No. 18-966
IN THE
Supreme Court of the United States
DEPARTMENT OF COMMERCE, et al.,
Petitioners,
v.
STATE OF NEW YORK, et al.,
Respondents.
O
N
W
RIT OF
C
ERTIORARI
B
EFORE
J
UDGMENT TO THE
UNITED STATES COURT OF APPEALS FOR THE SECOND CIRCUIT
BRIEF FOR GOVERNMENT RESPONDENTS
M
ATTHEW COLANGELO
Chief Counsel for
Federal Initiatives
E
LENA GOLDSTEIN
Acting Bureau Chief
Civil Rights Bureau
LETITIA JAMES
Attorney General
State of New York
B
ARBARA D. UNDERWOOD*
Solicitor General
S
TEVEN C. WU
Deputy Solicitor General
J
UDITH N. VALE
Senior Assistant
Solicitor General
SCOTT A. EISMAN
Assistant Solicitor General
28 Liberty Street
New York, NY 10005
(212) 416-8020
barbara.underwood@ag.ny.gov
*Counsel of Record
(Counsel list continues on signature pages.)
i
QUESTIONS PRESENTED
1. Whether the district court correctly concluded,
on the basis of well-settled principles of administra-
tive law, that the Secretary of Commerce’s decision to
add a citizenship question to the 2020 decennial
census questionnaire was arbitrary and capricious
and contrary to law, in violation of the Administrative
Procedure Act, 5 U.S.C. § 706(2).
2. Whether petitioners’ challenges to the district
court’s authorization of limited discovery beyond the
agency’s proffered administrative record are moot
and, in any event, meritless given that extraordinary
circumstances raised significant doubts about
whether the agency had provided the whole record or
an accurate account of its decision-making.
3. Whether the Secretary of Commerce’s decision
to add a citizenship question to the 2020 decennial
census questionnaire violated the Enumeration
Clause of the U.S. Constitution, art. I, § 2, cl. 3.
ii
TABLE OF CONTENTS
Page
Introduction ................................................................. 1
Statement .................................................................... 2
A. Modernization of the Decennial Census ........... 2
B. The Decision to Add a Citizenship
Question ............................................................. 8
C. Procedural History ........................................... 13
Summary of Argument .............................................. 17
Argument ................................................................... 21
I. Government Respondents Have Standing. ..... 21
II. The Secretary’s Decision Violated the APA. ... 25
A. The Secretary’s Decision Is Reviewable. ... 25
B. The Secretary’s Decision Was Arbitrary
and Capricious. ........................................... 31
1. The Secretary Disregarded Harms
to the Enumeration. .............................. 31
a. Undisputed evidence
demonstrated that a citizenship
question would depress response
rates. ................................................. 31
b. The Secretary’s reliance on a
purported lack of information
was unreasoned when he
abandoned testing procedures. ........ 37
2. The Secretary’s Reliance on DOJ’s
Purported Need for More Accurate
Citizenship Data Was Arbitrary and
Capricious. ............................................. 42
iii
Page
a. The Secretary failed to explain
why a citizenship question was
necessary when administrative
records would satisfy DOJ’s
request. ............................................. 42
b. The Secretary irrationally chose
an approach that would produce
less accurate citizenship
information than using
administrative records alone. .......... 45
3. The Secretary Failed to Explain
Why He Was Prioritizing
Citizenship Data over the
Enumeration. ........................................ 51
4. The Secretary’s Rationale Was
Pretextual............................................... 54
C. The Secretary’s Decision Was Contrary
to Law. ........................................................ 59
III. The Secretary’s Decision Violated the
Enumeration Clause. ....................................... 62
IV. The District Court Properly Authorized
Discovery. ......................................................... 68
Conclusion .................................................................. 69
iv
TABLE OF AUTHORITIES
Cases Page(s)
Alaska Dep’t of Envtl. Conservation v. EPA,
540 U.S. 461 (2004) .............................................. 28
American Wild Horse Pres. Campaign v.
Perdue, 873 F.3d 914 (D.C. Cir. 2017) ............ 39,52
Attias v. CareFirst Inc., 865 F.3d 620 (D.C.
Cir. 2017) .............................................................. 24
Baker v. Carr, 369 U.S. 186 (1962) ........................... 67
Bennett v. Spear, 520 U.S. 154 (1997) ................. 23,24
Biodiversity Legal Found. v. Babbitt, 146 F.3d
1249 (10th Cir. 1998) ........................................... 28
Block v. Meese, 793 F.2d 1303 (D.C. Cir. 1986) ........ 24
Burlington Truck Lines, Inc. v. United States,
371 U.S. 156 (1962) .................................... 51,55,69
Califano v. Yamasaki, 442 U.S. 682 (1979) .............. 62
California v. Ross, No. 18-cv-1865, 2018 WL
7142099 (N.D. Cal. Aug. 17, 2018) ...................... 66
California v. Ross, No. 18-cv-1865, 2019 WL
1052434 (N.D. Cal. 2019) ................................ 16,64
Carey v. Klutznick, 637 F.2d 834 (2d Cir. 1980) ....... 26
Center for Biological Diversity v. EPA, 722
F.3d 401 (D.C. Cir. 2013) ..................................... 52
Citizens to Pres. Overton Park, Inc. v. Volpe,
401 U.S. 402 (1971) .................................... 25-27,69
City of Philadelphia v. Klutznick, 503 F. Supp.
663 (E.D. Pa. 1980) .............................................. 63
City of Willacoochee v. Baldrige, 556 F. Supp.
551 (S.D. Ga. 1983) .............................................. 26
v
Cases Page(s)
Clapper v. Amnesty Int’l USA, 568 U.S. 398
(2013) .................................................................... 23
Clark County v. FAA, 522 F.3d 437 (D.C. Cir.
2008) ........................................................... 32,49,51
Davis v. FEC, 554 U.S. 724 (2008) ............................ 21
Delaware Dep’t of Nat. Res. & Envtl. Control v.
EPA, 785 F.3d 1 (D.C. Cir. 2015) ......................... 54
Department of Commerce v. Montana, 503 U.S.
442 (1992) ............................................................. 64
Department of Commerce v. United States
House of Representatives, 525 U.S. 316
(1999) .......................................................... 25-27,50
Encino Motorcars, LLC v. Navarro, 136 S. Ct.
2117 (2016) ................................................. 37,39,40
Evenwel v. Abbott, 136 S. Ct. 1120 (2016) ............. 3,63
FCC v. Fox Television Stations, Inc., 556 U.S.
502 (2009) ............................................................. 29
Federation for Am. Immigration Reform v.
Klutznick, 486 F. Supp. 564 (D.D.C. 1980) ........... 2
Franklin v. Massachusetts, 505 U.S. 788
(1992) ..................................................... 26,28,30,64
Heckler v. Chaney, 470 U.S. 821 (1985) .................... 30
Judulang v. Holder, 565 U.S. 42 (2011) ................... 40
Lambert v. Hartman, 517 F.3d 433 (6th Cir.
2008) ..................................................................... 24
Lexmark Int’l, Inc. v. Static Control
Components, Inc., 572 U.S. 118 (2014) ................ 23
Lujan v. Defenders of Wildlife, 504 U.S. 555
(1992) .................................................................... 23
vi
Cases Page(s)
McDonnell Douglas Corp. v. United States Dep’t
of Air Force, 375 F.3d 1182 (D.C. Cir. 2004) ....... 35
Michigan v. EPA, 135 S. Ct. 2699 (2015) ............ 28,52
Motor Vehicle Mfrs. Ass’n of the U.S., Inc. v.
State Farm Mut. Auto. Ins. Co., 463 U.S. 29
(1983) ............................................................ passim
NAACP v. Alabama ex rel. Patterson, 357 U.S.
449 (1958) ............................................................. 24
National Cable & Telecomms. Ass’n v. Brand X
Internet Servs., 545 U.S. 967 (2005) .................... 29
National Fuel Gas Supply Corp. v. FERC, 468
F.3d 831 (D.C. Cir. 2006) ............................. passim
National Lifeline Ass’n v. FCC, 915 F.3d 19
(D.C. Cir. 2019)..................................................... 42
National Treasury Emps. Union v. Horner, 854
F.2d 490 (D.C. Cir. 1988) ..................................... 28
New England Coal. on Nuclear Pollution v.
Nuclear Regulatory Comm’n, 727 F.2d 1127
(D.C. Cir. 1984)..................................................... 51
NRDC v. NHTSA, 894 F.3d 95 (2d Cir. 2018) .......... 24
Safe Extensions, Inc. v. FAA, 509 F.3d 593
(D.C. Cir. 2007)..................................................... 37
SEC v. Chenery Corp., 318 U.S. 80 (1943) ................ 47
Texas v. Mosbacher, 783 F. Supp. 308 (S.D. Tex.
1992) ..................................................................... 26
Thompson v. United States Dep’t of Labor, 885
F.2d 551 (9th Cir. 1989) ....................................... 68
Troy Corp. v. Browner, 120 F.3d 277 (D.C. Cir.
1997) ..................................................................... 39
vii
Cases Page(s)
United States v. Armstrong, 517 U.S. 456
(1996) .................................................................... 58
United States v. New York Tel. Co., 434 U.S.
159 (1977) ............................................................. 62
Utah v. Evans, 182 F. Supp. 2d 1165 (D. Utah
2001) ..................................................................... 26
Utah v. Evans, 536 U.S. 452 (2002) .................. passim
Webster v. Doe, 486 U.S. 592 (1988) ..................... 26,30
Weyerhaeuser Co. v. United States Fish &
Wildlife Serv., 139 S. Ct. 361 (2018) .......... 25,26,28
Wisconsin v. City of New York, 517 U.S. 1
(1996) ............................................................ passim
Woods Petroleum Corp. v. United States Dep’t
of Interior, 18 F.3d 854 (10th Cir. 1994) ............. 58
Constitutional Provisions
U.S. Const.
art. I, § 2, cl. 3 .............................................. 2,62, 67
amend. XIV, § 2 ................................................. 2,62
Laws
Pub. L. No. 94-171, 89 Stat. 1023 (1975) .................. 12
Pub. L. No. 94-521, 90 Stat. 2459 (1976) ................. 6,7
Pub. L. No. 105-119, 111 Stat. 2440 (1997) ................ 3
2 U.S.C. § 2a(a) .......................................................... 27
5 U.S.C.
§ 701(a)(2) ............................................................. 25
§ 706 ...................................................................... 68
viii
Laws Page(s)
13 U.S.C.
§ 6 .................................................................. passim
§ 141(a) ......................................................... passim
§ 141(b) ................................................................. 27
§ 141(d) ................................................................. 50
§ 141(f) ........................................................ 16,61,62
§ 195 ................................................................... 7,50
44 U.S.C.
§ 3501 ...................................................................... 8
§ 3506 ...................................................................... 8
Miscellaneous Authorities
Carroll Wright & William Hunt, The History
and Growth of the United States Census
(1900) (S. Doc. No. 194),
https://tinyurl.com/y3ouuthc .......................... 3,4,40
Census Bureau, 1980 Census of Population
and Housing Part A: History (1986),
https://tinyurl.com/yy7ahzyr............................. 8,38
Census Bureau, 2000 Census of Population
and Housing: History vol. 1 (2009), https://
tinyurl.com/yxqz6hkq ............................................. 8
Daniel Cork, Census Testing, in Encyclopedia
of the U.S. Census (Margo Anderson et al.
eds., 2d ed. 2012) .................................................... 7
H.R. Rep. 92-1288 (1972) .......................................... 60
H.R. Rep. 94-1719 (1976) (Conf.) ........................ 6,7,29
Margo Anderson & Stephen Fienberg, Who
Counts? (2001) ..................................................... 3,4
Margo Anderson, The American Census (2d ed.
2015) .................................................................... 3-5
ix
Miscellaneous Authorities Page(s)
Mid-Decade Census, Hr’g Before the H.
Subcomm. on Census & Statistics (1967) ....... 60-61
Miriam Rosenthal, Striving for Perfection: A
Brief History of Advances and Undercounts
in the U.S. Census, 17 Gov’t Info. Q. 193
(2000) ...................................................................... 4
Plans for Taking the 1960 Census: Hr’g Before
the House Subcomm. on Census &
Government Statistics (1959) ................................. 5
S. Rep. 79-752 (1945) ................................................. 55
S. Rep. 94-1256 (1976) .............................................. 6,7
INTRODUCTION
The Constitution and the Census Act require the
federal government to count every person in this
country every ten years. This enumeration has
momentous consequences: it determines the allocation
of congressional seats, state and local apportionment,
and the distribution of billions of dollars in federal
funds. And there is just one chance each decade to get
the enumeration right.
To ensure that this extraordinarily complex
process serves its important purposes, the Department
of Commerce and the Census Bureau have developed
rigorous, scientifically tested standards to achieve an
accurate and complete count. In applying those
standards for the last seventy years, Commerce and
the Bureau have emphatically declined to ask a
citizenship question of every household. As the Bureau
has long recognized, a citizenship question would
exacerbate the undercount of noncitizen and Hispanic
households, rendering the enumeration inaccurate in
some States more than others, and undermining its
constitutional and statutory purposes.
Secretary of Commerce Wilbur Ross disregarded
this longstanding bipartisan and scientific consensus
and ordered that a citizenship question be added to the
2020 questionnaire. In doing so, the Secretary rejected
uncontroverted evidence showing that the citizenship
question would reduce response rates among
noncitizen and Hispanic households and thus harm
the enumeration’s distributive accuracy. He also
sidestepped the Bureau’s well-established procedures
for testing changes to the questionnaire to avoid
undercounts. And while the Secretary purported to
rely on a Department of Justice (DOJ) request for
2
better citizenship information for Voting Rights Act
(VRA) enforcement, he ignored the unanimous
evidence before him showing that more accurate
citizenship information could be provided at lower cost
without asking a citizenship question, and failed to
disclose the active role that he and his staff had played
in soliciting and then generating DOJ’s supposedly
independent request.
The United States District Court for the Southern
District of New York (Furman, J.) made detailed
factual findings that adding a citizenship question
would affirmatively undermine the accuracy of the
decennial census (among other harms) for no
demonstrable benefit. The court correctly held that
the Secretary’s decision violated the Administrative
Procedure Act (APA) because it was arbitrary and
capricious, it was contrary to two provisions of the
Census Act, and the rationale provided by the
Secretary was pretextual. For similar reasons, the
Secretary’s decision violated the Enumeration Clause.
For each of these reasons, the judgment below
should be affirmed.
STATEMENT
A. Modernization of the Decennial Census
1. The Constitution requires an “actual
Enumeration” of the population every ten years. Art.
