California Asian Pacific Islander, Legislative Caucus, California Legislative Black Caucus, Oca-Greater Houston, Friendly House, Four Directions, Inc.

3 In a prior opinion, the Court referred to this as the "Census Clause," but will now adopt the terminology used by the parties in their briefs.

4 Unless otherwise noted, all citations to the docket are to No. 18-CV-1041.

5 Where no specific Administrative Record citation is provided but Plaintiffs' exhibit number is followed by the parenthetical "(AR)," the parties have stipulated that the exhibit is included in the Administrative Record and admitted into evidence. ECF No. 134.

6 That this July 25, 2017 call took place can be inferred from the Administrative Record cite, but it is confirmed by the Trial Record, PX-193 at 8, 40 (COM_DIS00021166, COM_DIS000211198).

7 The Memorandum was not produced in this case based on an assertion of attorney-client privilege.

8 While the Court can reasonably infer from the email's inclusion in the Administrative Record that the "DOJ-DOC" issue referred to in Gore's email to Teramoto, PX-59 (AR 2628) & PX-60 (AR 2634), was the addition of a citizenship question to the 2020 Decennial Census, the Trial Record confirms this fact, Gore Dep. at 96:17-97:2.

9 The Trial Record confirms that any discussions between the Department of Commerce and DOJ about a citizenship question addition were initiated by the Department of Commerce. Gore Dep. at 67-68.

10 Comstock testified that there were "[t]wo or three" meetings with representatives of the Census Bureau, the last of which was "somewhere in the vicinity of March 20th." Comstock Dep. 322-23. However, like Judge Furman, this Court declines to credit this testimony. As Judge Furman noted: "There is no evidence in the record of any such meetings. Further, Dr. Abowd testified that the February 12, 2018 meeting (which Comstock attended) was his one and only meeting with Secretary Ross, and that he could not recall a meeting with Comstock after March 1, 2018. Tr. 884, 991-92." New York v. U.S. Dep't of Commerce , No. 1:18-cv-02921-JMF, ECF No. 574 at 92 n. 27 (S.D.N.Y.).

11 Outside of showing that a citizenship question is likely to disparately impact Hispanics, Plaintiffs have not provided any evidence that Secretary Ross was motivated by animus towards Hispanics/Latinos.

12 See also PX-22 at 5-6; PX-69; PX-162 at 10; PX-1194 at 914:5-8, 916:1-918:9, 926:21-927:10, 930:9-24, 933:23-934:7, 938:22-939:17, 940:04-941:14; Census Bureau 30(b)(6) Dep. Vol. II at 361:6-363:4; ECF No. 138, Trial Tr. (Jan. 30) at 131:5-9 (Abowd); PX-152 at 22; Census Bureau 30(b)(6) Dep. Vol. II at 437:6-438:16, 449:5-13, 449:21-451:9; ECF No. 126, Trial Tr. (Jan. 23) at 106:10-107:11, 110:4-111:1 (Mathiowetz).

13 See also PX-25 (households that refuse to self-respond to the Census questionnaire because of the presence of a citizenship question are "particularly likely to refuse to respond in NRFU as well."); PX-162 at 41 ("[h]ouseholds deciding not to self-respond because of the citizenship question are likely to refuse to cooperate with enumerators coming to their door."); ECF No. 133-4, Navarrete Decl. ¶ 11 (community organizations are credibly concerned that the Latino and immigrant communities "will refuse to participate in the NRFU process out of fear of opening their doors to government workers.").

