Department of Commerce v. New York

(Slip Opinion) OCTOBER TERM, 2018 1

 Syllabus

 NOTE: Where it is feasible, a syllabus (headnote) will be released, as is
 being done in connection with this case, at the time the opinion is issued.
 The syllabus constitutes no part of the opinion of the Court but has been
 prepared by the Reporter of Decisions for the convenience of the reader.
 See United States v. Detroit Timber & Lumber Co., 200 U. S. 321, 337 .

SUPREME COURT OF THE UNITED STATES

 Syllabus

 DEPARTMENT OF COMMERCE ET AL. v. NEW YORK
 ET AL.

 CERTIORARI BEFORE JUDGMENT TO THE UNITED STATES
 COURT OF APPEALS FOR THE SECOND CIRCUIT

 No. 18–966. Argued April 23, 2019—Decided June 27, 2019
In order to apportion congressional representatives among the States,
 the Constitution requires an “Enumeration” of the population every
 10 years, to be made “in such Manner” as Congress “shall by Law di-
 rect,” Art. I, §2, cl. 3; Amdt. 14, §2. In the Census Act, Congress del-
 egated to the Secretary of Commerce the task of conducting the de-
 cennial census “in such form and content as he may determine.” 13
 U. S. C. §141 (a). The Secretary is aided by the Census Bureau, a sta-
 tistical agency in the Department of Commerce. The population
 count is also used to allocate federal funds to the States and to draw
 electoral districts. The census additionally serves as a means of col-
 lecting demographic information used for a variety of purposes.
 There have been 23 decennial censuses since 1790. All but one be-
 tween 1820 and 2000 asked at least some of the population about
 their citizenship or place of birth. The question was asked of all
 households until 1950, and was asked of a fraction of the population
 on an alternative long-form questionnaire between 1960 and 2000.
 In 2010, the citizenship question was moved from the census to the
 American Community Survey, which is sent each year to a small
 sample of households.
 In March 2018, Secretary of Commerce Wilbur Ross announced in
 a memo that he had decided to reinstate a citizenship question on the
 2020 census questionnaire at the request of the Department of Jus-
 tice (DOJ), which sought census block level citizenship data to use in
 enforcing the Voting Rights Act (VRA). The Secretary’s memo ex-
 plained that the Census Bureau initially analyzed, and the Secretary
 considered, three possible courses of action before he chose a fourth
 option that combined two of the proposed options: reinstate a citizen-
2 DEPARTMENT OF COMMERCE v. NEW YORK

 Syllabus

 ship question on the decennial census, and use administrative rec-
 ords from other agencies, e.g., the Social Security Administration, to
 provide additional citizenship data. The Secretary “carefully consid-
 ered” the possibility that reinstating a citizenship question would de-
 press the response rate, the long history of the citizenship question
 on the census, and several other factors before concluding that “the
 need for accurate citizenship data and the limited burden of the ques-
 tion” outweighed fears about a lower response rate.
 Here, two separate suits filed in Federal District Court in New
 York were consolidated: one filed by a group States, counties, cities,
 and others, alleging that the Secretary’s decision violated the Enu-
 meration Clause and the requirements of the Administrative Proce-
 dure Act; the other filed by non-governmental organizations, adding
 an equal protection claim. The District Court dismissed the Enu-
 meration Clause claim but allowed the other claims to proceed. In
 June 2018, the Government submitted the Commerce Department’s
 “administrative record”—materials that Secretary Ross considered in
 making his decision—including DOJ’s letter requesting reinstate-
 ment of the citizenship question. Shortly thereafter, at DOJ’s urging,
 the Government supplemented the record with a new memo from the
 Secretary, which stated that he had begun considering the addition of
 a citizenship question in early 2017 and had asked whether DOJ
 would formally request its inclusion. Arguing that the supplemental
 memo indicated that the record was incomplete, respondents asked
 the District Court to compel the Government to complete the admin-
 istrative record. The court granted that request, and the parties
 jointly stipulated to the inclusion of additional materials that con-
 firmed that the Secretary and his staff began exploring reinstate-
 ment of a citizenship question shortly after his 2017 confirmation, at-
 tempted to elicit requests for citizenship data from other agencies,
 and eventually persuaded DOJ to make the request. The court also
 authorized discovery outside the administrative record, including
 compelling a deposition of Secretary Ross, which this Court stayed
 pending further review. After a bench trial, the District Court de-
 termined that respondents had standing to sue. On the merits, it
 ruled that the Secretary’s action was arbitrary and capricious, based
 on a pretextual rationale, and violated the Census Act, and held that
 respondents had failed to show an equal protection violation.
Held:
 1. At least some respondents have Article III standing. For a legal
 dispute to qualify as a genuine case or controversy, at least one plain-
 tiff must “present an injury that is concrete, particularized, and actu-
 al or imminent; fairly traceable to the defendant’s challenged behav-
 ior; and likely to be redressed by a favorable ruling.” Davis v.
 Cite as: 588 U. S. ____ (2019) 3