I, § 2, cl. 3; amend. XIV, § 2. This enumeration must
count all residents, regardless of citizenship. See
Federation for Am. Immigration Reform v. Klutznick,
486 F. Supp. 564, 576 (D.D.C. 1980) (three-judge
court). The enumeration affects the apportionment of
representatives to Congress among the States; the
3
allocation of electors to the Electoral College; the
division of congressional, state, and local legislative
districts within each State; and the distribution of
hundreds of billions of dollars of federal funds.
Evenwel v. Abbott, 136 S. Ct. 1120, 1127-29 (2016).
Congress has delegated the conduct of the decennial
enumeration to the Secretary of Commerce, whose
decisions are constrained by both statutory restric-
tions and the constitutional requirement that the
census bear a “reasonable relationship to the accom-
plishment of an actual enumeration of the population.”
Wisconsin v. City of New York, 517 U.S. 1, 20 (1996);
see Pub. L. No. 105-119, § 209(a)(6), 111 Stat. 2440,
2481 (1997).
2. Before 1960, the decennial census was a
sprawling endeavor with two different and often
conflicting goals: counting the total population, and
collecting other demographic information. The census
questionnaire grew to include hundreds of questions
covering such disparate topics as occupations, literacy,
and health. See Carroll Wright & William Hunt, The
History and Growth of the United States Census 166
(1900) (S. Doc. No. 194), https://tinyurl.com/y3ouuthc.
The complexity of the questionnaire, administered by
in-person enumerators who interviewed respondents
and often misunderstood questions or answers, harmed
the accuracy of both the enumeration and demographic
data. Margo Anderson & Stephen Fienberg, Who
Counts? 19-23 (2001); Margo Anderson, The American
Census 206-07 (2d ed. 2015).
During this time, the census sometimes, but not
always, requested citizenship information. Before
4
1960, seven of the fifteen decennial censuses did not
seek citizenship status.
1
3. In 1960, the decennial census changed
dramatically into its modern form. Growing sophisti-
cation of data collection, statistical science, and
testing procedures had by then allowed Commerce and
the Bureau to understand that the census suffered
from serious data-accuracy problems, and to develop
ways to address those problems.
Increasingly robust evaluation procedures
demonstrated that the census undercounted the
population, and that this undercount was not spread
evenly across demographic groups or geographic
areas. Anderson, supra, at 215-20; Anderson &
Fienberg, supra, at 29-30. Evaluations of the 1950
census, for example, demonstrated that the census
had undercounted racial minorities at substantially
higher rates than others. Anderson & Fienberg, supra,
at 30.
Moreover, new data-collection techniques could
provide information as accurate as, and sometimes
more accurate than, demographic data collected via
the decennial census, while lowering costs and
lessening the burden on individual responders.
Miriam Rosenthal, Striving for Perfection: A Brief
History of Advances and Undercounts in the U.S.
Census, 17 Gov’t Info. Q. 193, 199-200 (2000). Govern-
ment records containing demographic information
(“administrative records”) had grown in number and
1
No citizenship inquiry appeared in 1790-1810, 1840-1860,
and 1880. Although some of these censuses asked about
birthplace, that question does not provide citizenship status.
Wright & Hunt, supra, at 132, 142-43, 147, 154, 166.
5
scope with the creation of agencies like the Social
Security Administration and the Immigration and
Naturalization Service. Anderson, supra, at 186-90.
And sampling—a technique that extrapolates informa-
tion about the entire population from data about a
representative subset—proved capable of producing
highly accurate demographic data without harming
the enumeration. Id. at 206; Plans for Taking the 1960
Census: Hr’g Before the House Subcomm. on Census &
Government Statistics (“1960 Plans”) 5-6 (1959).
Given these developments, the 1960 census was
dramatically changed to address the differential
undercount and to reduce burdens and costs. For the
first time, rather than relying on enumerators to visit
each household, the Bureau sent the questionnaire by
mail. See 1960 Plans, supra, at 6-7. Moreover, the
questionnaire sent to most households (the “short
form”) was reduced to a few simple, noncontroversial
questions, such as the number of individuals in each
household and their race, gender, and marital status.
All other demographic questions, including those
about citizenship, were removed and placed on a “long
form” questionnaire, initially sent to one of every four
households, and later to one of six households. (Pet.
App. 18a; see J.A.1211-1253 (2000 short- and long-
form questionnaires).)
The 1950 census was thus the last time the census
asked every household about citizenship. (Pet. App.
27a.) Ever since, Commerce and the Bureau have
vigorously opposed adding a citizenship question to the
questionnaire sent to every household, because doing
so will “inevitably jeopardize the overall accuracy of the
population count” by depressing responses from certain
populations and contributing to a differential under-
count. (Pet. App. 28a (quotation marks omitted).)
6
After the 2000 census, the long-form question-
naire was replaced by the American Community
Survey (ACS), a yearly survey conducted separately
from the decennial census, and distributed to about
one of every thirty-six households. (Pet. App. 18a-19a.)
4. Congress substantially reformed the Census Act
in 1976 to further modernize the census. Pub. L. No.
94-521, 90 Stat. 2459 (1976). These reforms permitted
Commerce to collect demographic information, but
placed important constraints on the use of the decen-
nial census questionnaire for that purpose and
prioritized other means of collecting such information.
To collect more up-to-date demographic
information than the decennial census provides,
Congress authorized the Secretary to conduct a mid-
decade census “in such form and content as he may
determine, including the use of sampling procedures
and special surveys.” § 7, 90 Stat. at 2461 (13 U.S.C.
§ 141(d)). Congress used the same quoted language in
the separate provision authorizing the Secretary to
conduct the decennial census. Id. (13 U.S.C. § 141(a)).
This delegation of authority was “essentially the
same” as “existing law,” H.R. Rep. 94-1719, at 11 (1976)
(Conf.), except that Congress added language to
“encourage the use of sampling,” S. Rep. 94-1256, at 4
(1976). The amendments also provided that the
Secretary may use the decennial census to collect
“other” information besides a “census of population,”
but only “as necessary.” § 7, 90 Stat. at 2461 (13 U.S.C.
§ 141(a)).
Congress also expressly limited the Secretary’s
ability to use the decennial census to collect demo-
graphic information (aside from total population), and
directed him to use more accurate and less costly
7
statistical techniques instead. First, in 13 U.S.C.
§ 6(c), Congress required the Secretary to use adminis-
trative records instead of census questions to collect
demographic data “[t]o the maximum extent possible”
given “the kind, timeliness, quality and scope of the
statistics required.” § 5, 90 Stat. at 2460. Second, where
“feasible,” the Secretary must use sampling instead of
census questions to obtain demographic information
beyond the enumeration. § 10, 90 Stat. at 2464 (13
U.S.C. § 195). By prioritizing other means of collecting
demographic information, Congress intended to mini-
mize the census’s burden on responding individuals
and thus maximize census responses. H.R. Rep. 94-
1719, at 10; S. Rep. 94-1256, at 1, 5.
5. Since at least 1940, the decennial census has
undergone extensive pretesting before census day.
Daniel Cork, Census Testing, in Encyclopedia of the
U.S. Census 79, 79 (Margo Anderson et al. eds., 2d ed.
2012). Such testing reflects not only the important
consequences of the census, but also the fact that the
decennial census is conducted only once every ten
years, with little room for correction if problems arise.
Pretesting is now a multi-year endeavor that subjects
nearly every aspect of the census to a battery of
evaluations, culminating in a comprehensive “dress
rehearsal” to understand how all aspects of the census
work together. Id. at 79-81. (J.A.1296.) A critical part
of that process is pretesting individual questions to
ensure that they yield accurate data without reducing
census responses. Pretesting includes administering
the questions to a sample of respondents to verify that
they “[c]an be understood,” “[a]re not unduly sensitive,”
and “do not cause undue burden.” (J.A.627-628.)
Overlapping statutes, guidelines, and agency
practices govern testing. For example, the Office of
8
Management and Budget’s (OMB) data-quality
standards direct Commerce to design the census “to
achieve the highest practical rates of response,” and
thus require pretesting of census questions. (J.A.657.)
See 44 U.S.C. §§ 3501(1)-(2), 3506(e). The Bureau’s
Statistical Quality Standards likewise require that
the census “be pretested with respondents to identify
problems” before implementation. (J.A.626.) When an
already-pretested survey undergoes “substantive
modifications,” including the addition of a new ques-
tion, “[p]retesting must be performed” again. (J.A.627.)
Commerce has consistently refused to add
questions to the decennial census that performed
poorly in pretesting. For the 1980 census, Commerce
declined to replace a question about Spanish origin
with a question about ethnicity after testing showed
that misunderstandings about the question resulted
in high nonresponse rates. Census Bureau, 1980
Census of Population and Housing Part A: History 2-
20 (1986), https://tinyurl.com/yy7ahzyr. For the 2000
census, Commerce declined to add a question
requesting Social Security numbers after pretesting
revealed a disproportionately distributed “3.4% decline
in self-response rates attributable to the question.”
(Pet. App. 30a; see J.A.892-893.) See 1 Census Bureau,
2000 Census of Population and Housing: History 46
(2009), https://tinyurl.com/yxqz6hkq.
B. The Decision to Add a Citizenship Question
In a memorandum dated March 26, 2018, the
Secretary announced his decision to add a citizenship
question to the 2020 census questionnaire sent to
every household. (Pet. App. 548a-563a.)
9
1. The Secretary represented that he began
assessing whether to add a citizenship question
“[f]ollowing receipt” of a December 2017 letter from
DOJ requesting block-level citizenship data to help
enforce § 2 of the VRA. (Pet. App. 548a-549a.) But as
the Secretary later acknowledged in a June 2018
supplemental decision memorandum, DOJ’s letter
had not initiated his decision-making. Rather, the
Secretary had begun his “deliberative process” soon
after his appointment in February 2017—almost a
year before DOJ’s letter. (Pet. App. 546a.) And DOJ
had not submitted the December 2017 letter on its
own initiative; rather, the Secretary and his staff had
approached DOJ to urge them to request a citizenship
question. (Pet. App. 82a-84a.)
The supplemental memorandum also failed to
fully disclose the Secretary’s engagement with the
issue before December 2017. (Pet. App. 74a-99a, 118a-
129a.) The Secretary actually “made the decision
months before DOJ sent its letter.” (Pet. App. 118a.)
The Secretary and his staff then “actively lobbied
other agencies” to request a citizenship question,
including both DOJ and the Department of Homeland
Security (DHS). (Pet. App. 121a.) After both agencies
declined (Pet App. 82a-84a), the Secretary reached out
to then–Attorney General Sessions, who discussed the
issue with John Gore, then the Acting Assistant
Attorney General for Civil Rights. (Pet. App. 89a-90a.)
The Attorney General’s senior counselor reassured the
Secretary’s Chief of Staff that DOJ would “do whatever
you all need us to do.” (J.A.254.) Gore then wrote
DOJ’s December 2017 letter, signed by another DOJ
official, requesting the addition of a citizenship ques-
tion to the decennial questionnaire to provide DOJ
10
with block-level citizenship data for VRA enforcement.
(Pet. App. 91a-95a, 564a-569a.)
2. In December 2017 and January 2018, in
response to DOJ’s letter, Dr. John Abowd—the
Bureau’s Chief Scientist—and his team of experts
analyzed the effects of adding a citizenship question to
the decennial questionnaire in a series of memoranda.
(J.A.104-122, 290-295, 301-318.) The Bureau conducted
this analysis without any awareness of the Secretary’s
involvement in generating DOJ’s letter (Pet. App.
116a-117a).
The memoranda warned the Secretary that
adding the question would not only depress the initial
response rate for all households, but would also
depress the response rate of households with a
noncitizen by at least 5.1 percentage points more than
for citizen households—approximately 1.6 million
more people not responding. (J.A.114-115 (630,000
households); J.A.1008 n.58 (2.54 persons per house-
hold).) The memoranda explained that because this
estimate was “cautious,” the actual differential
reduction in self-response rates would likely be much
greater. (J.A.114-116.) (The Bureau later updated its
analysis to warn that the differential decline in
response rates would be at least 5.8 percentage
points—approximately 6.5 million people (J.A. 1008).
2
)
While the Bureau attempts to address initial
nonresponses through Nonresponse Followup (NRFU)
2
This updated conclusion reflected both the increase from
5.1% to 5.8% and updated figures on the number of noncitizen
households and the average number of people per household.
(J.A.1008 & n.58.)
11
procedures,
3
the memoranda warned that NRFU
would be “very costly” (J.A.105, 115) and would not
resolve the problems introduced by the lower response
rate (J.A.113-116).
By contrast, the Bureau explained that it could
use existing administrative records to produce block-
level citizenship data as accurate as the block-level
race, age, and ethnicity data DOJ already uses for
VRA purposes. (J.A.105-107, 290-292, 317-318.) The
Bureau would use the “Numident,” a database contain-
ing “information on every person” with a Social Security
or Individual Taxpayer Identification Number.
(J.A.117.) This database contains highly reliable
citizenship information because individuals must
provide proof of citizenship or immigration status to
obtain these numbers. (J.A.117.) The Bureau would
then “link” individual census responses to these
database records by matching personal identifying
information. (J.A.156, 158, 954.) The Bureau could
already link roughly 90% of census respondents to
Numident records, and planned to obtain additional
records from other agencies to increase the number of
successful linkages. (J.A.120-121, 133-135, 154-155.)
The Bureau could then integrate this citizenship data
with the other block-level census data (known as
“PL94-171 data”) that the Secretary produces and
makes publicly available after every decennial census
for redistricting and that DOJ already uses for VRA
enforcement. (J.A.105-107, 290-292, 317-318, 860,
3
NRFU includes visits by enumerators; “use of adminis-
trative records; collection of information from ‘proxies,’ such as
neighbors or landlords; and ‘imputation,’ a process through which
the Census Bureau extrapolates data about households” from
comparable household data. (Pet. App. 151a.)
12
906-908.) See Pub. L. No. 94-171, 89 Stat. 1023 (1975)
(13 U.S.C. § 141(c)).
Because a citizenship question would generate
“substantially less accurate” citizenship data than
administrative records and would also impair the
enumeration, the Bureau recommended using admin-
istrative records to provide block-level citizenship
data to DOJ. (J.A.105.) The Bureau’s Acting Director
informed DOJ that this approach would provide
“higher quality [citizenship] data produced at lower
cost” than adding a citizenship question. (J.A.265.)
Although the Bureau sought to meet with DOJ to
discuss this recommendation, the Attorney General
directed DOJ to decline such a meeting. (Pet. App.
95a-97a; J.A.266.)
Meanwhile, the Secretary directed the Bureau to
analyze the effects of using both a citizenship question
and administrative records to generate citizenship
data. (Pet. App. 51a-58a.) In a memorandum dated
March 1, 2018, Dr. Abowd and his team provided the
Secretary with an analysis concluding that this hybrid
approach would “have all the negative cost and quality
implications” of adding the citizenship question—
including a decrease in self-response rates—while
“result[ing] in poorer quality citizenship data than”
using administrative records alone. (J.A.158-159.)