14 See also ECF No. 126, Tr. (Jan. 23) at 155:18-156:18 (Mathiowetz); ECF No. 133-3, Salas Decl. ¶ 10; PX-158 at 2.

15 The Court notes the same weaknesses with this step as discussed above.

16 See also Cal. Const. art. 21, § 1 ; Nev. Const. art. 15, § 13 ; Fla. Stat. § 11.031(1) ; Md. Code Ann., Elec. Law § 8-701 ; N.J. Const. art. 4, § 2 ¶¶ 1, 3 ; Tex. Const. art. 3, § 26 ; Georgia v. Ashcroft , 539 U.S. 461, 488 n.2, 123 S.Ct. 2498, 156 L.Ed.2d 428 (2003) ("When the decennial census numbers are released, states must redistrict to account for any changes or shifts in population."); Dep't of Commerce v. U.S. House of Representatives , 525 U.S. 316, 334, 119 S.Ct. 765, 142 L.Ed.2d 797 (1999) ("States use the population numbers generated by the federal decennial census for federal congressional redistricting.").

17 Administrative records are data collected by federal or state agencies while they are administering their programs and services, which may include information about an individual's residence and citizenship status. See Jennifer Ortman, Administrative Records Offset Declining Census Survey Response , U.S. Census Bureau (Nov. 20, 2018), https://www.census.gov/library/stories/2018/11/administrative-records-offset-declining-census-survey-response.html.

18 As to the redressability of Plaintiffs' claims, Defendants do not even dispute that Plaintiffs' injuries will be redressed if Defendants are enjoined from including a citizenship question on the 2020 Census.

19 See also PX-1194 at 887:16-20.

20 Extra-record evidence presented at trial-including the testimony of Plaintiffs' experts and of Defendants' witness, Dr. Abowd-also demonstrates that adding a citizenship question will have material adverse effects on self-response rates, particularly those of immigrant and Hispanic households. See, e.g. , PX-162 at 39 (revising "conservative" estimate of decline in self-response to 5.8 percentage points); PX-297 at RFA 96-97 (the Census Bureau found empirical evidence that a citizenship question could reduce response rates); PX-662 (updated CBAMS presentation); PX-1194 at 881-82, 897, 919-20; Trial Tr. (Jan. 23) at 126:13-127:10 (Mathiowetz); PX-696 at 621:21-622:8; 615:20-621:15; 644:1-7;; PX-670; PX-671. Although this additional evidence is not needed to establish that the evidence in the Administrative Record on self-response rates ran directly counter to the Secretary's claims in the Ross Memo, it serves to confirm the material significance of the Secretary's failure in negating any reasonable basis for the decision.

21 The Trial Record confirms that the Census Bureau's initial analysis predicting a 5.1 percentage point decline in self response as the result of a citizenship question represented a conservative estimate. PX-1194 at 893:3-4. With the benefit of additional research and analysis, Census Bureau researchers have concluded that the expected decline in self-response due to the citizenship question is higher-at least 5.8 percentage points, although the Census Bureau maintained that even this figure is conservative. PX-162 at 38-39. The Brown et al. paper containing the 5.8 percentage point prediction has been subjected to rigorous internal review by the Census Bureau and has been submitted to a peer-reviewed journal for publication. See Trial Tr. (Jan. 30) at 64:7-15, 68:12-16, 129:12-14 (Abowd).

22 Extra-record evidence casts even more doubt upon the Secretary's assertions about his conversation with a Nielsen executive. Christine Pierce, the Nielsen Company Vice President of Data Science referred to in the Ross Memo, testified through a written affidavit that she expressed her "unequivocal[ ]" concern "that a citizenship question would negatively impact self-response rates," and that the Secretary misrepresented her statements in the Ross Memo. ECF No. 95-1, Pierce Decl. at ¶¶ 9, 15, 12-18 (admitted for limited purposes per the Court's January 25, 2019 Letter Order, ECF No. 128, as a prior sworn statement of an unavailable witness.). She claims that Secretary Ross's statement that a Nielsen Executive cited evidence that a citizenship question would not cause a decline in self-response rate is false. Id. ¶¶ 9, 12-18.

23 Alternative D refers to the fourth alternative analyzed by the Census Bureau and devised by Secretary Ross, which "combined Alternative B (asking the citizenship question of every household on the 2020 Census) with Alternative C (do not ask the question, link reliable administrative data on citizenship status instead)." PX-132 at 2 (AR 9813).