 Syllabus

Federal Election Comm’n, 554 U. S. 724, 733 . The District Court con-
cluded that the evidence at trial established a sufficient likelihood
that reinstating a citizenship question would result in noncitizen
households responding to the census at lower rates than other
groups, which would cause them to be undercounted and lead to
many of the injuries respondents asserted—diminishment of political
representation, loss of federal funds, degradation of census data, and
diversion of resources. For purposes of standing, these findings of
fact were not so suspect as to be clearly erroneous. Several state re-
spondents have shown that if noncitizen households are undercount-
ed by as little as 2%, they will lose out on federal funds that are dis-
tributed on the basis of state population. That is a sufficiently
concrete and imminent injury to satisfy Article III, and there is no
dispute that a ruling in favor of respondents would redress that
harm. Pp. 8–11.
 2. The Enumeration Clause permits Congress, and by extension
the Secretary, to inquire about citizenship on the census question-
naire. That conclusion follows from Congress’s broad authority over
the census, as informed by long and consistent historical practice that
“has been open, widespread, and unchallenged since the early days of
the Republic.” NLRB v. Noel Canning, 573 U. S. 513 , 572 (Scalia, J.,
concurring in judgment). Pp. 11–13.
 3. The Secretary’s decision is reviewable under the Administrative
Procedure Act. The APA instructs reviewing courts to set aside agen-
cy action that is “arbitrary, capricious, an abuse of discretion, or oth-
erwise not in accordance with law,” 5 U. S. C. §706 (2)(A), but it
makes review unavailable “to the extent that” the agency action is
“committed to agency discretion by law,” §701(a)(2). The Census Act
confers broad authority on the Secretary, but it does not leave his
discretion unbounded. The §701(a)(2) exception is generally limited
to “certain categories of administrative decisions that courts tradi-
tionally have regarded as ‘committed to agency discretion,’ ” Lincoln
v. Vigil, 508 U. S. 182, 191 . The taking of the census is not one of
those areas. Nor is the statute drawn so that it furnishes no mean-
ingful standard by which to judge the Secretary’s action, which is
amenable to review for compliance with several Census Act provi-
sions according to the general requirements of reasoned agency deci-
sionmaking. Because this is not a case in which there is “no law to
apply,” Citizens to Preserve Overton Park, Inc. v. Volpe, 401 U. S. 402,
410 , the Secretary’s decision is subject to judicial review. Pp. 13–16.
 4. The Secretary’s decision was supported by the evidence before
him. He examined the Bureau’s analysis of various ways to collect
improved citizenship data and explained why he thought the best
course was to both reinstate a citizenship question and use citizen-
4 DEPARTMENT OF COMMERCE v. NEW YORK

 Syllabus

 ship data from administrative records to fill in the gaps. He then
 weighed the value of obtaining more complete and accurate citizen-
 ship data against the uncertain risk that reinstating a citizenship
 question would result in a materially lower response rate, and ex-
 plained why he thought the benefits of his approach outweighed the
 risk. That decision was reasonable and reasonably explained, partic-
 ularly in light of the long history of the citizenship question on the
 census. Pp. 16–20.
 5. The District Court also erred in ruling that the Secretary violat-
 ed two particular provisions of the Census Act, §6(c) and §141(f ).
 Section 6’s first two subsections authorize the Secretary to acquire
 administrative records from other federal agencies and state and lo-
 cal governments, while subsection (c) requires the Secretary, to the
 maximum extent possible, to use that information “instead of con-
 ducting direct inquiries.” Assuming that §6(c) applies, the Secretary
 complied with it for essentially the same reasons that his decision
 was not arbitrary and capricious: Administrative records would not,
 in his judgment, provide the more complete and accurate data that
 DOJ sought. The Secretary also complied with §141(f ), which re-
 quires him to make a series of reports to Congress about his plans for
 the census. And even if he had violated that provision, the error
 would be harmless because he fully informed Congress of, and ex-
 plained, his decision. Pp. 20–23.
 6. In order to permit meaningful judicial review, an agency must
 “ ‘disclose the basis’ ” of its action. Burlington Truck Lines, Inc. v.
 United States, 371 U. S. 156 , 167–169. A court is ordinarily limited
 to evaluating the agency’s contemporaneous explanation in light of
 the existing administrative record, Vermont Yankee Nuclear Power
 Corp. v. Natural Resources Defense Council, Inc., 435 U. S. 519 , but it
 may inquire into “the mental processes of administrative deci-
 sionmakers” upon a “strong showing of bad faith or improper behav-
 ior,” Overton Park, 401 U. S., at 420 . While the District Court pre-
 maturely invoked that exception in ordering extra-record discovery
 here, it was ultimately justified in light of the expanded administra-
 tive record. Accordingly, the District Court’s ruling on pretext will be
 reviewed in light of all the evidence in the record, including the extra-
 record discovery.
 It is hardly improper for an agency head to come into office with
 policy preferences and ideas, discuss them with affected parties,
 sound out other agencies for support, and work with staff attorneys
 to substantiate the legal basis for a preferred policy. Yet viewing the
 evidence as a whole, this Court shares the District Court’s conviction
 that the decision to reinstate a citizenship question cannot adequate-
 ly be explained in terms of DOJ’s request for improved citizenship
 Cite as: 588 U. S. ____ (2019) 5

 Syllabus

 data to better enforce the VRA. Several points, taken together, re-
 veal a significant mismatch between the Secretary’s decision and the
 rationale he provided. The record shows that he began taking steps
 to reinstate the question a week into his tenure, but gives no hint
 that he was considering VRA enforcement. His director of policy at-
 tempted to elicit requests for citizenship data from the Department of
 Homeland Security and DOJ’s Office of Immigration Review before
 turning to the VRA rationale and DOJ’s Civil Rights Division. For its
 part, DOJ’s actions suggest that it was more interested in helping the
 Commerce Department than in securing the data. Altogether, the ev-
 idence tells a story that does not match the Secretary’s explanation
 for his decision. Unlike a typical case in which an agency may have
 both stated and unstated reasons for a decision, here the VRA en-
 forcement rationale—the sole stated reason—seems to have been con-
 trived. The reasoned explanation requirement of administrative law
 is meant to ensure that agencies offer genuine justifications for im-
 portant decisions, reasons that can be scrutinized by courts and the
 interested public. The explanation provided here was more of a dis-
 traction. In these unusual circumstances, the District Court was
 warranted in remanding to the agency. See Florida Power & Light
 Co. v. Lorion, 470 U. S. 729, 744 . Pp. 23–28.
 351 F. Supp. 3d 502 , affirmed in part, reversed in part, and remanded.

