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
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.
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
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.
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.
3. Between 1970 and 2000, the Census Bureau used both the short-form questionnaire
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.
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."
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);
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."
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."
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."
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).
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)."
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."
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."
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
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."
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."
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
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.
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).
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."
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),
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."
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
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.
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.
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.
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.
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.
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."
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.
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
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.
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
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.
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
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. Statisticiansoften 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.
40. The Census Bureau anticipated needing to model the citizenship status of approximately 35 million people under Alternative C.
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.
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."
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.
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?"
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 requirementsestablished 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.
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
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."
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
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."
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.
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
59. In one of Gore's several calls with the Secretary's advisors,
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.
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
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
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;
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.
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.
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
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.
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 ,
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.
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.
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
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 ,
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).
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
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
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.
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
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 whenwe 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.
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
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
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.
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
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).
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
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
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
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
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).
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
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
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
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. ,
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.
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,
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."
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.
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 ,
The Census Bureau is a principal statistical agency within the federal statistical system.
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 ,
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.
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 ,
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 ,
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.
a. Intrastate Vote Dilution
Turning to the first injury Plaintiffs allege, the Supreme Court has found that plaintiffs challenging a proposed Census
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 ;
b. Apportionment
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 ,
c. Loss of Federal Funding
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 ,
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.
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.
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 ,
Plaintiffs have proven a "causal connection" between their injuries and the conduct they challenge in this lawsuit. Lujan ,
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 ,
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 ,
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 ,
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. ,
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. ,
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. ,
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 ,
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 ,
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 ,
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 ,
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 ,
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
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,
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).
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
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.
• 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
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. ----,
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).
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
• 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. ,
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
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).
• 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]."
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."
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. ----,
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 ,
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
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).
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.
Mississippi Commission on Environmental Quality v. EPA ,
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),either 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 ,
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.
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 ,
Plaintiffs have established that the addition of a citizenship question to the Census does not bear "a reasonable relationship to the accomplishment of an
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.
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
To establish a violation of the equal protection component of the Due
(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,
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.
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
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. ,
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.
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.
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 ,
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.
Notes
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.
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.
Unless otherwise noted, all citations to the docket are to No. 18-CV-1041.
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.
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).
The Memorandum was not produced in this case based on an assertion of attorney-client privilege.
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.
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.
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.).
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.
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).
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.").
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.
The Court notes the same weaknesses with this step as discussed above.
See also Cal. Const. art. 21, § 1 ; Nev. Const. art. 15, § 13 ;
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.
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.
See also PX-1194 at 887:16-20.
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.
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).
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.
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).
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.
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).
The Court is not convinced by Defendants' argument that Plaintiffs' waived allegations related to the Secretary's violation of
The Trial Record also shows that the Secretary concealed his personal efforts to solicit a legitimate rationale for the citizenship question. PX-491.
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.
The Court is not persuaded by Defendants' argument that if Plaintiffs may pursue a claim against federal employees in their official capacity under