24 In the Trial Record, Dr. Abowd walked through the flaws in Defendants' post-hoc explanation and explained why the Census Bureau's experts unanimously favored Alternative C over Alternative D. PX-1194 at 978-985.

25 The Trial Record supplements this Administrative Record evidence. In August of 2018, the Census Bureau reported that between 30% and 37% of all people identified as noncitizens by administrative records reported themselves as citizens on the ACS. PX-162. These statistics confirm that the citizenship question has not performed adequately on the ACS-a conclusion with which the Census Bureau's Chief Scientist, Dr. Abowd agreed "without qualification" at trial. See Trial Tr. (Jan. 30) at 166:10-25 (Abowd); see also Trial Tr. (Jan. 23) at 90:8-13, 126:20-25 (Mathiowetz). And there is no basis to expect that the citizenship question would perform better on the Decennial Census. Trial Tr. (Jan. 30) at 163:15-20 (Abowd).

26 The Court is not convinced by Defendants' argument that Plaintiffs' waived allegations related to the Secretary's violation of 13 U.S.C. § 6 by not specifically referencing the statute in their Complaints. Plaintiffs' § 6(c) allegations are encompassed in their allegations that Defendants' decision to add a citizenship question was not in accordance with law, in excess of statutory jurisdiction and authority, and in violation of binding law and regulations regarding statistical collection procedures and methods.

27 The Trial Record also shows that the Secretary concealed his personal efforts to solicit a legitimate rationale for the citizenship question. PX-491.

28 The Administrative Record provides ample evidence of Defendants' Enumeration Clause violation. Thus, for the sake of efficiency, the Court relies only on Administrative Record evidence in this section and does not discuss Defendants' argument that the Court may not look to extra-record evidence when reviewing an Enumeration Clause claim. ECF No. 150 ¶ 490.

29 The Court is not persuaded by Defendants' argument that if Plaintiffs may pursue a claim against federal employees in their official capacity under 42 U.S.C. § 1985(3) then their claims under the APA must be dismissed because they have an adequate alternative remedy. The APA provides that "[a]gency action made reviewable by statute and final agency action for which there is no other adequate remedy in a court are subject to judicial review." 5 U.S.C. § 704. "An alternative remedy will not be adequate under § 704 if the remedy offers only 'doubtful and limited relief.' " Garcia v. Vilsack , 563 F.3d 519, 522 (D.C. Cir. 2009) (quoting Bowen v. Massachusetts , 487 U.S. 879, 901, 108 S.Ct. 2722, 101 L.Ed.2d 749 (1988) ). In support of their position, Defendants cite only one case-Garcia v. Vilsack , 563 F.3d 519, 522 (D.C. Cir. 2009) -which is readily distinguishable from the situation here. In Garcia , Plaintiffs brought an APA claim as an end-run-around their failure to obtain relief under the Equal Credit Opportunity Act (ECOA). Unlike § 1985(3) -a broad civil rights statute with unclear application to Plaintiffs' claims-the ECOA provided a specific remedy for the Garcia plaintiffs' precise injuries. Thus, the D.C. Circuit found that the plaintiffs could not bring an APA claim because they failed to show they lacked an adequate alternative remedy. The court held that allowing the plaintiffs to also bring an APA claim "would effectively rewrite the statute that Congress specifically enacted in response to the USDA's failure to address discrimination complaints" because the statute's "plain text" required complainants to choose between "going to court immediately or first renewing their administrative complaints." Garcia , 563 F.3d at 524. Here, allowing Plaintiffs to proceed with their claims under the APA in no way rewrites § 1985-a statute that was not obviously designed to remedy violations of the Census Act, Census Bureau guidelines, or OMB regulations. Whereas in Garcia , the Plaintiffs had an adequate alternative remedy under a statute specifically designed to remedy their injuries but made strategic missteps about how best to resolve their claims, here Plaintiffs do not.