 ROBERTS, C. J., delivered the opinion for a unanimous Court with
respect to Parts I and II, and the opinion of the Court with respect to
Parts III, IV–B, and IV–C, in which THOMAS, ALITO, GORSUCH, and KAV-
ANAUGH, JJ., joined; with respect to Part IV–A, in which THOMAS,
GINSBURG, BREYER, SOTOMAYOR, KAGAN, and KAVANAUGH, JJ., joined;
and with respect to Part V, in which GINSBURG, BREYER, SOTOMAYOR,
and KAGAN, JJ., joined. THOMAS, J., filed an opinion concurring in part
and dissenting in part, in which GORSUCH and KAVANAUGH, JJ., joined.
BREYER, J., filed an opinion concurring in part and dissenting in part,
in which GINSBURG, SOTOMAYOR, and KAGAN, JJ., joined. ALITO, J., filed
an opinion concurring in part and dissenting in part.
 Cite as: 588 U. S. ____ (2019) 1

 Opinion of the Court

 NOTICE: This opinion is subject to formal revision before publication in the
 preliminary print of the United States Reports. Readers are requested to
 notify the Reporter of Decisions, Supreme Court of the United States, Wash-
 ington, D. C. 20543, of any typographical or other formal errors, in order
 that corrections may be made before the preliminary print goes to press.

SUPREME COURT OF THE UNITED STATES
 _________________

 No. 18–966
 _________________

DEPARTMENT OF COMMERCE, ET AL., PETITIONERS
 v. NEW YORK, ET AL.
ON WRIT OF CERTIORARI BEFORE JUDGMENT TO THE UNITED
 STATES COURT OF APPEALS FOR THE SECOND CIRCUIT
 [June 27, 2019]

 CHIEF JUSTICE ROBERTS delivered the opinion of the
Court.
 The Secretary of Commerce decided to reinstate a ques-
tion about citizenship on the 2020 census questionnaire.
A group of plaintiffs challenged that decision on constitu-
tional and statutory grounds. We now decide whether the
Secretary violated the Enumeration Clause of the Consti-
tution, the Census Act, or otherwise abused his discretion.
 I
 A
 In order to apportion Members of the House of Repre-
sentatives among the States, the Constitution requires an
“Enumeration” of the population every 10 years, to be
made “in such Manner” as Congress “shall by Law direct.”
Art. I, §2, cl. 3; Amdt. 14, §2. In the Census Act, Congress
delegated to the Secretary of Commerce the task of con-
ducting the decennial census “in such form and content as
he may determine.” 13 U. S. C. §141 (a). The Secretary is
aided in that task by the Census Bureau, a statistical
agency housed within the Department of Commerce. See
§§2, 21.
2 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of the Court

 The population count derived from the census is used
not only to apportion representatives but also to allocate
federal funds to the States and to draw electoral districts.
Wisconsin v. City of New York, 517 U. S. 1 , 5–6 (1996).
The census additionally serves as a means of collecting
demographic information, which “is used for such varied
purposes as computing federal grant-in-aid benefits, draft-
ing of legislation, urban and regional planning, business
planning, and academic and social studies.” Baldrige v.
Shapiro, 455 U. S. 345 , 353–354, n. 9 (1982). Over the
years, the census has asked questions about (for example)
race, sex, age, health, education, occupation, housing, and
military service. It has also asked about radio ownership,
age at first marriage, and native tongue. The Census Act
obliges everyone to answer census questions truthfully
and requires the Secretary to keep individual answers
confidential, including from other Government agencies.
§§221, 8(b), 9(a).
 There have been 23 decennial censuses from the first
census in 1790 to the most recent in 2010. Every census
between 1820 and 2000 (with the exception of 1840) asked
at least some of the population about their citizenship or
place of birth. Between 1820 and 1950, the question was
asked of all households. Between 1960 and 2000, it was
asked of about one-fourth to one-sixth of the population.
That change was part of a larger effort to simplify the
census by asking most people a few basic demographic
questions (such as sex, age, race, and marital status) on a
short-form questionnaire, while asking a sample of the
population more detailed demographic questions on a
long-form questionnaire. In explaining the decision to
move the citizenship question to the long-form question-
naire, the Census Bureau opined that “general census
information on citizenship had become of less importance
compared with other possible questions to be included in
the census, particularly in view of the recent statutory
 Cite as: 588 U. S. ____ (2019) 3

 Opinion of the Court

requirement for annual alien registration which could
provide the Immigration and Naturalization Service, the
principal user of such data, with the information it needed.”
Dept. of Commerce, Bureau of Census, 1960 Censuses of
Population and Housing 194 (1966). 1
 In 2010, the year of the latest census, the format
changed again. All households received the same ques-
tionnaire, which asked about sex, age, race, Hispanic
origin, and living arrangements. The more detailed demo-
graphic questions previously asked on the long-form ques-
tionnaire, including the question about citizenship, were
instead asked in the American Community Survey (or
ACS), which is sent each year to a rotating sample of
about 2.6% of households.
 The Census Bureau and former Bureau officials have
resisted occasional proposals to resume asking a citizen-
ship question of everyone, on the ground that doing so
would discourage noncitizens from responding to the
census and lead to a less accurate count of the total popu-
lation. See, e.g., Federation of Am. Immigration Reform v.
Klutznick, 486 F. Supp. 564, 568 (DC 1980) (“[A]ccording
to the Bureau[,] any effort to ascertain citizenship will
inevitably jeopardize the overall accuracy of the popula-
tion count”); Brief for Former Directors of the U. S. Census
Bureau as Amici Curiae in Evenwel v. Abbott, O. T. 2014,
No. 14–940, p. 25 (inquiring about citizenship would “in-
variably lead to a lower response rate”).
 B
 In March 2018, Secretary of Commerce Wilbur Ross
announced in a memo that he had decided to reinstate a
question about citizenship on the 2020 decennial census
questionnaire. The Secretary stated that he was acting at

——————
 1 The annual alien registration requirement was repealed in 1981.