13
3. The Secretary then issued his March 26, 2018,
decision memorandum announcing that he would both
add a citizenship question to the decennial question-
naire and use administrative records. (Pet. App. 548a-
563a.) The memorandum asserted that “limited empir-
ical evidence exists about whether adding a citizen-
ship question would decrease response rates” (Pet.
App. 557a), disregarding the multiple empirical
analyses demonstrating that adding the question would
disproportionately decrease response rates and harm
the enumeration’s accuracy. The memorandum also
stated that the advantage of the Secretary’s approach
was that it would provide DOJ the “most complete and
accurate” citizenship data (Pet. App. 556a), contrary
to evidence that this approach would provide less
complete and less accurate citizenship data than using
administrative records alone. The Secretary further
claimed that the citizenship question was sufficiently
“well tested” (Pet. App. 550a), even though the
question had not undergone any of the testing that
governs the census questionnaire.
C. Procedural History
1. New York, seventeen other States, sixteen local
governments, and the U.S. Conference of Mayors
(“government respondents”) filed a complaint alleging
that the Secretary’s decision to add a citizenship ques-
tion violated the APA and the Enumeration Clause.
14
2. Petitioners’ initial Administrative Record
contained scarcely any documents preceding DOJ’s
December 2017 letter, despite the Secretary’s acknowl-
edgment in his supplemental decision memorandum
that he had been deliberating the citizenship question
during that time. (Pet. App. 546a.) On July 3, 2018,
the district court ordered petitioners to complete the
Administrative Record, authorized limited expert
discovery, and authorized additional discovery based
on the irregularity of petitioners’ initial record and on
a strong showing of petitioners’ bad faith and
improper behavior. (Pet. App. 523a-531a.) On August
17, the district court authorized a deposition of Gore
(Pet. App. 452a-455a), and on September 21 authorized
a deposition of the Secretary (Pet. App. 437a-439a).
This Court stayed the Secretary’s deposition but
declined to stay Gore’s deposition or other discovery.
139 S. Ct. 16 (2018). The Court then granted certiorari
to review the pretrial discovery orders, 139 S. Ct. 566
(2018) (No. 18-557), but declined to stay the trial, 139
S. Ct. 452 (2018).
3. Meanwhile, the district court denied petitioners’
motion to dismiss the APA claims, concluding that the
Secretary’s decision to add a citizenship question to
the decennial census was reviewable. (Pet. App. 21a-
25a, 402a-408.)
The court did, however, dismiss respondents’
Enumeration Clause claim for failure to state a claim.
(Pet. App. 408a-424a.) While the court recognized that
the Enumeration Clause reflects “a strong constitu-
tional interest in accuracy” (Pet. App. 423a (quoting
Utah v. Evans, 536 U.S. 452, 478 (2002))), it held that
the existence of a citizenship inquiry before 1960
precluded the argument that such a question was
15
altogether forbidden by the Constitution (Pet. App.
412a).
4. After an eight-day trial, the district court issued
an opinion containing detailed findings of fact and
conclusions of law. (Pet. App. 1a-353a.) The court
entered final judgment vacating the Secretary’s
decision, enjoining the addition of a citizenship
question to the 2020 census unless the legal defects
identified by the court were cured, and remanding to
Commerce. (Pet. App. 352a.)
First, the court evaluated what evidence it could
properly consider. As the parties had agreed, the court
considered the Administrative Record for any purpose
(Pet. App. 250a) and extra-record evidence to deter-
mine respondents’ standing (Pet. App. 129a-130a).
The court did not consider extra-record evidence to
resolve whether petitioners had violated the APA,
except where such material illuminated technical
matters or showed a failure to consider important
factors. (Pet. App. 260a-261a.) The court further deter-
mined that, while it could permissibly consider extra-
record material to decide whether the Secretary’s
decision was pretextual, it did not need to do so because
it “would reach the same conclusions” based solely on
the Administrative Record. (Pet. App. 261a.)
Second, the court determined that respondents
had standing. (Pet. App. 130a-239a.)
Third, the court ruled that the Secretary’s decision
violated the APA in multiple independent ways. The
decision was arbitrary and capricious, and based on a
pretextual rationale. The decision was also contrary to
law because it violated two statutes: one requiring the
Secretary, “[t]o the maximum extent possible,” to
16
acquire demographic information using administra-
tive records rather than direct inquiries, 13 U.S.C.
§ 6(c); and another precluding the Secretary from
altering previously reported census topics without
making certain findings and filing a new report with
Congress, id. § 141(f). (Pet. App. 261a-321a.) The court
noted that extra-record evidence confirmed, but was
not essential to, its conclusions on the APA claims.
(Pet. App. 313a-315a, 320a-321a.)
Fourth, the court rejected the Fifth Amendment
equal protection claim brought by private respondents
in a consolidated case. (Pet. App. 322a.)
Finally, the court vacated as moot its order
authorizing the Secretary’s deposition. (Pet. App.
352a-353a.)
5. In No. 18-557, the parties submitted opening
briefs in this Court addressing the pretrial discovery
issues. After the district court entered final judgment,
respondents moved to dismiss the writ as improvi-
dently granted. The Court removed No. 18-557 from
its argument calendar and suspended further briefing.
The Court then granted certiorari before judgment in
this case.
6. In a separate proceeding brought by different
plaintiffs to challenge the addition of a citizenship
question, the United States District Court for the
Northern District of California (Seeborg, J.), on March
6, 2019, issued a post-trial decision holding that the
Secretary’s decision violated both the APA and the
Enumeration Clause. California v. Ross, No. 18-cv-
1865, 2019 WL 1052434 (N.D. Cal. 2019). This Court
directed the parties here to brief and argue, as an
alternative ground for affirmance, whether the
Secretary’s decision violated the Enumeration Clause.
17
SUMMARY OF ARGUMENT
I. Government respondents have standing.
Petitioners’ opening brief does not dispute the district
court’s extensive findings about injury and redressa-
bility. Instead, petitioners argue only that respondents’
injuries are not traceable to the Secretary’s decision to
add a citizenship question because they are more
proximately caused by the unlawful and irrational
failure of third parties to respond to the census
questionnaire. But this Court has long held that the
intervening acts of third parties do not break the
causal chain if those acts predictably result from
challenged conduct. Here, the district court made
factual findings—uncontested by petitioners—that
the citizenship question will cause differential
nonresponse rates for noncitizens and Hispanics,
leading to a differential undercount and a decrease in
data quality that will concretely injure respondents.
Given this proof that third parties will react in ways
that harm government respondents, it is immaterial
whether their reactions are unlawful or irrational.
II. The Secretary’s decision to add a citizenship
question violates the APA.
A. The Secretary’s decision is reviewable.
Petitioners misread § 141(a) of the Census Act, 13
U.S.C. § 141(a), as conferring unreviewable discretion
on the Secretary to place whatever questions he wants
on the decennial questionnaire. This Court has
repeatedly reviewed the Secretary’s actions under this
provision. And petitioners’ argument is inconsistent
with the 1976 Census Act, which is the source of the
language that petitioners rely on here. That enactment
included multiple provisions that constrained the
Secretary’s conduct of the census and provide judicially
18
manageable standards here, including the require-
ment that the Secretary rely, to the “maximum extent
possible,” on administrative records rather than
census questions to collect demographic information,
13 U.S.C. § 6(c).
B. The Secretary’s decision was arbitrary and
capricious.
First, the Secretary unreasonably ignored the
uncontroverted empirical evidence that the citizen-
ship question would make the enumeration less
accurate. All the evidence in the Administrative
Record demonstrates that a citizenship question
would cause millions of noncitizens and Hispanics to
not respond to the census, undermining the accuracy
of the constitutionally required headcount.
The Secretary was not entitled to dismiss this
evidence as “inconclusive.” That label is contradicted
by the firm conclusions of the analyses themselves.
More fundamentally, even if the evidence of an under-
count were inconclusive, the Secretary’s actions would
still be unreasonable because he abandoned the well-
established process for testing proposed changes to the
questionnaire. The testing process ensures that no
change is made to the census without understanding its
effects. Under this longstanding conservative
approach, inconclusive evidence of harm does not
permit altering the questionnaire without further
testing.
Second, the Secretary acted contrary to the
evidence by concluding that a citizenship question was
necessary to provide DOJ with information for VRA
enforcement. The Bureau informed the Secretary that
it could obtain data sufficient to address DOJ’s
purported needs, without adding a citizenship question,
19
by linking highly reliable administrative records
containing citizenship status to individual census
responses. The Secretary rejected that proposal in
favor of a solution that purportedly would provide
even more accurate citizenship information. But the
Secretary never explained how using his preferred
approach rather than the administrative-records
approach would result in any improvement to VRA
enforcement.
In any event, all the evidence in the Administrative
Record demonstrates that the Secretary’s solution—
using administrative records and adding a citizenship
question—would provide less accurate citizenship
data, at greater cost, than relying on administrative
records alone. The Secretary disregarded the fact that
the citizenship question will introduce significant
errors not present under the administrative-records
approach: at least 9.5 million wrong responses on
citizenship status; the loss of information about
additional millions of noncitizens and Hispanics due
to the undercount; and an inability to link an
additional one million individuals to administrative
records. These harms are not offset by an increase in
the number of census responses on citizenship status,
as petitioners claim: the Administrative Record shows
that a citizenship question will often trigger inaccurate
responses, and the Bureau expressly concluded that
sophisticated modeling based on available adminis-
trative records would produce comparatively more
accurate citizenship information.
Third, the Secretary failed to justify his conclusion
that the purported benefit of providing DOJ more
citizenship data “outweighs” any harm to the accuracy
of the enumeration. The Constitution and the Census
Act require the Secretary to prioritize an accurate
20
enumeration due to the momentous consequences of
the headcount, including its effect on the allocation of
House seats and billions of dollars in federal funds.
The Secretary provided no explanation for his
judgment to subordinate that priority in favor of
generating what he asserts (counter to the evidence)
would be incrementally more accurate citizenship data.
Fourth, the Secretary’s stated rationale was
pretextual. While he claimed to be relying on DOJ’s
independent judgment about the need for a citizenship
question, the district court found that it was in fact the
Secretary and his staff who engineered DOJ’s request
from the outset. The Secretary’s supposed reliance on
DOJ’s expertise thus could not provide the necessary
rationale for his decision.
C. The Secretary’s decision was also contrary to
law. First, given the evidence that administrative
records alone would satisfy DOJ’s VRA-enforcement
needs, the Secretary’s decision to add a citizenship
question violated 13 U.S.C. § 6(c), which requires the
Secretary to collect demographic information using
administrative records to the maximum extent
possible instead of by posing direct inquiries through
the decennial census. Second, the Secretary violated
13 U.S.C. § 141(f) by adding the citizenship question
without submitting the mandated report to Congress
or making the required findings that new circum-
stances necessitated a change.
III. The Secretary’s decision violated the
Enumeration Clause. That provision requires the
Secretary’s decisions about the census to be reason-
ably related to the pursuit of an accurate enumeration
of the total population. The Secretary flouted this
21
constitutional obligation by adding a citizenship ques-
tion that would affirmatively undermine the accuracy
of the headcount. And the Secretary’s justification
that the question would provide valuable information
to DOJ is contradicted by the Administrative Record
and the evidence produced at trial.
IV. Insofar as petitioners’ challenge to extra-
record discovery is not moot, this Court should reject
it. The district court properly authorized discovery
beyond the Administrative Record because petitioners
had concededly failed to disclose the full basis for the
Secretary’s decision and had in fact obfuscated the
Secretary’s decision-making process. Discovery was
therefore essential to provide the “whole record” that
the APA requires.
ARGUMENT
I. GOVERNMENT RESPONDENTS HAVE STANDING.
To have standing, “a claimant must present an
injury that is concrete, particularized, and actual or
imminent; fairly traceable to the defendant’s
challenged behavior; and likely to be redressed by a
favorable ruling.” Davis v. FEC, 554 U.S. 724, 733
(2008). The district court correctly held that govern-
ment respondents have standing here.
A. Petitioners do not challenge injury or
redressability in their opening brief. The district court
made extensive factual findings demonstrating that
the addition of a citizenship question to the 2020
census will both cause a net differential undercount of
noncitizen and Hispanic households and—separate
from the undercount—will irreparably harm the accu-
racy of census data used by government respondents
22
for essential governmental functions. (Pet. App. 141a-
147a, 150a-168a, 184a-187a, 233a-236a). Specifically,
the district court found that adding a citizenship
question will reduce noncitizen self-responses by “at
least 5.8%”—or roughly 6.5 million people—and also
significantly reduce self-responses from Hispanic
households (Pet. App. 150a, 169a; J.A.1008). Because
NRFU would not cure these differential declines, these
reductions would result in a net incremental under-
count of noncitizens and Hispanics. (Pet. App. 169a.)
That differential undercount and other harms to
data accuracy will injure government respondents in
at least four ways: (a) loss of seats in Congress and in
state and local legislatures; (b) loss of federal funding;
(c) harm to accurate population data used to distribute
government services; and (d) forced diversion of
resources. (Pet. App. 173a-194a.) Petitioners contest
neither these factual findings nor the district court’s
legal conclusion (Pet. App. 194a-239a) that these
injuries satisfy Article III.
4
B. Instead, petitioners argue (Br. 17-21) that
government respondents’ injuries would not be fairly
traceable to the Secretary’s decision because they are
more proximately caused by individuals’ unlawful and
irrational reactions to the addition of a citizenship
question.
This argument misconceives the requirements for
traceability. When a third party’s actions are part of
4
These findings rebut petitioners’ assertion (Br. 19-20) that
recognizing standing here would permit challenges to “any
demographic question on the decennial census.” Under the
district court’s reasoning, plaintiffs would have standing to
challenge only census questions proved to cause such harms, not
any question to which some people refuse to respond.
23
the causal chain, all that is required is a showing that
the challenged conduct had a “determinative…effect”
on that third party. Bennett v. Spear, 520 U.S. 154,
169 (1997). Because “[p]roximate causation is not a
requirement” for standing, Lexmark Int’l, Inc. v. Static
Control Components, Inc., 572 U.S. 118, 134 n.6
(2014), challenged conduct need not be the only step—
or even “the very last step”—in the causal chain,
Bennett, 520 U.S. at 169. Here, the district court
found, and petitioners do not challenge, that the citizen-
ship question will affect the census responses of
noncitizens and Hispanics, leading directly to
government respondents’ injuries.
Petitioners are thus wrong to characterize the
district court’s reasoning as relying on “‘speculation
about the decisions of independent actors.’” Br. 18
(quoting Clapper v. Amnesty Int’l USA, 568 U.S. 398,
414 (2013)). When a plaintiff has no direct evidence
about how a third party will react to challenged
conduct, it may be speculative to assume that the third
party will react in an irrational or illegal way. But
here, because respondents proved how noncitizens and
Hispanics would react to the citizenship question (Pet.