See §11, 95 Stat. 1617 (1981).
4 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of the Court

the request of the Department of Justice (DOJ), which
sought improved data about citizen voting-age population
for purposes of enforcing the Voting Rights Act (or VRA)—
specifically the Act’s ban on diluting the influence of mi-
nority voters by depriving them of single-member districts
in which they can elect their preferred candidates. App. to
Pet. for Cert. 548a. DOJ explained that federal courts
determine whether a minority group could constitute a
majority in a particular district by looking to the citizen
voting-age population of the group. According to DOJ, the
existing citizenship data from the American Community
Survey was not ideal: It was not reported at the level of
the census block, the basic component of legislative dis-
tricting plans; it had substantial margins of error; and it
did not align in time with the census-based population
counts used to draw legislative districts. DOJ therefore
formally requested reinstatement of the citizenship ques-
tion on the census questionnaire. Id., at 565a–569a.
 The Secretary’s memo explained that the Census Bu-
reau initially analyzed, and the Secretary considered,
three possible courses of action. The first was to continue
to collect citizenship information in the American Com-
munity Survey and attempt to develop a data model that
would more accurately estimate citizenship at the census
block level. The Secretary rejected that option because the
Bureau “did not assert and could not confirm” that such
ACS-based data modeling was possible “with a sufficient
degree of accuracy.” Id., at 551a.
 The second option was to reinstate a citizenship ques-
tion on the decennial census. The Bureau predicted that
doing so would discourage some noncitizens from respond-
ing to the census. That would necessitate increased “non-
response follow up” operations—procedures the Bureau
uses to attempt to count people who have not responded to
the census—and potentially lead to a less accurate count
of the total population.
 Cite as: 588 U. S. ____ (2019) 5

 Opinion of the Court

 Option three was to use administrative records from
other agencies, such as the Social Security Administration
and Citizenship and Immigration Services, to provide DOJ
with citizenship data. The Census Bureau recommended
this option, and the Secretary found it a “potentially ap-
pealing solution” because the Bureau has long used ad-
ministrative records to supplement and improve census
data. Id., at 554a. But the Secretary concluded that
administrative records alone were inadequate because
they were missing for more than 10% of the population.
 The Secretary ultimately asked the Census Bureau to
develop a fourth option that would combine options two
and three: reinstate a citizenship question on the census
questionnaire, and also use the time remaining until the
2020 census to “further enhance” the Bureau’s “adminis-
trative record data sets, protocols, and statistical models.”
 Id., at 555a. The memo explained that, in the Secretary’s
judgment, the fourth option would provide DOJ with the
“most complete and accurate” citizen voting-age popula-
tion data in response to its request. Id., at 556a.
 The Secretary “carefully considered” the possibility
that reinstating a citizenship question would depress
the response rate. Ibid. But after evaluating the
Bureau’s “limited empirical evidence” on the question—
evidence drawn from estimated non-response rates to
previous American Community Surveys and census
questionnaires—the Secretary concluded that it was not
possible to “determine definitively” whether inquiring
about citizenship in the census would materially affect
response rates. Id., at 557a, 562a. He also noted the long
history of the citizenship question on the census, as well
as the facts that the United Nations recommends collect-
ing census-based citizenship information, and other major
democracies such as Australia, Canada, France, Indonesia,
Ireland, Germany, Mexico, Spain, and the United Kingdom
inquire about citizenship in their censuses. Altogether,
6 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of the Court

the Secretary determined that “the need for accurate
citizenship data and the limited burden that the rein-
statement of the citizenship question would impose out-
weigh fears about a potentially lower response rate.” Id.,
at 557a.
 C
 Shortly after the Secretary announced his decision, two
groups of plaintiffs filed suit in Federal District Court in
New York, challenging the decision on several grounds.
The first group of plaintiffs included 18 States, the District
of Columbia, various counties and cities, and the United
States Conference of Mayors. They alleged that the Secre-
tary’s decision violated the Enumeration Clause of the
Constitution and the requirements of the Administrative
Procedure Act. The second group of plaintiffs consisted of
several non-governmental organizations that work with
immigrant and minority communities. They added an
equal protection claim. The District Court consolidated
the two cases. Both groups of plaintiffs are respondents
here.
 The Government moved to dismiss the lawsuits, arguing
that the Secretary’s decision was unreviewable and that
respondents had failed to state cognizable claims under
the Enumeration Clause and the Equal Protection Clause.
The District Court dismissed the Enumeration Clause
claim but allowed the other claims to proceed. 315 F.
Supp. 3d 766 (SDNY 2018).
 In June 2018, the Government submitted to the District
Court the Commerce Department’s “administrative rec-
ord”: the materials that Secretary Ross considered in
making his decision. That record included DOJ’s Decem-
ber 2017 letter requesting reinstatement of the citizenship
question, as well as several memos from the Census Bu-
reau analyzing the predicted effects of reinstating the
question. Shortly thereafter, at DOJ’s urging, the Gov-
 Cite as: 588 U. S. ____ (2019) 7

 Opinion of the Court

ernment supplemented the record with a new memo from
the Secretary, “intended to provide further background
and context regarding” his March 2018 memo. App. to
Pet. for Cert. 546a. The supplemental memo stated that
the Secretary had begun considering whether to add the
citizenship question in early 2017, and had inquired
whether DOJ “would support, and if so would request,
inclusion of a citizenship question as consistent with and
useful for enforcement of the Voting Rights Act.” Ibid.
According to the Secretary, DOJ “formally” requested
reinstatement of the citizenship question after that in-
quiry. Ibid.
 Respondents argued that the supplemental memo indi-
cated that the Government had submitted an incomplete
record of the materials considered by the Secretary. They
asked the District Court to compel the Government to
complete the administrative record. The court granted
that request, and the parties jointly stipulated to the
inclusion of more than 12,000 pages of additional materi-
als in the administrative record. Among those materials
were emails and other records confirming that the Secre-
tary and his staff began exploring the possibility of rein-
stating a citizenship question shortly after he was con-
firmed in early 2017, attempted to elicit requests for
citizenship data from other agencies, and eventually per-
suaded DOJ to request reinstatement of the question for
VRA enforcement purposes.
 In addition, respondents asked the court to authorize
discovery outside the administrative record. They claimed
that such an unusual step was warranted because they
had made a strong preliminary showing that the Secretary
had acted in bad faith. See Citizens to Preserve Overton
Park, Inc. v. Volpe, 401 U. S. 402, 420 (1971). The court
also granted that request, authorizing expert discovery
and depositions of certain DOJ and Commerce Depart-
ment officials.
8 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of the Court