App. 228a), there is no need for judicial “guesswork as
to how independent decisionmakers will exercise their
judgment,” Clapper, 568 U.S. at 413.
C. Petitioners’ arguments (Br. 17-21) about the
rationality or lawfulness of individuals’ responses to
the citizenship question are thus beside the point. The
rationality or lawfulness of a third party’s reaction to
challenged conduct has never been a barrier to standing
so long as a plaintiff can show that the third party will
react “in such manner as to produce causation.” Lujan
v. Defenders of Wildlife, 504 U.S. 555, 562 (1992).
24
This Court and others have consistently
recognized standing based on third parties’ irrational
or illegal responses to challenged governmental action.
In NAACP v. Alabama ex rel. Patterson, for example,
this Court found standing to challenge a law compelling
disclosure of the NAACP’s membership based on proof
that past disclosures had caused third parties to
respond irrationally—and often illegally—through
“economic reprisal, [termination] of employment,” and
threats of physical injury. 357 U.S. 449, 462 (1958).
And in Block v. Meese, the D.C. Circuit held that a
distributor could challenge the government’s classifi-
cation of a film as “political propaganda” based on the
anticipated, albeit irrational, public reaction to that
classification. 793 F.2d 1303, 1307-09 (D.C. Cir. 1986)
(Scalia, J.).
Similarly, plaintiffs have standing to challenge a
defendant’s failure to safeguard private information
even when a data thief was “the most immediate cause
of plaintiffs’ injuries.” Attias v. CareFirst Inc., 865
F.3d 620, 629 (D.C. Cir. 2017), cert. denied, 138 S. Ct.
981 (2018); Lambert v. Hartman, 517 F.3d 433, 437-38
(6th Cir. 2008). And plaintiffs have standing to
challenge a federal agency’s reduction of a financial
penalty when the reduction will predictably increase
the likelihood that regulated entities will fail to
comply with the law. NRDC v. NHTSA, 894 F.3d 95,
104 (2d Cir. 2018). Contrary to petitioners’ attempted
distinction (Br. 20), these rulings did not turn on any
finding that the defendant breached a legal duty to
protect the plaintiffs, but instead recognized that the
third parties’ irrational or illegal action was simply
one step in the causal chain connecting plaintiffs’
harm to defendants’ conduct, see Bennett, 520 U.S. at
169.
25
D. As petitioners acknowledged below (Pet. App.
480a-481a), their position would effectively preclude
anyone from having standing to challenge census deci-
sions that reduce participation, even decisions made
with the intent and predictable effect of doing so. But
this Court’s precedents do not support petitioners’
claim that nonresponses are legally irrelevant to
standing. The Court has repeatedly heard census-
related cases in which the asserted harm to the
plaintiff resulted from third parties’ failure to respond
to the census. See Evans, 536 U.S. at 457-58 (imputing
data for people who failed to respond); Department of
Commerce v. United States House of Representatives,
525 U.S. 316, 324 (1999) (sampling as part of NRFU).
Petitioners have a constitutional and statutory duty to
address such nonresponses, not to ignore them.
II. THE SECRETARY’S DECISION VIOLATED THE APA.
A. The Secretary’s Decision Is Reviewable.
The district court properly rejected petitioners’
argument that Congress vested the Secretary with
unreviewable discretion over the decennial census
questionnaire. (Pet. App. 398a-408a.) Agency action is
subject to a “strong presumption” of judicial review.
Weyerhaeuser Co. v. United States Fish & Wildlife
Serv., 139 S. Ct. 361, 370 (2018). The APA’s narrow
exception to this presumption for decisions “committed
to agency discretion by law,” 5 U.S.C. § 701(a)(2),
applies only in “rare instances” where Congress has
provided “clear and convincing evidence” that it vested
an agency with unfettered discretion, Citizens to Pres.
26
Overton Park, Inc. v. Volpe, 401 U.S. 402, 410 (1971).
No such evidence exists here.
5
1. Petitioners rely (Br. 21-22) on language in 13
U.S.C. § 141(a) mandating that the Secretary “shall…
take a decennial census of population…in such form
and content as he may determine.” But that language
merely requires the Secretary to conduct a decennial
census and was not “intended to effect a new, unreview-
able commitment to agency discretion.” Franklin v.
Massachusetts, 505 U.S. 788, 816 n.16 (1992) (Stevens,
J., concurring in part & in judgment). Requiring an
agency to exercise some discretion does not confer
unfettered discretion. See Weyerhaeuser, 139 S. Ct. at
370.
Indeed, this Court and others have repeatedly
considered challenges involving the Secretary’s
authority under § 141(a) to conduct the decennial
census—the same authority he invokes here—and
squarely rejected the argument that there are no
manageable standards to apply.
6
Petitioners attempt
to distinguish these cases by asserting (Br. 25) that
none of them involved the census questionnaire, but
that distinction finds no support in the statute. The
“form and content” that § 141(a) requires the Secretary
to “determine” is not the content of the questionnaire
specifically, but rather the conduct of the decennial
5
The APA’s exception to reviewability does not apply to
respondents’ Enumeration Clause claim. See Franklin v.
Massachusetts, 505 U.S. 788, 801 (1992); Webster v. Doe, 486 U.S.
592, 601-02 (1988).
6
See House of Representatives, 525 U.S. 316; Carey v.
Klutznick, 637 F.2d 834 (2d Cir. 1980) (per curiam); Utah v.
Evans, 182 F. Supp. 2d 1165 (D. Utah 2001), aff’d, 536 U.S. 452;
Texas v. Mosbacher, 783 F. Supp. 308 (S.D. Tex. 1992); City of
Willacoochee v. Baldrige, 556 F. Supp. 551 (S.D. Ga. 1983).
27
census generally—as confirmed by § 141(a)’s inclusion
of “sampling” and “special surveys” within “form and
content.” Just as other aspects of the Secretary’s
conduct of the census under § 141(a) are reviewable,
so too are his decisions regarding the questionnaire.
That conclusion is consistent with other statutory
language and the broader context of the 1976 Census
Act, which added the quoted language to § 141(a). The
1976 Act not only continued to require the Secretary
to conduct the decennial census, but more specifically
confirmed his duty to pursue an accurate enumeration
and expressly constrained his authority to use the
census to gather demographic information other than
the enumeration. In particular, Congress directed the
Secretary to use other techniques besides the census
to collect demographic information such as citizenship
status. These “narrower and more specific” statutory
provisions inform the Secretary’s authority under
§ 141(a), House of Representatives, 525 U.S. at 338,
and provide ample “law to apply,” Overton Park, 401
U.S. at 410 (quotation marks omitted).
First, the 1976 Act, consistent with the
Constitution, imposed a duty to pursue an accurate
enumeration. The “strong constitutional interest in
accuracy,” Evans, 536 U.S. at 478, requires that
Congress’s census-related decisions bear a “reasonable
relationship to the accomplishment of an actual
enumeration of the population,” Wisconsin, 517 U.S.
at 20. Congress, in turn, built the pursuit of an accurate
enumeration into the language and structure of the
Census Act by requiring the Secretary to produce a
“tabulation of total population” in each State “as
required for the apportionment of Representatives in
Congress,” § 141(b). See 2 U.S.C. § 2a(a). “This statu-
tory command…embodies a duty to conduct a census
28
that is accurate….” Franklin, 505 U.S. at 819-20
(Stevens, J.).
Second, the 1976 Act required the Secretary to
rely on administrative records to obtain demographic
data, “instead of conducting direct inquiries” on the
decennial census, “[t]o the maximum extent possible
and consistent with the kind, timeliness, quality and
scope of the statistics required.” 13 U.S.C. § 6(c).
Petitioners’ contention that § 6(c) provides no judi-
cially manageable standards (Br. 45-46) ignores the
“maximum extent possible” requirement—directive
language of the type that courts “routinely assess,”
Weyerhaeuser, 139 S. Ct. at 371. See Alaska Dep’t of
Envtl. Conservation v. EPA, 540 U.S. 461, 496-97
(2004) (“best available control technology”); Biodiver-
sity Legal Found. v. Babbitt, 146 F.3d 1249, 1254 (10th
Cir. 1998) (“maximum extent practicable”). And the
Secretary’s reliance on effective VRA enforcement as
the rationale for adding a citizenship question provides
judicially manageable standards to evaluate whether
the question is needed for that purpose and whether
administrative records would produce “the kind,
timeliness, quality and scope of the statistics
required” for that purpose. (As explained infra at 43,
59-61, the Secretary’s reasoning failed to satisfy those
standards.)
Third, the Act provides that the Secretary may use
the decennial questionnaire to collect “other” informa-
tion besides a “census of population,” but only “as
necessary.” § 141(a). Courts routinely interpret similar
language as imposing judicially enforceable constraints.
See Michigan v. EPA, 135 S. Ct. 2699, 2707 (2015)
(“appropriate and necessary”); National Treasury
Emps. Union v. Horner, 854 F.2d 490, 495 (D.C. Cir.
1988) (“necessary”).
29
Congress imposed these limitations on the use of
the census questionnaire based on the growing under-
standing that administrative records and sampling
could enable the Secretary to obtain sufficiently
accurate demographic data without undermining the
accuracy of the enumeration. See H.R. Rep. 94-1719,
at 10. This history demonstrates that, by adding the
language of § 141(a) in the 1976 Act, Congress did not
intend to grant the Secretary unreviewable discretion
to collect demographic information from the decennial
census without regard to whether that information is
obtainable through other means.
2. The established testing procedures that have
long governed the decennial census provide further
standards by which to evaluate whether the Secretary’s
decision was arbitrary and capricious. Under applica-
ble statutes, guidelines, and practices, pretesting is
required even for minor changes to the census
questionnaire to preserve the accuracy of the enumera-
tion and other census data. See supra at 7-8. Peti-
tioners miss the mark in asserting (Br. 39-40) that
these requirements are not legally binding on the
Secretary. Petitioners conceded below that OMB’s
directives, which require pretesting, do legally bind the
Secretary. (Pet. App. 308a-309a.) In any event, courts
do not look only to legally binding obligations to
evaluate APA claims; to the contrary, this Court has
routinely relied on an agency’s “past practice,”
National Cable & Telecomms. Ass’n v. Brand X
Internet Servs., 545 U.S. 967, 981 (2005), or “serious
reliance interests” by regulated entities, FCC v. Fox
Television Stations, Inc., 556 U.S. 502, 515 (2009).
OMB’s directives and the Bureau’s Statistical Quality
Standards likewise provide a basis by which to
evaluate the reasonableness of the Secretary’s
30
decision to add a citizenship question without
adequate testing.
3. These statutory restrictions and established
standards together create a regime far different from
the one in Webster v. Doe, a case that arose in the
distinct context of national security and that involved
a single provision authorizing the CIA director to
terminate an employee whenever he “shall deem such
termination necessary or advisable in the interests of
the United States.” 486 U.S. 592, 594 (1988) (quota-
tion marks omitted). The Census Act contains “[n]o
language equivalent to ‘deem…advisable,’” Franklin,
505 U.S. at 817 (Stevens, J.), and instead contains
specific provisions constraining the Secretary’s collec-
tion of demographic information.
Petitioners also misplace reliance on Heckler v.
Chaney, arguing that the Secretary’s decision involves
a “complicated balancing of a number of factors”
immune from judicial review. Br. 24 (quoting 470 U.S.
821, 831 (1985)). Heckler referred only to the
“complicated balancing” of factors inherent in an
agency’s refusal to exercise enforcement power—
authority that, unlike the conduct of the census, is
traditionally reserved to the unreviewable discretion
of the executive branch. 470 U.S. at 831-32. By
contrast, judicial review of census-related decisions
has long helped to ensure “public confidence in the
integrity” of the census and to “strengthen this
mainstay of our democracy.” Franklin, 505 U.S. at 817
(Stevens, J.).
31
B. The Secretary’s Decision Was
Arbitrary and Capricious.
In deciding to add a citizenship question to the
2020 census questionnaire, the Secretary disregarded
uncontroverted evidence about the question’s impact
on response rates, unreasonably concluded that the
question was necessary to provide DOJ block-level
citizenship information for VRA enforcement, and
made an unexplained policy judgment that any
purported benefits to DOJ were “of greater importance”
than any harm to the enumeration. (Pet. App. 562a.)
Each step of this process was a “classic, clear-cut APA
violation[].” (Pet. App. 10a.)
1. The Secretary Disregarded Harms
to the Enumeration.
a. Undisputed evidence demonstrated
that a citizenship question would
depress response rates.
As the Secretary recognized, it was “incumbent”
upon him “to make every effort to provide a complete
and accurate decennial census” given the constitu-
tional and practical importance of an accurate
enumeration. (Pet. App. 549a.) The Secretary further
recognized that “[a] significantly lower response rate
by non-citizens” or Hispanics “could reduce the accu-
racy of the decennial census.” (Pet. App. 552a.)
But the Secretary then acted arbitrarily and
capriciously in finding that he lacked empirical evi-
dence that adding a citizenship question will dispro-
portionately depress response rates of noncitizen and
Hispanic households. (Pet. App. 552a-557a, 560a-
561a.) That conclusion was contrary to the Adminis-
trative Record, which contains uncontroverted
32
empirical evidence that the question will dispropor-
tionately depress response rates and thus will “harm[]
the quality of the census count” (J.A.105; see Pet. App.
42a-50a, 141a-144a, 285a-286a). See Motor Vehicle
Mfrs. Ass’n of the U.S., Inc. v. State Farm Mut. Auto.
Ins. Co., 463 U.S. 29, 43 (1983). Because “the only
evidence in the record available…actually supports
the opposite conclusion[],” the Secretary’s decision
fails to “satisfy the APA’s reasoned decisionmaking
requirement.” Clark County v. FAA, 522 F.3d 437, 442
(D.C. Cir. 2008) (Kavanaugh, J.).
i. In multiple empirical analyses, the Bureau’s
Chief Scientist and his technical staff informed the
Secretary that adding the citizenship question would
depress the response rate of noncitizen households by
at least 5.1 percentage points more than it would
depress the response rate of citizen households.
(J.A.114; see J.A.104-159, 292, 310.) The analyses
warned that this “cautious estimate” likely did not
reflect the full extent of the actual disproportionate
reduction in response rates caused by a citizenship
question.
7
(J.A.114-116.)
The Bureau reached this conclusion by comparing
the response rates of citizens and noncitizens to the
2010 census, which did not contain a citizenship
question, and the 2010 ACS, which did contain such a
question. The data showed that among citizens, the
drop in response rate from census to ACS was 13.8
percentage points (79.9% to 66.1%), whereas for
noncitizens the drop was 18.9 points (71.5% to 52.6%)
7
The district court found that this differential decline was
large enough to affect, among other things, legislative
apportionment and federal funding. See supra at 21-22.
33
—a differential decrease of 5.1 points for noncitizens
(18.9% minus 13.8%). (J.A.310.)
The Bureau attributed this disproportionate
decline specifically to the citizenship question. That
conclusion was confirmed by empirical analyses
demonstrating that Hispanic households—which
contain a higher proportion of noncitizens—were twice
as likely as non-Hispanic white households to refuse
to answer the citizenship question on the mail-in ACS,
and nine times as likely as non-Hispanic white
households to stop responding to the internet version
of the ACS once they reached the citizenship question.
Because these analyses demonstrated that the citizen-
ship question specifically had a strong deterrent effect
on Hispanic and noncitizen respondents, the Bureau
concluded that the question rather than other factors
(such as general distrust of government or the ACS’s
length) was driving down response rates for those
populations. (J.A.109-112.)
The analyses also established that these
disproportionate reductions in response rates are
significant and will thus “harm the quality of the
census count” (J.A.105)—i.e., reduce its accuracy
(J.A.113). The analyses explained that decreasing
response rates would cause more noncitizen and
Hispanic households who would otherwise self-
respond to enter NRFU instead. (J.A.113-114.) And
because the data on these households produced by
NRFU would be less accurate than these households’
self-responses—in part because NRFU responses may
come from a proxy, such as a landlord, rather than
directly from a household member—increased NRFU
at the expense of self-responses would “reduce the
quality of the resulting data.” (J.A.113-114; see Pet.
App. 153a, 158a-166a.)
34
Finally, the Bureau emphasized that its analyses
of the harms to census accuracy also applied to the
Secretary’s proposal to both add a citizenship question
and use administrative records to generate block-level
citizenship data. (J.A.159.) By contrast, using only
administrative records to generate block-level citizen-
ship data would not harm the enumeration’s accuracy.
(J.A.159.) The Secretary’s assertion that he lacked
empirical evidence about the citizenship question’s
effect on response rates thus runs “counter to the
evidence before” him, State Farm, 463 U.S. at 43.
ii. The Secretary’s attempts to disregard this
evidence (Br. 30-31) cannot withstand scrutiny. He
claimed that the analyses of the relative decline in
response rates for noncitizens compared to citizens
were “inconclusive” (Br. 30; see Pet. App. 554a), but
that characterization cannot be squared with the plain
language of every memorandum prepared by the
Bureau during this time, all of which conclude—
unequivocally—that the citizenship question would
cause a measurable and disproportionate decline in
noncitizen response rates.
8
And extensive trial evidence
confirms this point: Dr. Abowd testified that the
Bureau’s subsequent research showed that the
disproportionate decline in response rates would be
worse than initially projected (J.A.854), and several
other experts likewise testified that the citizenship
question will disproportionately depress noncitizen
and Hispanic response rates (Pet. App. 146a-148a).
8
Indeed, the evidence showed that the differential decline is
getting worse: the disproportionate decline in noncitizen response
rates in 2000 was only 3.3 percentage points, but rose to 5.1
percentage points in 2010. (J.A.110-111.)
35
There is no reasoned basis for the Secretary’s
contention (Pet. App. 553a) that comparisons of the
short-form census questionnaire to the ACS or to the
long-form questionnaire were too “challenging” given
general differences between these instruments. The
Bureau specifically controlled for such differences to
produce strong empirical evidence that it was the
citizenship question, rather than other aspects of the
ACS or long-form questionnaire, that produced
disproportionate declines in noncitizen and Hispanic
response rates. (J.A.111.) The Secretary likewise
misplaced reliance on the assertion that other ACS
questions have overall nonresponse rates purportedly
“comparable” to the citizenship question’s overall
nonresponse rate. See Br. 30. The analyses do not rest
on the overall nonresponse rate to the ACS citizenship
question but rather the difference in nonresponse
rates between noncitizens/Hispanics and citizens/non-
Hispanic whites—a difference that demonstrates that
the citizenship question is causing the dispropor-
tionate depression in response rates.
9
(J.A.109-110.)
The district court thus properly declined to defer to the
Secretary’s “conclusory and unsupported” dismissal,
McDonnell Douglas Corp. v. United States Dep’t of Air
Force, 375 F.3d 1182, 1187 (D.C. Cir. 2004), of the very
empirical evidence that he said would present a
serious concern about the citizenship question’s
9
Indeed, petitioners acknowledged that the Secretary lacked
evidence that the difference in nonresponse rates for the ACS
citizenship question was comparable to the difference in
nonresponse rates for other questions. (PX-297, at 28-29.)
36
“reduc[ing] the accuracy of the decennial census” (Pet.
App. 552a).
10
Disregarding the strong evidence of the
citizenship question’s negative effects on census
accuracy was particularly arbitrary because there is
no evidence in the Administrative Record at all
“supporting a conclusion that addition of the citizen-
ship question will not harm the response rate.” (Pet.
App. 286a.) The Secretary purported to base his
conclusion on a conversation with the Senior Vice
President of data science from the Nielsen Company,
who told him that the response rate to a privately
operated survey had not declined when Nielsen
“added questions on place of birth” and arrival in the
United States. (Pet. App. 559a.) But the Administrative
Record shows that Nielsen’s survey was not remotely
comparable to the decennial census because Nielsen,
unlike the Bureau, (a) paid survey participants, and
(b) had no obligation to count total population in any
event. (J.A.238-240.) See National Fuel Gas Supply
Corp. v. FERC, 468 F.3d 831, 842 (D.C. Cir. 2006)
(Kavanaugh, J.). (rejecting agency reliance on
examples that had “no bearing” on relevant issue).
Moreover, despite the Secretary’s assertion that
“empirical evidence” from Nielsen supported his
decision (Pet. App. 559a), no such evidence exists in
the Administrative Record (Pet. App. 109a-112a). The
Secretary also purported to rely (Pet. App. 559a) on an
example shared by a former Bureau official about a
10
The Secretary’s criticism of the Bureau’s comparative
analyses was also unreasonable because it was his eleventh-hour
request to evaluate a citizenship question—and his failure to
engage the Bureau for nearly a year beforehand—that forced the
Bureau to rely on such analyses rather directly testing the
question’s impact on responses. See infra at 37-42.
37
prior controversial Bureau decision to share data with
another agency, but that example did not involve an
alteration to the decennial census (J.A.236)—and the
same official informed the Secretary that “asking a
citizenship question on the Decennial Census would
diminish response rates and degrade the quality of
responses” (J.A.235). The Secretary’s conclusion is thus
arbitrary because he “provided absolutely no evidence
to back it up.” Safe Extensions, Inc. v. FAA, 509 F.3d
593, 605 (D.C. Cir. 2007).
b. The Secretary’s reliance on a
purported lack of information was
unreasoned when he abandoned
testing procedures.
Even if the Secretary could reasonably conclude
that he lacked conclusive empirical evidence about the
citizenship question’s effect on response rates (Pet.
App. 554a), he still acted arbitrarily and capriciously
in adding the question without conducting any of the
established testing procedures that are designed to
provide him with precisely such empirical evidence.
See State Farm, 463 U.S. at 52 (requiring reasoned
explanation for changing course without “engaging in
a search for further evidence”). The policy underlying
these rigorous testing procedures is a conservative
one: that a survey as large, complex, and important as
the decennial census should not be altered without a
firm understanding of the effects of any such change.
Thus, uncertainty about the effects of changing the
questionnaire is itself a compelling reason not to make
the change. But without even a “minimal level of
analysis,” Encino Motorcars, LLC v. Navarro, 136 S.
Ct. 2117, 2125 (2016), the Secretary repudiated nearly
sixty years of bipartisan and scientific consensus
38
about the way to evaluate and achieve an accurate
decennial census.
i. As the Administrative Record makes clear, and
trial evidence confirms, the Secretary failed to follow
the “well-proven multi-year” testing process that
traditionally governs the decennial census question-
naire.
11
(J.A.204; see J.A.278-280, 597-600, 887-892.)
These procedures are not mere technicalities. They
ensure that the enumeration and other census data
are as accurate as possible by evaluating whether any
proposed change to the census questionnaire, however
minor, would drive down response rates. (J.A.204-205,
626-628.) And the Bureau has consistently declined to
make changes—including to the questionnaire—when
testing has demonstrated reductions in response rates
even less severe than the Bureau found here. See
supra at 8 (3.4% reduction).
The Secretary ignored these procedures here,
adding the citizenship question without subjecting it
to any testing. (Pet. App. 100a-101a; J.A.204-205.)
Moreover, the Secretary had ample opportunity to
conduct at least some testing (Pet. App. 546a), and
indeed was presented with the Bureau’s proposal to
conduct a test that would have further “isolate[d]” the
effects of the citizenship question on response rates—
but declined to do so (Pet. App. 554a; Trial Tr. 1001-
1004). The Secretary’s stark departure “from decades-
11
The Bureau’s empirical analyses, while persuasive evidence
of the likely effect of the question, were no substitute for this
testing process. The analyses were done on the basis of existing
data about past responses to the decennial census and other
surveys. By contrast, testing involves generating new data, such
as by randomized controlled trials or field testing of a specific
question. 1980 Census, supra, at 2-19–2-20.
39
long past practices and official policies” of testing
without any reasoned explanation or even acknowl-
edgment of the change was arbitrary and capricious.
American Wild Horse Pres. Campaign v. Perdue, 873
F.3d 914, 923 (D.C. Cir. 2017); see Encino, 136 S. Ct.
at 2125-26.
Even worse, the Secretary invoked, as part of his
reason for adding the citizenship question, the
purported absence of the very empirical evidence that
the testing process would have produced. Courts
would not defer to the Secretary of Health and Human
Services if he abandoned the Food and Drug Adminis-
tration’s rigorous testing procedures and then approved
a new drug based on a purported lack of evidence. Cf.
Troy Corp. v. Browner, 120 F.3d 277, 293 (D.C. Cir.
1997) (arbitrary and capricious to list chemical as toxic
without following testing guidelines). The district court
properly declined to defer to similarly unreasoned
decision-making here.
ii. Petitioners’ efforts to excuse the Secretary’s
disregard for established testing procedures are
unavailing.
First, petitioners cannot rely on an asserted
“tradition” of a citizenship question (Br. 39) because
there is no tradition comparable to what the Secretary
seeks to do here. No questionnaire mailed to every
household has ever included a citizenship question.
Before 1960, enumerators surveyed households in
person, interviewing individual respondents; and from
1960 onwards, a citizenship question appeared only on
the long-form questionnaire or the ACS sent to a
fraction of the population. The Secretary’s addition of
a citizenship question on the short-form questionnaire
mailed to every household is thus unprecedented.
40
More fundamentally, the era when a citizenship
question was sometimes asked (in person) of every
resident was a time before testing procedures existed,
before sampling and administrative records were
shown to provide useful demographic data without
harming the enumeration’s accuracy, and before it
was understood that questions about sensitive topics
could reduce the accuracy of the enumeration. Indeed,
prior censuses asked many questions that today would
be rejected as likely to deter certain people from
participating, including questions about slaves,
Wright & Hunt, supra, at 154; “grown daughters who
assist in the household duties,” id. at 190; and
“[i]diots,” including their head size, id. at 200. But
shortly after improved statistical methodologies
became widely accepted, the Census Bureau removed
citizenship (and many other topics) from the questions
asked of every resident. And since the application of
better statistical science and more robust testing
procedures, Commerce and the Bureau have strongly
opposed adding a citizenship question to the short-
form questionnaire because it will harm the enume-
ration’s accuracy, and because citizenship data is
available from other sources. Simply stating that a
citizenship question (like many others) was asked
before 1960 under dramatically different circum-
stances thus does not come close to satisfying the
Secretary’s obligation to provide a reasoned explana-
tion for reversing decades of considered agency
judgment. See Encino, 136 S. Ct. at 2125-26; Judulang
v. Holder, 565 U.S. 42, 61 (2011).
Second, petitioners are mistaken in asserting (Br.
39-40) that the Secretary did follow testing proce-
dures. Notwithstanding the Secretary’s assertion (Pet.
App. 550a), testing a question for use on the ACS and
41
long-form questionnaire does not qualify it for use on
the decennial census. (Pet. App. 305a-308a.) The
Administrative Record—including a letter from a
bipartisan group of former Bureau Directors—
demonstrates that these other instruments differ in
scope and kind from the short-form questionnaire
mailed to every household, and that the testing done
for those instruments is no substitute for the proven,
multi-year testing process applicable to the decennial
census questionnaire specifically. (J.A.194-196, 204-
205.) Indeed, the Bureau’s own Statistical Quality
Standards anticipate that questions will be tested in
the context of the specific survey on which they
appear. (J.A.626-631.)
Moreover, even if testing for these other surveys
were relevant, petitioners incorrectly assert (Br. 40)
that the citizenship question “performed adequately”
on those surveys (J.A.627). To the contrary, the Admin-
istrative Record demonstrates that the citizenship
question has not performed adequately on the ACS (or
the long-form questionnaire) because noncitizens
“inaccurately mark ‘citizen’ about 30 percent of the
time.” (Pet. App. 555a; see J.A.147.) Petitioners miss
the mark in relying on Dr. Abowd’s statement that the
Bureau “would accept” the testing performed on the
ACS citizenship question. (J.A.108.) Dr. Abowd made
clear that the citizenship question has not performed
adequately on the ACS (J.A.117, 930-932), but
explained that he had no better option but to rely on
ACS testing, given “the quality, cost, [and] risk
constraints that [the Bureau was] facing to make this
decision” (Trial Tr. 1108; see Trial Tr. 1290-1291).
Third, the Secretary’s assertion that placing “the
citizenship question last on the decennial census form”
would “minimize” any decrease in response rates (Pet.
42
App. 562a) is pure speculation, given that no evidence
supports it and no testing was done to produce such
evidence. (Pet. App. 288a.) The Secretary’s reliance on
unfounded speculation in place of evidence was thus
arbitrary and capricious. See National Lifeline Ass’n v.
FCC, 915 F.3d 19, 32 (D.C. Cir. 2019).
2. The Secretary’s Reliance on DOJ’s
Purported Need for More Accurate
Citizenship Data Was Arbitrary and
Capricious.
The Secretary’s sole justification for undermining
the accuracy of the 2020 enumeration was that adding
a citizenship question was necessary to provide DOJ
the data it claimed to need for VRA enforcement. (Pet.
App. 562a.) As the district court correctly concluded,
this rationale was both unreasoned and contrary to
the evidence.
a. The Secretary failed to explain why
a citizenship question was necessary
when administrative records would
satisfy DOJ’s request.