 In August and September 2018, the District Court
issued orders compelling depositions of Secretary Ross and
of the Acting Assistant Attorney General for DOJ’s Civil
Rights Division. We granted the Government’s request to
stay the Secretary’s deposition pending further review, but
we declined to stay the Acting AAG’s deposition or the
other extra-record discovery that the District Court had
authorized.
 The District Court held a bench trial and issued findings
of fact and conclusions of law on respondents’ statutory
and equal protection claims. After determining that re-
spondents had standing to sue, the District Court ruled
that the Secretary’s action was arbitrary and capricious,
based on a pretextual rationale, and violated certain
provisions of the Census Act. On the equal protection
claim, however, the District Court concluded that re-
spondents had not met their burden of showing that the
Secretary was motivated by discriminatory animus. The
court granted judgment to respondents on their statutory
claims, vacated the Secretary’s decision, and enjoined him
from reinstating the citizenship question until he cured
the legal errors the court had identified. 351 F. Supp. 3d
502 (SDNY 2019).
 The Government appealed to the Second Circuit, but
also filed a petition for writ of certiorari before judgment,
asking this Court to review the District Court’s decision
directly because the case involved an issue of imperative
public importance, and the census questionnaire needed to
be finalized for printing by the end of June 2019. We
granted the petition. 586 U. S. ___ (2019). At the Gov-
ernment’s request, we later ordered the parties to address
whether the Enumeration Clause provided an alternative
basis to affirm. 586 U. S. ___ (2019).
 II
 We begin with jurisdiction. Article III of the Constitu-
 Cite as: 588 U. S. ____ (2019) 9

 Opinion of the Court

tion limits federal courts to deciding “Cases” and “Contro-
versies.” For a legal dispute to qualify as a genuine case
or controversy, at least one plaintiff must have standing to
sue. The doctrine of standing “limits the category of liti-
gants empowered to maintain a lawsuit in federal court to
seek redress for a legal wrong” and “confines the federal
courts to a properly judicial role.” Spokeo, Inc. v. Robins,
 578 U. S. ___ , ___ (2016) (slip op., at 6). To have standing,
a plaintiff must “present an injury that is concrete, partic-
ularized, and actual or imminent; fairly traceable to the
defendant’s challenged behavior; and likely to be re-
dressed by a favorable ruling.” Davis v. Federal Election
Comm’n, 554 U. S. 724, 733 (2008).
 Respondents assert a number of injuries—diminishment
of political representation, loss of federal funds, degrada-
tion of census data, and diversion of resources—all of
which turn on their expectation that reinstating a citizen-
ship question will depress the census response rate and
lead to an inaccurate population count. Several States
with a disproportionate share of noncitizens, for example,
anticipate losing a seat in Congress or qualifying for less
federal funding if their populations are undercounted.
These are primarily future injuries, which “may suffice if
the threatened injury is certainly impending, or there is a
substantial risk that the harm will occur.” Susan B.
Anthony List v. Driehaus, 573 U. S. 149 , 158 (2014) (inter-
nal quotation marks omitted).
 The District Court concluded that the evidence at trial
established a sufficient likelihood that the reinstatement
of a citizenship question would result in noncitizen house-
holds responding to the census at lower rates than other
groups, which in turn would cause them to be undercounted
and lead to many of respondents’ asserted injuries. For
purposes of standing, these findings of fact were not so
suspect as to be clearly erroneous.
 We therefore agree that at least some respondents have
10 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of the Court

Article III standing. Several state respondents here have
shown that if noncitizen households are undercounted by
as little as 2%—lower than the District Court’s 5.8% pre-
diction—they will lose out on federal funds that are dis-
tributed on the basis of state population. That is a suffi-
ciently concrete and imminent injury to satisfy Article III,
and there is no dispute that a ruling in favor of respond-
ents would redress that harm.
 The Government contends, however, that any harm to
respondents is not fairly traceable to the Secretary’s deci-
sion, because such harm depends on the independent
action of third parties choosing to violate their legal duty
to respond to the census. The chain of causation is made
even more tenuous, the Government argues, by the fact
that such intervening, unlawful third-party action would
be motivated by unfounded fears that the Federal Gov-
ernment will itself break the law by using noncitizens’
answers against them for law enforcement purposes. The
Government invokes our steady refusal to “endorse stand-
ing theories that rest on speculation about the decisions of
independent actors,” Clapper v. Amnesty Int’l USA, 568
U. S. 398, 414 (2013), particularly speculation about
future unlawful conduct, Los Angeles v. Lyons, 461 U. S.
95, 105 (1983).
 But we are satisfied that, in these circumstances, re-
spondents have met their burden of showing that third
parties will likely react in predictable ways to the citizen-
ship question, even if they do so unlawfully and despite
the requirement that the Government keep individual
answers confidential. The evidence at trial established
that noncitizen households have historically responded to
the census at lower rates than other groups, and the Dis-
trict Court did not clearly err in crediting the Census
Bureau’s theory that the discrepancy is likely attributable
at least in part to noncitizens’ reluctance to answer a
citizenship question. Respondents’ theory of standing
 Cite as: 588 U. S. ____ (2019) 11