DOJ’s December 2017 letter asked the Secretary
to add a citizenship question to resolve specific
concerns with existing ACS data on citizenship. In
response, the Bureau informed both DOJ and the
Secretary that all of these concerns would be addressed,
without adding a citizenship question, by linking highly
reliable administrative records containing citizenship
information to individual census responses. The Secre-
tary rejected this administrative-records approach
(Pet. App. 554a-555a), but he failed to explain why
this approach did not fully resolve DOJ’s purported
problems and thus provide adequate citizenship
43
information for VRA enforcement. That complete lack
of reasoned decision-making violated both the APA
and the Secretary’s statutory obligation to rely on
administrative records “[t]o the maximum extent
possible.” 13 U.S.C. § 6(c); see State Farm, 463 U.S. at
48 (“agency must cogently explain” decision).
In its letter, DOJ raised specific objections about
the citizenship data then available. DOJ claimed that
the adoption of the ACS in 2010 had deprived DOJ of
the citizenship data it previously received from the
long-form questionnaire “sent to approximately one in
every six households.” (Pet. App. 566a.) DOJ
complained that, unlike citizenship data from the long-
form questionnaire, ACS citizenship data was not part
of the same database as other decennial-census data,
was not reported at the census-block level, and did not
cover the same timeframe. (Pet. App. 566a-568a).
The Bureau informed both DOJ and the Secretary
that there was an available solution to all of these
concerns that would not require the addition of a
citizenship question. As explained supra at 11-12, the
Bureau proposed linking individual citizenship infor-
mation from the Numident database (and other admin-
istrative records) to the PL94-171 data that DOJ uses
for VRA purposes and that States and their subdivi-
sions use for redistricting. That approach would resolve
all of DOJ’s purported concerns with ACS citizenship
data by producing “block-level tables of citizen voting
age population [CVAP] by race and ethnicity” in the
same database (J.A.291), at the same time, and with
“essentially the same accuracy” as the decennial-
census data DOJ and jurisdictions conducting
redistricting already use (J.A.317; see J.A.107).
44
The Secretary never explained why a citizenship
question would still be necessary given that adminis-
trative records alone would resolve DOJ’s concerns. As
petitioners do not dispute, the administrative-records
approach will provide direct evidence of citizenship
status for about 90% of the population—295 million
people—with sophisticated modeling inferring citizen-
ship for the remaining 10%. (J.A.146.) Petitioners now
argue (Br. 33) that asking a citizenship question will
somewhat improve the accuracy of the information for
the 10% who cannot be linked to administrative
records. But even if that were true—and it is not (see
infra at 45-51)—neither petitioners’ brief nor the
Secretary’s decision memorandum provides any expla-
nation why such an incremental change would make
any meaningful difference for VRA enforcement. In
other words, despite having “staked [his] rationale” on
DOJ’s purported concerns with citizenship data, the
Secretary provided no reasoned explanation or evi-
dence that DOJ would still have any such concerns
under the administrative-records approach. See
National Fuel, 468 F.3d at 843.
DOJ certainly provided no such explanation or
evidence. To the contrary, its letter suggests that DOJ
would have been satisfied with the administrative-
records approach because that approach would produce
data far more accurate than the long-form question-
naire (sent to only one of six households) that the
letter favorably mentions (Pet. App. 566a). And when
the Bureau sought to discuss the administrative-
records approach with DOJ, the Attorney General
forbade any meeting. (J.A.266.) DOJ thus never made
any request for data better than that supplied by the
Bureau’s administrative-records approach; it asked
45
only for data better than that supplied by the ACS, a
request that the Bureau’s approach fully satisfied.
Moreover, the Administrative Record demon-
strates, and trial evidence confirms, that any
incremental increase in citizenship-data accuracy for
just 10% of the population would not make any
meaningful difference given that ACS-derived citizen-
ship data is already sufficient for VRA enforcement
(Pet. App. 295a-297a; see infra at 52-53) and that the
administrative-records approach would significantly
improve that accuracy in any event. Neither the
Secretary nor DOJ nor petitioners have identified a
single VRA case that DOJ failed to bring or lost
because of the absence of whatever supposed marginal
improvement a citizenship question might contribute.
The Secretary’s decision to add a citizenship question
nonetheless, for reasons divorced from any concrete
connection to DOJ’s stated interests, was arbitrary
and capricious. Cf. National Fuel, 468 F.3d at 843
(“Professing that an order ameliorates a real industry
problem but then citing no evidence demonstrating
that there is in fact an industry problem is not reasoned
decisionmaking.”)
b. The Secretary irrationally chose an
approach that would produce less
accurate citizenship information
than using administrative records
alone.
In any event, the Secretary’s conclusion that a
citizenship question would produce more accurate
information than the administrative-records approach
(Pet. App. 556a) was directly contrary to the evidence
before him. As the district court found, “all of the
relevant evidence before Secretary Ross—all of it—
46
demonstrated that using administrative records”
alone will “actually produce more accurate block-level
CVAP data than” using both a citizenship question
and administrative records. (Pet. App. 290a.)
Under the administrative-records approach, the
Bureau would link 295 million census respondents
(out of 330 million total) to administrative records
containing citizenship information, and use sophisti-
cated modeling to determine the citizenship of the
remaining 35 million census respondents for whom
administrative records cannot be linked (“unlinked
respondents”). (Pet. App. 54a-55a; J.A.146.) The
Bureau explained that this modeling would be very
accurate. (J.A.106, 135-136, 146.)
The Secretary did not contest that, where
administrative records exist, they provide extremely
reliable evidence of citizenship status. (Pet. App. 554a-
555a.) But he reasoned that adding a citizenship
question would produce additional information about
citizenship among the 35 million unlinked respondents,
in the form of 22.2 million direct responses to the
question from that group. (Pet. App. 56a, 555a-556a.)
Petitioners now argue that “logic” (Br. 32) compels the
conclusion that more data from census responses
about this group’s citizenship status is better than
less. But that argument is directly contradicted by the
Administrative Record, which shows that adding a
citizenship question will in fact impair the Bureau’s
use of administrative records without contributing any
meaningful additional information about citizenship
status—thus making the net effect of using both
methods less accurate than using administrative
records alone. (Pet. App. 57a.)
47
i. This impairment derives from several factors.
First, the Administrative Record makes clear that
asking the citizenship question will result in inaccurate
responses that the Bureau must accept. Survey
responses to the citizenship question are “highly
suspect.” (Pet. App. 291a.) As the Secretary acknowl-
edged based on responses to the ACS citizenship
question (Pet. App. 560a), noncitizens inaccurately
respond as citizens “at a very high rate” (J.A.120)—
“often more than 30%” (J.A.117). The Bureau thus
found here that a citizenship question would generate
9.5 million responses that conflict with those respon-
dents’ citizenship data in administrative records, and
are likely incorrect given the high accuracy of such
records. (J.A.148, 294.) And the question would
generate another 500,000 responses that cannot be
matched to administrative records but are also likely
inaccurate given the general unreliability of survey
responses on citizenship. (J.A.148, 150, 157.) Adding a
citizenship question would thus “create a problem that
would not exist” under the administrative-records
approach. (Pet. App. 57a; see J.A.157.)
Petitioners attempt to minimize the significance of
this error by asserting, for the first time on appeal,
that “nothing prevents” the Secretary from discarding
the 9.5 million citizenship-question responses that
conflict with administrative records and using the
records’ citizenship information instead. Br. 34. But
petitioners are mistaken. This newly described option
was not part of the Secretary’s analysis, and was not
even presented to the district court, so it cannot be
invoked as a post hoc rationale. See SEC v. Chenery
Corp., 318 U.S. 80, 87-88 (1943). Moreover, this
modification of long-settled procedures has not itself
been fully analyzed or tested. The Administrative
48
Record shows that the “long established” practice of
the Bureau is to accept all responses given to the
decennial census questionnaire, and to resort to
administrative records only when no response is
provided. (J.A.130.) See Evans, 536 U.S. at 466-67.
Only this established approach has undergone testing,
whereas the Bureau has not even begun “explor[ing]
the possibility of checking or changing” responses to a
census question (J.A.131)—a technique that might
open the door to further errors or political manipu-
lation. The accuracy of the census is too important to
be governed by last-minute adjustments by appellate
counsel.
Second, in addition to leading to millions of
inaccurate responses, adding a citizenship question
will also reduce the number of responses by causing a
net undercount of noncitizen and Hispanic house-
holds, which the Bureau estimated at 6.5 million
people. (Pet. App. 152a, 169a-171a.) See supra at 21-
22. Because the enumeration will miss these people
entirely, the Bureau will have no information about
them. This undercount will thus lead not only to the
loss of citizenship information for these individuals,
but also to the loss of the remaining data necessary for
VRA enforcement.
Third, the citizenship question will also increase
by one million the total number of individuals who
cannot be linked to administrative records (J.A.149-
150) because the question would degrade the accuracy
of personal identifying information obtained from the
census, making it harder to match such information to
individuals’ administrative records (J.A.146-147, 158).
The Secretary’s decision memorandum entirely
fails to mention these “important aspect[s] of the
49
problem,” State Farm, 463 U.S. at 43. Given “the lack
of any coherent explanation” to address these major
errors introduced by adding a citizenship question, the
Secretary’s determination does “not satisfy the
reasoned decisionmaking requirement,” Clark County,
522 F.3d at 443.
ii. Petitioners instead rely exclusively on the
purported benefit of obtaining 22.2 million census
responses on citizenship from individuals whose citi-
zenship is not otherwise established by administrative
records. Whereas reliance on administrative records
alone would require modeling to determine the citizen-
ship status of 35 million individuals for whom adminis-
trative records cannot be linked, petitioners assert
(Br. 33) that it is “an obvious improvement” to obtain
responses for 22.2 million of those individuals and
therefore model for only 13.8 million of them. But there
are several fundamental flaws in petitioners’ claim.
First, the Administrative Record demonstrates
that the 22.2 million responses to the citizenship
question will be less accurate than the information
produced by modeling. (J.A.146-148.) These responses
will include both self-responses to the citizenship
question, and responses derived from in-person visits
and proxies. But as explained supra at 47, noncitizens
often self-respond to a citizenship question by inaccu-
rately claiming to be citizens—a problem that will
introduce at least 500,000 inaccurate responses into
the 22.2 million unlinked responses. (J.A.148.) And in-
person visits and proxy responses will introduce
additional errors beyond these incorrect self-responses,
such as when proxies inaccurately describe the
number or citizenship status of household members.
(J.A.157-158.) By contrast, data from administrative
records does not contain such errors and thus allows
50
highly accurate modeling for individuals whose
citizenship status is not reflected in such records.
(J.A.135-136, 146.)
Second, even setting aside the peculiar sensitivities
of the citizenship question, petitioners are wrong to
assume (Br. 33) that survey responses are “obvious[ly]”
more reliable than other statistical methods. This
Court has recognized that statistical estimates may
produce results at least as accurate as direct survey
questions. See Evans, 536 U.S. at 472; cf. House of
Representatives, 525 U.S. at 348 (Scalia J., concurring
in part) (Framers “must have…known that various
methods of estimating unreachable people would be
more accurate than assuming that all unreachable
people did not exist”). Congress reached a similar
judgment when, in the 1976 Act, it directed the
Secretary to use sampling, rather than census
questions, whenever feasible to collect data other than
the enumeration. See 13 U.S.C. §§ 141(a), (d), 195. The
traditional use of the decennial census questionnaire
rather than other techniques, such as sampling and
modeling, thus derives not from any inherently
superior accuracy of questionnaire responses but
rather from the unique requirements applicable to the
actual enumeration. See House of Representatives, 525
U.S. at 343.
Third, the addition of a citizenship question will
substantially diminish the accuracy of any modeling.
As petitioners concede (Br. 34), modeling for the 13.8
million (with a citizenship question) will be worse than
modeling of the 35 million (without the citizenship
question) because the underlying data on which the
model is based will be inherently less accurate.
(J.A.148, 150; Pet. App. 291a.)
51
Because the “record evidence actually undermines”
the Secretary’s statement that more accurate citizen-
ship data would be produced from using both a
citizenship question and administrative records, the
Secretary’s decision to ask the question nonetheless
was arbitrary and capricious. See Clark County, 522
F.3d at 443 n.2; New England Coal. on Nuclear
Pollution v. Nuclear Regulatory Comm’n, 727 F.2d
1127, 1130-31 (D.C. Cir. 1984) (Scalia, J.).
3. The Secretary Failed to Explain Why
He Was Prioritizing Citizenship Data
over the Enumeration.
In addition to irrationally assessing the harm to
the enumeration and the benefits to citizenship data
attributable to a citizenship question, the Secretary
also arbitrarily concluded that providing DOJ with
allegedly “more accurate” citizenship data “outweighs”
and “is of greater importance than any adverse effect
that may result” from the citizenship question (Pet.
App. 562a (emphasis added)). Assuming the Secretary
can choose to sacrifice the accuracy of the decennial
census for some other nonconstitutional objective, he
must still provide a reasoned explanation for why the
benefits gained are worth the harms incurred given
the constitutional, statutory, and practical importance
of pursuing an accurate enumeration. See Burlington
Truck Lines, Inc. v. United States, 371 U.S. 156, 168
(1962). The Secretary failed to do so here.
Discretion to balance harms and benefits is not
unfettered. Rather, an agency must rationally account
for the relative importance that the Constitution or
Congress has assigned to certain factors. Moreover,
when, as here, the agency purports to act in service of
some real-world benefit (improved VRA enforcement)
52
and is faced with credible claims of real-world harm
(inaccurate enumeration), it must reasonably assess
the practical effects of the balance it strikes. See State
Farm, 463 U.S. at 55 (agency must “bear in mind that
Congress intended safety to be the preeminent
factor”); Center for Biological Diversity v. EPA, 722
F.3d 401, 414 (D.C. Cir. 2013) (Kavanaugh, J.,
concurring) (agency “not permitted to substitute its
view of the costs and benefits of regulation for
Congress’s view”). Merely identifying each side of the
balance does not provide a reasoned explanation for
choosing one over the other. Cf. Michigan, 135 S. Ct.
at 2707 (“One would not say that it is even
rational…to impose billions of dollars in economic
costs in return for a few dollars in health or
environmental benefits.”). The Secretary thus
“need[ed] to explain” why a theoretical improvement
to citizenship-data accuracy “justifies” concrete harms
to the enumeration. National Fuel, 468 F.3d at 844.
The Secretary provided no such explanation. He
made no findings whatsoever about the “broader, real-
world impact” of either side of his purported balan-
cing. See American Wild Horse, 873 F.3d at 931. (Pet.
App. 294a-297a.) As to harm to the enumeration, the
Secretary focused only on potential impacts on
response rates (Pet. App. 556a) but nowhere considered
the severe and irreparable consequences of even a
small decrease to the distributional accuracy of the
enumeration, including the loss of congressional seats
and federal funding. See supra at 21-22.
The Secretary also made no findings about the
concrete benefits of improving the accuracy of
citizenship data for VRA purposes. Contrary to
petitioners’ assertion (Br. 30), it is not self-evident
that incrementally more complete and accurate
53
citizenship data would have any measurable effect on
DOJ’s ability to enforce the VRA. See supra at 42-45.
To the contrary, the Administrative Record shows that
the absence of a citizenship question since the VRA’s
enactment in 1965 has not hampered DOJ’s or private
advocates’ efforts to enforce the VRA given the
availability of ACS-derived citizenship data.
12
(Pet.
App. 295a-296a n.71; J.A.193-196, 397-400, 269-272;
see also Dep. of Pamela Karlan 49-53, 66, California,
No. 18-cv-1865, N.D. Cal. ECF:145; id. at 59 (no cases
that plaintiffs “could bring and win if they had”
citizenship data from the census “that they can’t bring
and win now”).) Even if ACS data were not sufficient,
the Secretary never explained why the improvements
of the Bureau’s recommended administrative-records
approach would not suffice. And the Secretary also
failed to address the practical impact of the Bureau’s
disclosure-avoidance protocols, which introduce
margins of error into data shared with other agencies
to protect census respondents’ privacy. (Pet. App.
297a-299a.)
The Secretary thus failed to provide any
explanation for why a theoretical improvement to the
12
The trial evidence confirms that DOJ does not need more
accurate citizenship data than the ACS provides—let alone more
accurate data than administrative records would provide. Dr.
Lisa Handley testified that her work on VRA matters has never
been impeded by using ACS-derived citizenship data, and that
she is unaware of any VRA claim rejected based on shortcomings
with the ACS. (J.A.797-802.) And Gore admitted that he did not
know of any case questioning the adequacy of ACS citizenship
data, or any changes in the law or statistical science supporting
his request for a citizenship question. (J.A.1024-1025, 1079-1086,
1105-1109.)
54
accuracy of citizenship data “outweighs” or is “of
greater importance than” the enormous practical
consequences of harming the accuracy of the enume-
ration (Pet. App. 562a). His conclusory balancing of
the harms and benefits of a citizenship question does
not constitute reasoned decision-making.
13
4. The Secretary’s Rationale Was Pretextual.
a. The district court also properly found the
Secretary’s decision to be arbitrary and capricious
because it relied on a pretextual rationale. The Secre-
tary purported to defer to a genuine, independent
request by DOJ for additional citizenship data to
improve VRA enforcement. But in fact, the Secretary
had already decided to add a citizenship question
before receiving DOJ’s request. And petitioners never
disclosed—until this litigation—that it was the
Secretary and his staff who provided the VRA-
enforcement rationale to DOJ and then worked closely
with DOJ to draft the December 2017 letter articu-
lating that rationale. The Secretary’s rationale for
adding a citizenship question thus misrepresented
that he was deferring to DOJ’s expert judgment, when
in fact the decision was driven by the Secretary and
his staff.
Settled principles of administrative law foreclose
any deference when a decision-maker falsely claims to
rely on the expertise of another agency to defend its
13
DOJ’s December 2017 letter is no substitute for such
reasoning. DOJ did not conduct the balancing that the Secretary
purported to do here, and in any event the Secretary cannot
blindly defer to another agency’s request to add a question to the
decennial census. See Delaware Dep’t of Nat. Res. & Envtl.
Control v. EPA, 785 F.3d 1, 16 (D.C. Cir. 2015).
55
determination. First, the APA requires an agency to
disclose the actual rationale for its action so that the
reviewing court may understand “the basis on which
the [agency] exercised its expert discretion.” Burlington,
371 U.S. at 167; see S. Rep. 79-752, at 15 (1945)
(agency must “explain the actual basis” of its rules).
As petitioners conceded below, presenting a false
rationale for a decision would violate the APA. (Pet.
App. 312a-313a.)
Second, a decision-maker acts arbitrarily by
purporting to rely on another agency’s expertise when,
in fact, the decision-maker instructed that agency
rather than the other way around. Such illusory
reliance undercuts the foundational premise for
judicial deference to administrative action: that the
decision resulted from an exercise of specialized
expertise that courts lack. See State Farm, 463 U.S. at
54. When a decision-maker purports to rely on an
exercise of expert judgment that never happened,
there is nothing to which the courts can defer.
The Secretary’s decision to add a citizenship
question transgressed both these principles. While the
Secretary represented that he was responding to
DOJ’s December 2017 request, the Administrative
Record demonstrates that the Secretary decided to
add a citizenship question months before that request,
as the district court explained in detail. (Pet. App.
118a-129a.) The Secretary and his staff then engaged
in extensive discussions—both internally and with
outside parties—that presumed the decision to add a
citizenship question had already been made, and
turned to how best to “execut[e]” that decision.
(J.A.402.) Absent from these discussions was “any
mention, at all, of VRA enforcement.” (Pet. App. 313a.)
56
The Administrative Record demonstrates that the
Secretary and his staff also went to “extraordinary
lengths” (Pet. App. 318a) to find any other agency to
request the citizenship question and thus provide
“cover for a decision” that had already been made (Pet.
App. 124a). In particular, after both DOJ and DHS
initially refused to request a citizenship question
(J.A.414), the Secretary personally called the Attorney
General (J.A.252-253, 281-282), leading the Attorney
General’s Chief of Staff to assure Commerce that DOJ
would “do whatever you all need us to do” (J.A.254).
And after the Secretary’s staff provided “DOJ with the
[VRA] rationale” (Pet. App. 121a), Gore drafted DOJ’s
letter.
The trial evidence reinforces the district court’s
findings from the Administrative Record. For instance,
a key member of the Secretary’s staff “all but
admit[ted] that Secretary Ross had made up his mind
to add the citizenship question in the spring of 2017,”
and that his task in soliciting support from other
agencies was to “‘find the best rationale’ to support”
that predetermined result. (Pet. App. 314a.) Gore
admitted that he drafted DOJ’s letter solely in response
to the Secretary’s request and principally based on
Commerce’s written work product and advice, rather
than any expertise of DOJ staff. (J.A.1077-1078, 1114-
1115; Pet. App. 125a.) And Gore admitted not knowing
whether a citizenship question would result in
citizenship data more accurate than the data DOJ
already uses. (J.A.1100-1103.)
The district court thus properly found that DOJ’s
letter reflected the Secretary’s assertions about the
VRA rationale rather than any independent judgment
by DOJ. But courts should not defer to the Secretary
of Commerce’s judgment about VRA enforcement. Cf.
57
National Fuel, 468 F.3d at 843 (no deference to Federal
Trade Commission’s report that “relied largely on
[Federal Energy Regulatory Commission’s] assertions,
not the FTC’s independent examination”).
b. Petitioners offer no persuasive answer to the
district court’s factual findings or legal reasoning.
Instead, they repeatedly mischaracterize the nature of
the Secretary’s decision, as the district court found.
For example, petitioners argue (Br. 41-42) that the
district court faulted the Secretary merely for having
“additional” reasons beyond the purportedly “rational
and supported” reason he gave. But the district court
did no such thing; it found that the sole reason the
Secretary provided—deference to DOJ’s independent
judgment about its VRA-enforcement needs—was
neither rational nor supported because DOJ did not
exercise independent judgment, and the VRA rationale
was inadequate. Petitioners similarly argue (Br. 43)
that the Secretary merely had an “inclin[ation]
towards a certain policy position” when he “reached
out to DOJ to ask if it would support that policy.” But
the court found that the Secretary had already decided
to add the citizenship question, manufactured the
VRA-enforcement rationale, provided it to DOJ to
present as its own, and then made it appear as if DOJ
had independently exercised judgment to request
citizenship data. (Pet. App. 120a-121a.) These circum-
stances, along with other evidence, supported the
district court’s finding that the rationale given by the
Secretary was pretextual.
c. There is no basis for petitioners’ contention (Br.
42) that a finding of pretext requires evidence that the
Secretary subjectively disbelieved the stated grounds
for the decision, irreversibly prejudged the decision, or
58
was otherwise driven by some legally forbidden motive.
Such evidence could support a finding of bad faith that
renders an agency decision arbitrary and capricious.
See Woods Petroleum Corp. v. United States Dep’t of
Interior, 18 F.3d 854, 859-60 (1994), adhered to on
reh’g en banc, 47 F.3d 1032 (10th Cir. 1995). But an
agency’s decision is also arbitrary when it recites a
false rationale, or purports to rely on a nonexistent
exercise of expert judgment—for instance, if an agency
were to claim that its decision was based on studies
that had never been conducted.
In any event, the district court’s pretext finding
would be supported even if such a finding required
proof that the Secretary had an unalterably closed
mind or subjectively disbelieved the VRA-enforcement
rationale. Ample evidence demonstrated that the
Secretary had “decided to add the question for reasons
entirely unrelated to VRA enforcement well before he
persuaded DOJ” to send its letter. (Pet. App. 318a.)
And evidence likewise demonstrated that the Secre-
tary did not believe the VRA-enforcement rationale,
such as evidence that the Secretary urged officials who
lacked VRA-enforcement responsibilities to request
the citizenship question. (Pet. App. 120a-121a.) The
district court thus properly found that any
presumption of regularity was rebutted under the
exceptional facts presented here. See United States v.
Armstrong, 517 U.S. 456, 470 (1996).
59
C. The Secretary’s Decision Was
Contrary to Law.
The district court correctly concluded that the
Secretary’s decision violated two statutes.
1. The Secretary violated 13 U.S.C. § 6(c) by
adding a citizenship question to the decennial census
even though administrative records would provide
block-level citizenship data sufficient to satisfy DOJ’s
purported VRA-enforcement needs. Section 6(c)
requires that “[t]o the maximum extent possible and
consistent with the kind, timeliness, quality and scope
of statistics required,” the Secretary must “acquire
and use information available” from administrative
records “instead of conducting direct inquiries.” The
Secretary violated this statutory mandate here
because, as explained supra at 43-45, administrative
records will provide DOJ with block-level citizenship
data of the “kind, timeliness, quality and scope” that
it purported to need.
14
Petitioners’ contrary arguments are unavailing.
First, petitioners claim (Br. 45-46) that § 6(c) leaves to
the Secretary the “policy choice” to use census ques-
tions whenever he deems administrative records to be
“incomplete.” But that argument cannot be squared
with the statute’s mandatory language, which directs
that the Secretary “shall” use administrative records,
shall do so “instead of conducting direct inquiries,” and
shall do so “[t]o the maximum extent possible.”
14
Section 6(c) thus provides a standard for reviewing the
Secretary’s decision (see supra at 28); a basis to find the decision
arbitrary and capricious (see supra at 43-45); and a ground to find
an independent statutory violation.
60
Congress thus left “the Secretary no room to choose.”
(Pet. App. 266a.)
Moreover, mere “gaps in [the] data” about
citizenship from administrative records (Br. 46)
cannot justify the Secretary’s choice to burden the
entire population with a citizenship question when
administrative records will produce direct citizenship
data for 295 million people and will produce highly
reliable citizenship data for the remaining 35 million
through modeling. (J.A.149.) Petitioners’ contrary
argument mistakenly presumes that § 6(c) has no
application if administrative records do not on their
face contain information. That interpretation makes
little sense given the context and purpose of the 1976
amendments to the Census Act, which expressly
prioritized the use of sampling and other statistical
techniques to extrapolate information for the entire
population based on information about a subset of the
population. See supra at 6-7. Congress’s command
that the Secretary use administrative records “[t]o the
maximum extent possible” “instead of conducting
direct inquiries” thus cannot reasonably be read to
excuse the Secretary from using statistical methods
based on such records in place of the decennial census
to derive information about the whole population.
Second, contrary to petitioners’ assertion (Br. 46-
47), the presence of a citizenship question on the long-
form questionnaire before 1976 does not remotely
suggest that Congress intended to exempt a citizen-
ship question from § 6(c). In 1976, the use of adminis-
trative records, sampling, and modeling was relatively
new. Congress fully expected that these methods could
displace the then-current uses of the decennial census
questionnaire to obtain demographic data. See H.R.
Rep. 92-1288, at 15-16 (1972); Mid-Decade Census,
61
Hr’g Before the H. Subcomm. on Census & Statistics 7
(1967). Indeed, by 1976, the citizenship question had
already been removed from the short-form question-
naire distributed to all households and placed on the
long-form questionnaire—a sample survey that used
modeling to generate citizenship data for the vast
majority of residents. Section 6(c) merely continued a
development that had already begun.
Third, there is no merit to petitioners’ assertion
(Br. 47) that finding a § 6(c) violation here will invali-
date “every demographic question” on the decennial
census. Whether administrative records will suffice in
a given case is a fact-specific inquiry that depends on
the reliability and availability of administrative
records for the particular type of information required,
and the ease of linking those records to census
responses. Here, the district court’s finding of a § 6(c)
violation properly rested on the substantial and
uncontroverted evidence that administrative records
alone would provide highly reliable information about
citizenship status specifically.
2. The district court also correctly set aside the
Secretary’s decision as contrary to 13 U.S.C. § 141(f).
That provision precludes the Secretary from altering
the census subjects that he previously reported to
Congress unless he “finds new circumstances exist
which necessitate” such a change and submits a new
report. 13 U.S.C. § 141(f)(1), (3). The Secretary violated
this provision because he failed to include citizenship
as a census subject in his initial report under
§ 141(f)(1), and then failed to issue a new report or
issue any findings that “new circumstances” warranted
adding citizenship as a subject, as required by
§ 141(f)(3). (Pet. App. 274a.) The Secretary’s inclusion
of a citizenship question in his separate report of such
62
questions under § 141(f)(2) (Br. 52-53) did not satisfy
the statute’s distinct requirements to separately report
census subjects under § 141(f)(1) and to report specific
findings justifying any change in such subjects under
§ 141(f)(3). (Pet. App. 275a-276a.)
Petitioners are incorrect in arguing (Br. 49-51)
that only Congress may enforce § 141(f). As the
district court explained (Pet. App. 276a-284a), § 141(f)
is unlike the purely informational reporting
requirements in the cases on which petitioners rely
because the provision here imposes a substantive
constraint on the Secretary’s ability to surprise
Congress or the public by altering the subjects of the
decennial census belatedly, and without making
findings to justify the change.
III. THE SECRETARY’S DECISION VIOLATED
THE
ENUMERATION CLAUSE.
The Secretary’s decision also violated the
Enumeration Clause. This Court need not address this
constitutional claim if it holds that the Secretary’s
action violates the APA. See Califano v. Yamasaki,
442 U.S. 682, 692 (1979). But if the Court reaches the
constitutional claim, it should affirm the judgment
below on this alternative ground. See United States v.
New York Tel. Co., 434 U.S. 159, 166 n.8 (1977).