 Opinion of the Court

thus does not rest on mere speculation about the decisions
of third parties; it relies instead on the predictable effect of
Government action on the decisions of third parties. Cf.
Bennett v. Spear, 520 U. S. 154 , 169–170 (1997); Davis,
554 U. S., at 734–735. Because Article III “requires no
more than de facto causality,” Block v. Meese, 793 F. 2d
1303, 1309 (CADC 1986) (Scalia, J.), traceability is satis-
fied here. We may therefore consider the merits of re-
spondents’ claims, at least as far as the Constitution is
concerned.
 III
 The Enumeration Clause of the Constitution does not
provide a basis to set aside the Secretary’s decision. The
text of that clause “vests Congress with virtually unlim-
ited discretion in conducting the decennial ‘actual Enu-
meration,’ ” and Congress “has delegated its broad authority
over the census to the Secretary.” Wisconsin, 517 U. S., at
19 . Given that expansive grant of authority, we have
rejected challenges to the conduct of the census where the
Secretary’s decisions bore a “reasonable relationship to the
accomplishment of an actual enumeration.” Id., at 20 .
 Respondents ask us to evaluate the Secretary’s decision
to reinstate a citizenship question under that “reasonable
relationship” standard, but we agree with the District
Court that a different analysis is needed here. Our cases
applying that standard concerned decisions about the
population count itself—such as a postcensus decision not
to use a particular method to adjust an undercount, id., at
4 , and a decision to allocate overseas military personnel to
their home States, Franklin v. Massachusetts, 505 U. S.
788 , 790–791 (1992). We have never applied the standard
to decisions about what kinds of demographic information
to collect in the course of taking the census. Indeed, as the
District Court recognized, applying the “reasonable rela-
tionship” standard to every census-related decision “would
12 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of the Court

lead to the conclusion that it is unconstitutional to ask any
demographic question on the census” because “asking such
questions bears no relationship whatsoever to the goal of
an accurate headcount.” 315 F. Supp. 3d, at 804–805. Yet
demographic questions have been asked in every census
since 1790, and questions about citizenship in particular
have been asked for nearly as long. Like the District
Court, we decline respondents’ invitation to measure the
constitutionality of the citizenship question by a stand-
ard that would seem to render every census since 1790
unconstitutional.
 We look instead to Congress’s broad authority over the
census, as informed by long and consistent historical
practice. All three branches of Government have under-
stood the Constitution to allow Congress, and by extension
the Secretary, to use the census for more than simply
counting the population. Since 1790, Congress has sought,
or permitted the Secretary to seek, information about
matters as varied as age, sex, marital status, health,
trade, profession, literacy, and value of real estate owned.
See id., at 801 . Since 1820, it has sought, or permitted the
Secretary to seek, information about citizenship in partic-
ular. Federal courts have approved the practice of collect-
ing demographic data in the census. See, e.g., United
States v. Moriarity, 106 F. 886, 891 (CC SDNY 1901) (duty
to take a census of population “does not prohibit the gath-
ering of other statistics, if ‘necessary and proper,’ for the
intelligent exercise of other powers enumerated in the
constitution”). While we have never faced the question
directly, we have assumed that Congress has the power to
use the census for information-gathering purposes, see
Legal Tender Cases, 12 Wall. 457, 536 (1871), and we have
recognized the role of the census as a “linchpin of the
federal statistical system by collecting data on the charac-
teristics of individuals, households, and housing units
throughout the country,” Department of Commerce v.
 Cite as: 588 U. S. ____ (2019) 13

 Opinion of the Court

United States House of Representatives, 525 U. S. 316, 341
(1999) (internal quotation marks omitted).
 That history matters. Here, as in other areas, our
interpretation of the Constitution is guided by a Govern-
ment practice that “has been open, widespread, and un-
challenged since the early days of the Republic.” NLRB v.
Noel Canning, 573 U. S. 513 , 572 (2014) (Scalia, J., con-
curring in judgment); see also Wisconsin, 517 U. S., at 21
(noting “importance of historical practice” in census con-
text). In light of the early understanding of and long
practice under the Enumeration Clause, we conclude that
it permits Congress, and by extension the Secretary, to
inquire about citizenship on the census questionnaire. We
need not, and do not, decide the constitutionality of any
other question that Congress or the Secretary might de-
cide to include in the census.
 IV
 The District Court set aside the Secretary’s decision to
reinstate a citizenship question on the grounds that the
Secretary acted arbitrarily and violated certain provisions
of the Census Act. The Government contests those rul-
ings, but also argues that the Secretary’s decision was not
judicially reviewable under the Administrative Procedure
Act in the first place. We begin with that contention.
 A
 The Administrative Procedure Act embodies a “basic
presumption of judicial review,” Abbott Laboratories v.
Gardner, 387 U. S. 136, 140 (1967), and instructs review-
ing courts to set aside agency action that is “arbitrary,
capricious, an abuse of discretion, or otherwise not in
accordance with law,” 5 U. S. C. §706 (2)(A). Review is not
available, however, “to the extent that” a relevant statute
precludes it, §701(a)(1), or the agency action is “committed
to agency discretion by law,” §701(a)(2). The Government
14 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of the Court

argues that the Census Act commits to the Secretary’s
unreviewable discretion decisions about what questions to
include on the decennial census questionnaire.
 We disagree. To be sure, the Act confers broad authority
on the Secretary. Section 141(a) instructs him to take “a
decennial census of population” in “such form and content
as he may determine, including the use of sampling proce-
dures and special surveys.” 13 U. S. C. §141 . The Act
defines “census of population” to mean “a census of popu-
lation, housing, and matters relating to population and
housing,” §141(g), and it authorizes the Secretary, in
“connection with any such census,” to “obtain such other
census information as necessary,” §141(a). It also states
that the “Secretary shall prepare questionnaires, and shall
determine the inquiries, and the number, form, and subdi-
visions thereof, for the statistics, surveys, and censuses
provided for in this title.” §5. And it authorizes him to
acquire materials, such as administrative records, from
other federal, state, and local agencies in aid of conducting
the census. §6. Those provisions leave much to the
Secretary’s discretion. See Wisconsin, 517 U. S., at 19
(“Through the Census Act, Congress has delegated its
broad authority over the census to the Secretary.”).
 But they do not leave his discretion unbounded. In
order to give effect to the command that courts set aside
agency action that is an abuse of discretion, and to honor
the presumption of judicial review, we have read the
§701(a)(2) exception for action committed to agency discre-
tion “quite narrowly, restricting it to ‘those rare circum-
stances where the relevant statute is drawn so that a
court would have no meaningful standard against which to
judge the agency’s exercise of discretion.’ ” Weyerhaeuser
Co. v. United States Fish and Wildlife Serv., 586 U. S. ___ ,
___ (2018) (slip op., at 12) (quoting Lincoln v. Vigil, 508
U. S. 182, 191 (1993)). And we have generally limited the
exception to “certain categories of administrative decisions
 Cite as: 588 U. S. ____ (2019) 15