A. The Enumeration Clause requires an “actual
Enumeration” of the population every ten years that
must be conducted by “counting the whole number of
persons in each State.” U.S. Const. art. I, § 2, cl. 3; id.
amend. XIV, § 2. While this provision leaves substan-
tial discretion to Congress (or its delegate, the
Secretary) to determine the “methodological details” of
conducting the required headcount, that discretion is
63
constrained by the “strong constitutional interest in
accuracy,” Evans, 536 U.S. at 474, 478, with a
“preference for distributive accuracy,” Wisconsin, 517
U.S. at 20. Thus, decisions by Congress or the Secretary
about the conduct of the decennial enumeration must
bear “a reasonable relationship to the accomplishment
of an actual enumeration of the population.” Id.
The history and purpose of the Enumeration
Clause confirm the central importance of accuracy as
a limiting constitutional principle. The Framers delib-
erately chose the objective measure of total population
as the relevant constitutional metric to avoid the use
of the census for political manipulation. Evans, 536
U.S. at 478; see id. at 503 (Thomas, J., concurring &
dissenting in part) (Framers’ “principal concern was
that the Constitution establish a standard resistant to
manipulation”); cf. Evenwel, 136 S. Ct. at 1142 (tally
of “total population” is “more reliable and less subject
to manipulation and dispute than statistics concerning
eligible voters”). This anti-manipulation purpose would
be severely undermined if, as petitioners argued
below, the Enumeration Clause would be satisfied by
any “person-by-person headcount,” however poorly
planned or implemented. (S.D.N.Y. ECF:155 at 30.)
Under that extreme interpretation, petitioners could
conduct a census that dramatically and foreseeably
undermines the enumeration’s accuracy because of
“bias, manipulation, fraud or similarly grave abuse,”
City of Philadelphia v. Klutznick, 503 F. Supp. 663,
675 (E.D. Pa. 1980), without transgressing any
constitutional line. But such a deviation from the
objective goal of an accurate headcount was “exactly
the type of conduct and temptation the Framers
wished to avoid.” Id.
64
B. Here, the Secretary’s decision to add a
citizenship question to the 2020 census violated the
Enumeration Clause because it would affirmatively
undermine the accuracy of the enumeration. See also
California, 2019 WL 1052434, at *2, *67-69 (finding
violation of Enumeration Clause because inclusion of
a citizenship question “affirmatively interferes with
the actual enumeration and fulfills no reasonable
government purpose”). As explained supra at 21-22,
the evidence here demonstrates that the addition of a
citizenship question would lead to a differential
undercount severe enough to cause several States to
lose congressional seats, among other injuries. Such
loss of representation is precisely the type of injury
that the Enumeration Clause was designed to prevent.
See, e.g., Franklin, 505 U.S. at 790; Department of
Commerce v. Montana, 503 U.S. 442, 445 (1992).
The Secretary cannot defeat the constitutional
claim by asserting that adding a citizenship question
would provide useful data for DOJ’s enforcement of
the VRA. While this Court has deferred to the
Secretary’s judgment about how best to achieve an
accurate enumeration, see Evans, 536 U.S. at 478, his
decision here was not made to improve the accuracy of
the enumeration. Instead, he decided that collecting
data for VRA enforcement “is of greater importance
than any adverse effect [on the enumeration] that may
result” from the citizenship question.
15
(Pet. App.
562a.)
15
Wisconsin’s reference to the Secretary’s “‘virtually unlim-
ited discretion,’” 517 U.S. at 19, does not hold, as petitioners
suggest (Br. 21), that any census-related decision is unreviewable
under the Constitution. In Wisconsin, the Court deferred to the
65
Even if it were permissible for the Secretary to
trade the enumeration’s accuracy for some other policy
objective, the evidence here demonstrates that adding
a citizenship question would not enhance VRA
enforcement. As explained supra at 43-45, the district
court found that the Bureau could provide DOJ the
block-level citizenship data that it claimed to need for
VRA enforcement, without adding a citizenship
question, by linking highly reliable administrative
records containing citizenship information to census
responses. Nothing in the Administrative Record
suggests that the data collected by a citizenship
question would enable more effective VRA enforce-
ment than the data collected from administrative
records. Indeed, there is not even evidence that a
citizenship question would be an improvement over
the citizenship data currently collected by the ACS
(and earlier by the long-form questionnaire). See
supra at 52-53 & n.12. Because there is no evidence
that a citizenship question would provide any
meaningful improvement, the Clause bars the
Secretary from relying on that justification to sacrifice
the accuracy of the enumeration.
C. The district court misconstrued respondents’
constitutional claim as a challenge to any decennial-
census question “unrelated” to the Enumeration
Clause’s goal of conducting a headcount of all
residents—including any demographic question. (Pet.
App. 418a.) But the defect at issue here is not that the
citizenship question is merely “unrelated” to the
Secretary’s judgment about which treatment of census data
would be most accurate for apportionment. 517 U.S. at 20-24. It
did not defer to a decision about whether to pursue accuracy at
all—let alone to undermine accuracy in pursuit of some other
objective.
66
headcount of total population, but rather that adding
this question to the 2020 census would affirmatively
undermine the accuracy of the headcount. And the
proof of this harm derives from the particular circum-
stances of this case, not from some broad-based
challenge to any demographic question. See supra at
21-22. See also California v. Ross, No. 18-cv-1865,
2018 WL 7142099, at *15 (N.D. Cal. Aug. 17, 2018)
(constitutional claim arises from “the effect of asking a
question about citizenship in the context of this
decennial census taking”).
For similar reasons, the district court misplaced
reliance on the fact that the decennial census form has
in the past included demographic questions, including
questions related to citizenship. (Pet. App. 412a-419a.)
The early use of demographic questions on the census,
including questions about citizenship, occurred before
the modernization of the census process provided a
clear scientific understanding of the potential harms
to the enumeration of asking particular questions. See
supra at 3-4. Accordingly, for those prior forms, there
is no indication that the Secretary had before him
concrete and unrebutted evidence that the inclusion of
a particular demographic question would lead to a
materially less accurate headcount, as is the case here.
D. Petitioners asserted below that respondents’
Enumeration Clause claim was a nonjusticiable politi-
cal question. The district court correctly rejected that
argument. (Pet. App. 391a-398a.)
Under the political question doctrine, courts may
not adjudicate a constitutional dispute where there is
“a lack of judicially discoverable and manageable
standards for resolving it,” or where there is “a textu-
ally demonstrable constitutional commitment of the
67
issue to a coordinate political department.” Baker v.
Carr, 369 U.S. 186, 217 (1962). Here, the constitu-
tional interest in accuracy provides a judicially
manageable standard to evaluate the Secretary’s
decision to add a citizenship question. See supra at 62-
63. And contrary to petitioners’ arguments below, the
Constitution does not commit the conduct of the
census entirely to the unreviewable discretion of
either Congress or the Secretary.
The Enumeration Clause provides that the
decennial census shall be conducted “in such Manner
as [Congress] shall by Law direct.” Art. I, § 2, cl. 3. The
district court correctly observed that this Court and
others have never found this language to completely
insulate the Secretary’s conduct of the census from
judicial review. (Pet. App. 392a-393a.) Indeed, peti-
tioners concede (Br. 27) that the Secretary’s decisions
concerning the census are not entirely unreviewable.
In particular, petitioners acknowledged below that
courts may review whether the Secretary is in fact
conducting a “person-by-person headcount of the
population” but contended that courts are powerless to
evaluate “the manner” by which the Secretary conducts
such a headcount. (S.D.N.Y. ECF:155 at 21.)
Petitioners’ dichotomy is “a false one.” (Pet. App.
395a.) The core error in petitioners’ argument below is
that the “manner” of conducting the decennial census
can and does have consequences for whether the
Secretary is in fact conducting a “person-by-person
headcount of the population.” That connection lies at
the heart of both respondents’ Enumeration Clause
and APA claims: respondents alleged, and proved at
trial, that a citizenship inquiry would make the
person-by-person enumeration less accurate.
68
Petitioners’ attempt to label respondents’ constitu-
tional claim as a challenge to the “manner” of conduct-
ing the decennial census thus does not distinguish this
case from the challenges that are indisputably
justiciable under the Enumeration Clause.
IV. THE DISTRICT COURT PROPERLY AUTHORIZED
DISCOVERY.
The entry of final judgment has largely mooted the
parties’ discovery dispute. See No. 18-557 Gov’t Resp.
Br. 26-28. The district court vacated its order
authorizing the Secretary’s deposition (Pet. App. 352a-
353a), and respondents withdrew that deposition
request (S.D.N.Y. ECF:577). And because the Adminis-
trative Record alone supports the district court’s judg-
ment (Pet. App. 260a-261a), this Court may affirm
without resolving whether extra-record discovery was
warranted.
If the Court reaches the question, it should affirm
the district court’s discovery orders. Petitioners
misconstrue (Br. 55) the basis for discovery as an
attempt to “probe the Secretary’s mental processes.”
But discovery was justified to uncover objective facts
about the decision-making process that should have
been disclosed as part of the “whole record” that the
APA requires. 5 U.S.C. § 706; see 18-557 Gov’t Resp.
Br. 30-39. As the district court explained (Pet. App.
126a-129a), petitioners obscured their decision-
making, depriving respondents—and the courts—of
the information that the Secretary “directly or
indirectly” considered. Thompson v. United States
Dep’t of Labor, 885 F.2d 551, 555 (9th Cir. 1989). On
July 3, when the first discovery order was entered,
petitioners had concededly failed to disclose the
Secretary’s deliberations before December 2017; the
69
Secretary had provided contradictory accounts of
those deliberations; and there was evidence that the
Secretary had prejudged the decision to add a
citizenship question and used the VRA-enforcement
rationale as a pretext. See supra at 55-56; 18-557 Gov’t
Resp. Br. 40-48. Given this “strong showing of bad
faith or improper behavior,” Overton Park, 401 U.S. at
420, extra-record discovery was warranted to under-
stand “the basis on which the” Secretary reached his
decision. Burlington, 371 U.S. at 167.
CONCLUSION
For each of these reasons, the Court should affirm
the judgment below.
M
ATTHEW COLANGELO
Chief Counsel for
Federal Initiatives
E
LENA GOLDSTEIN
Acting Bureau Chief
Civil Rights Bureau
Respectfully submitted,
L
ETITIA JAMES
Attorney General
State of New York
B
ARBARA D. UNDERWOOD*
Solicitor General
S
TEVEN C. WU
Deputy Solicitor General
J
UDITH N. VALE
Senior Assistant
Solicitor General
S
COTT A. EISMAN
Assistant Solicitor General
barbara.underwood@ag.ny.gov
April 2019
* Counsel of Record
(Counsel list continues on next page.)
70
P
HIL
W
EISER
Attorney General
State of Colorado
136 State Capitol Bldg.
Denver, CO 80203
T
HOMAS
J.
M
ILLER
Attorney General
State of Iowa
1305 E. Walnut St.
Des Moines, IA 50319
WILLIAM TONG
Attorney General
State of Connecticut
55 Elm St.
P.O. Box 120
Hartford, CT 06106
BRIAN E. FROSH
Attorney General
State of Maryland
200 St. Paul Pl.
20th Floor
Baltimore, MD 21202
KATHLEEN JENNINGS
Attorney General
State of Delaware
Department
of Justice
6th Floor
820 N. French St.
Wilmington, DE 19801
MAURA HEALEY
Attorney General
Commonwealth of
Massachusetts
One Ashburton Pl.
Boston, MA 02108
KARL A. RACINE
Attorney General
District of Columbia
441 4th St., NW
Suite 630 South
Washington, DC 20001
KEITH ELLISON
Attorney General
State of Minnesota
102 State Capitol
75 Rev. Dr. Martin Luther
King Jr. Blvd.
St. Paul, MN 55155
K
WAME
R
AOUL
Attorney General
State of Illinois
100 W. Randolph St.
Chicago
, IL 60601
G
URBIR
S.
G
REWAL
Attorney General
State of New Jersey
25 Market St.
8th Floor, West Wing
Trenton, NJ 08625-0080
71
H
ECTOR
H.
B
ALDERAS
Attorney General
State of New Mexico
408 Galisteo St.
Santa Fe, NM 87501
T
HOMAS
J.
D
ONOVAN
,
J
R
.
Attorney General
State of Vermont
109 State St.
Montpelier, VT 05609
JOSHUA H. STEIN
Attorney General
State of North Carolina
Department of Justice
114 W. Edenton St.
Raleigh, NC 27603
ROBERT W. FERGUSON
Attorney General
State of Washington
800 Fifth Avenue
Suite 2000
Seattle, WA 98104
ELLEN F. ROSENBLUM
Attorney General
State of Oregon
1162 Court St. N.E.
Salem, OR 97301
MATTHEW JERZYK
City Solicitor
City of Central Falls
580 Broad St.
Central Falls, RI 02863
JOSH SHAPIRO
Attorney General
Commonwealth of
Pennsylvania
16th Floor
Strawberry Sq.
Harrisburg, PA 17120
EDWARD N. SISKEL
Corporation Counsel
City of Chicago
30 N. LaSalle St.
Suite 800
Chicago, IL 60602
PETER F. NERONHA
Attorney General
State of Rhode Island
150 South
Main St.
Providence, RI 02903
ZACHARY M. KLEIN
City Attorney
City of Columbus
77 North Front St.
4th Floor
Columbus, OH 43215
72
D
ENNIS
J.
H
ERRERA
City Attorney
City and County of San
Francisco
City Hall, Room 234
1 Dr. Carlton B. Goodlett Pl.
San Francisco, CA 94102
Z
ACHARY
W.
C
ARTER
Corporation Counsel
City of New York
100 Church St.
New York, NY 10007
ROLANDO L. RIOS
Special Counsel
Counties of Cameron and
Hidalgo
110 Broadway
Suite 355
San Antonio, TX 78205
MARCEL S. PRATT
City Solicitor
City of Philadelphia
1515 Arch St.
17th Floor
Philadelphia, PA 19102
JO ANNE BERNAL
County Attorney
County of El Paso
500 E. San Antonio
Room 503
El Paso, TX 79901
CRIS MEYER
City Attorney
City of Phoenix
200 W. Washington
Suite 1300
Phoenix, AZ 85003
CHARLES J. MCKEE
County Counsel
County of Monterey
168 West Alisal St. 3rd Fl.
Salinas, CA 93901
YVONNE S. HILTON
City Solicitor
City of Pittsburgh
414 Grant St.
Room 323
Pittsburgh, PA 15219
J
OHN
D
ANIEL
R
EAVES
General Counsel
U.S. Conference of Mayors
1750 K St., N.W.
11th Floor
Washington, DC 20006
J
EFFREY
D
ANA
City Solicitor
City of Providence
444 Westminster St.
Providence, RI 02903
73
P
ETER
S.
H
OLMES
City Attorney
City of Seattle
701 Fifth Avenue
Suite 2050
Seattle, WA 98104-7097