 Opinion of the Court

that courts traditionally have regarded as ‘committed to
agency discretion,’ ” id., at 191 , such as a decision not to
institute enforcement proceedings, Heckler v. Chaney, 470
U. S. 821 , 831–832 (1985), or a decision by an intelligence
agency to terminate an employee in the interest of national
security, Webster v. Doe, 486 U. S. 592 , 600–601 (1988).
 The taking of the census is not one of those areas tradi-
tionally committed to agency discretion. We and other
courts have entertained both constitutional and statutory
challenges to census-related decisionmaking. See, e.g.,
Department of Commerce, 525 U. S. 316 ; Wisconsin, 517
U. S. 1 ; Carey v. Klutznick, 637 F. 2d 834 (CA2 1980).
 Nor is the statute here drawn so that it furnishes no
meaningful standard by which to judge the Secretary’s
action. In contrast to the National Security Act in Web-
ster, which gave the Director of Central Intelligence dis-
cretion to terminate employees whenever he “deem[ed]” it
“advisable,” 486 U. S., at 594 , the Census Act constrains
the Secretary’s authority to determine the form and con-
tent of the census in a number of ways. Section 195, for
example, governs the extent to which he can use statistical
sampling. Section 6(c), which will be considered in more
detail below, circumscribes his power in certain circum-
stances to collect information through direct inquiries
when administrative records are available. More generally,
by mandating a population count that will be used to
apportion representatives, see §141(b), 2 U. S. C. §2a, the
Act imposes “a duty to conduct a census that is accurate
and that fairly accounts for the crucial representational
rights that depend on the census and the apportionment.”
Franklin, 505 U. S., at 819–820 (Stevens, J., concurring in
part and concurring in judgment).
 The Secretary’s decision to reinstate a citizenship ques-
tion is amenable to review for compliance with those and
other provisions of the Census Act, according to the gen-
eral requirements of reasoned agency decisionmaking.
16 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of the Court

Because this is not a case in which there is “no law to
apply,” Overton Park, 401 U. S., at 410 , the Secretary’s
decision is subject to judicial review.
 B
 At the heart of this suit is respondents’ claim that the
Secretary abused his discretion in deciding to reinstate a
citizenship question. We review the Secretary’s exercise of
discretion under the deferential “arbitrary and capricious”
standard. See 5 U. S. C. §706 (2)(A). Our scope of review
is “narrow”: we determine only whether the Secretary
examined “the relevant data” and articulated “a satisfac-
tory explanation” for his decision, “including a rational
connection between the facts found and the choice made.”
Motor Vehicle Mfrs. Assn. of United States, Inc. v. State
Farm Mut. Automobile Ins. Co., 463 U. S. 29, 43 (1983)
(internal quotation marks omitted). We may not substi-
tute our judgment for that of the Secretary, ibid., but
instead must confine ourselves to ensuring that he re-
mained “within the bounds of reasoned decisionmaking,”
Baltimore Gas & Elec. Co. v. Natural Resources Defense
Council, Inc., 462 U. S. 87, 105 (1983).
 The District Court set aside the Secretary’s decision for
two independent reasons: His course of action was not
supported by the evidence before him, and his stated
rationale was pretextual. We focus on the first point here
and take up the question of pretext later.
 The Secretary examined the Bureau’s analysis of vari-
ous ways to collect improved citizenship data and ex-
plained why he thought the best course was to both rein-
state a citizenship question and use citizenship data from
administrative records to fill in the gaps. He considered
but rejected the Bureau’s recommendation to use adminis-
trative records alone. As he explained, records are lacking
for about 10% of the population, so the Bureau would still
need to estimate citizenship for millions of voting-age
 Cite as: 588 U. S. ____ (2019) 17

 Opinion of the Court

people. Asking a citizenship question of everyone, the
Secretary reasoned, would eliminate the need to estimate
citizenship for many of those people. And supplementing
census responses with administrative record data would
help complete the picture and allow the Bureau to better
estimate citizenship for the smaller set of cases where it
was still necessary to do so.
 The evidence before the Secretary supported that deci-
sion. As the Bureau acknowledged, each approach—using
administrative records alone, or asking about citizenship
and using records to fill in the gaps—entailed tradeoffs
between accuracy and completeness. Without a citizen-
ship question, the Bureau would need to estimate the
citizenship of about 35 million people; with a citizenship
question, it would need to estimate the citizenship of only
13.8 million. Under either approach, there would be some
errors in both the administrative records and the Bureau’s
estimates. With a citizenship question, there would also
be some erroneous self-responses (about 500,000) and
some conflicts between responses and administrative
record data (about 9.5 million).
 The Bureau explained that the “relative quality” of the
citizenship data generated by each approach would depend
on the “relative importance of the errors” in each, but it
was not able to “quantify the relative magnitude of the
errors across the alternatives.” App. 148. The Bureau
nonetheless recommended using administrative records
alone because it had “high confidence” that it could develop
an accurate model for estimating the citizenship of the 35
million people for whom administrative records were not
available, and it thought the resulting citizenship data
would be of superior quality. Id., at 146 , 158–159. But
when the time came for the Secretary to make a decision,
the model did not yet exist, and even if it had, there was
no way to gauge its relative accuracy. As the Bureau put
it, “we will most likely never possess a fully adequate
18 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of the Court

truth deck to benchmark” the model—which appears to be
bureaucratese for “maybe, maybe not.” Id., at 146 . The
Secretary opted instead for the approach that would yield
a more complete set of data at an acceptable rate of accu-
racy, and would require estimating the citizenship of
fewer people.
 The District Court overruled that choice, agreeing with
the Bureau’s assessment that its recommended approach
would yield higher quality citizenship data on the whole.
But the choice between reasonable policy alternatives in
the face of uncertainty was the Secretary’s to make. He
considered the relevant factors, weighed risks and bene-
fits, and articulated a satisfactory explanation for his
decision. In overriding that reasonable exercise of discre-
tion, the court improperly substituted its judgment for
that of the agency.
 The Secretary then weighed the benefit of collecting
more complete and accurate citizenship data against the
risk that inquiring about citizenship would depress census
response rates, particularly among noncitizen households.
In the Secretary’s view, that risk was difficult to assess.
The Bureau predicted a 5.1% decline in response rates
among noncitizen households if the citizenship question
were reinstated. 2 It relied for that prediction primarily on
studies showing that, while noncitizens had responded at
lower rates than citizens to the 2000 short-form and 2010
censuses, which did not ask about citizenship, they re-
sponded at even lower rates than citizens to the 2000 long-
form census and the 2010 American Community Survey,
which did ask about citizenship. The Bureau thought it
was reasonable to infer that the citizenship question
accounted for the differential decline in noncitizen re-
——————
 2 Several months after the Secretary made his decision, the Bureau

updated its prediction to 5.8%, the figure the District Court later relied
on in its standing analysis. See 351 F. Supp. 3d 502, 579 (SDNY 2019).
 Cite as: 588 U. S. ____ (2019) 19

 Opinion of the Court

sponses. But, the Secretary explained, the Bureau was
unable to rule out other causes. For one thing, the evi-
dence before the Secretary suggested that noncitizen
households tend to be more distrustful of, and less likely
to respond to, any government effort to collect infor-
mation. For another, both the 2000 long-form census and
2010 ACS asked over 45 questions on a range of topics,
including employment, income, and housing characteris-
tics. Noncitizen households might disproportionately fail
to respond to a lengthy and intrusive Government ques-
tionnaire for a number of reasons besides reluctance to
answer a citizenship question—reasons relating to educa-
tion level, socioeconomic status, and less exposure to
Government outreach efforts. See App. to Pet. for Cert.
553a–554a, 557a–558a.
 The Secretary justifiably found the Bureau’s analysis
inconclusive. Weighing that uncertainty against the value
of obtaining more complete and accurate citizenship data,
he determined that reinstating a citizenship question was
worth the risk of a potentially lower response rate. That
decision was reasonable and reasonably explained, partic-
ularly in light of the long history of the citizenship ques-
tion on the census.
 JUSTICE BREYER would conclude otherwise, but only by
subordinating the Secretary’s policymaking discretion to
the Bureau’s technocratic expertise. JUSTICE BREYER’s
analysis treats the Bureau’s (pessimistic) prediction about
response rates and (optimistic) assumptions about its data
modeling abilities as touchstones of substantive reason-
ableness rather than simply evidence for the Secretary to
consider. He suggests that the Secretary should have
deferred to the Bureau or at least offered some special
justification for drawing his own inferences and adopting
his own assumptions. But the Census Act authorizes the
Secretary, not the Bureau, to make policy choices within
the range of reasonable options. And the evidence before
20 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of the Court

the Secretary hardly led ineluctably to just one reasonable
course of action. It called for value-laden decisionmaking
and the weighing of incommensurables under conditions of
uncertainty. The Secretary was required to consider the
evidence and give reasons for his chosen course of action.
He did so. It is not for us to ask whether his decision was
“the best one possible” or even whether it was “better than
the alternatives.” FERC v. Electric Power Supply Assn.,
 577 U. S. ___ , ___ (2016) (slip op., at 30). By second-
guessing the Secretary’s weighing of risks and benefits
and penalizing him for departing from the Bureau’s infer-
ences and assumptions, JUSTICE BREYER—like the District
Court—substitutes his judgment for that of the agency.
 C
 The District Court also ruled that the Secretary violated
two particular provisions of the Census Act, §6(c) and
§141(f ).
 Section 6 has three subsections. Subsections (a) and (b)
authorize the Secretary to acquire administrative records
from other federal agencies and from state and local gov-
ernments. 3 Subsection (c) states:
 “To the maximum extent possible and consistent
 with the kind, timeliness, quality and scope of the sta-
 tistics required, the Secretary shall acquire and use
 information available from any source referred to in
——————
 3 The full text of subsections (a) and (b) provides:
 “(a) The Secretary, whenever he considers it advisable, may call upon
any other department, agency, or establishment of the Federal
Government, or of the government of the District of Columbia, for
information pertinent to the work provided for in this title.
 “(b) The Secretary may acquire, by purchase or otherwise, from
States, counties, cities, or other units of government, or their instru-
mentalities, or from private persons and agencies, such copies of
records, reports, and other material as may be required for the efficient
and economical conduct of the censuses and surveys provided for in this
title.” 13 U. S. C. §6 .
 Cite as: 588 U. S. ____ (2019) 21

 Opinion of the Court

 subsection (a) or (b) of this section instead of conduct-
 ing direct inquiries.” 13 U. S. C. §6 (c).
 The District Court held, and respondents argue, that the
Secretary failed to comply with §6(c) because he opted to
collect citizenship data using direct inquiries when it was
possible to provide DOJ with data from administrative
records alone.
 At the outset, §6(c) may not even apply here. It governs
the Secretary’s choices with respect to “statistics re-
quired.” The parties have assumed that phrase refers to
census-related data that the Secretary wishes to acquire,
but it may instead refer to particular kinds of statistics
that other provisions of the Census Act actually do require
the Secretary to collect and publish. See, e.g., §41 (“The
Secretary shall collect and publish statistics concerning
[cotton and cotton production].”); §61 (“The Secretary shall
collect, collate, and publish monthly statistics concerning
[vegetable and animal oils and the like].”); §91 (“The Sec-
retary shall collect and publish quarterly financial statis-
tics of business operations, organization, practices, man-
agement, and relation to other businesses.”). If so, §6(c)
would seem to have nothing to say about the Secretary’s
collection of census-related citizenship data, which is not a
“statistic” he is “required” to collect.
 Regardless, assuming the provision applies, the Secre-
tary complied with it, for essentially the same reasons
that his decision was not arbitrary and capricious. As he
explained, administrative records would not, in his judg-
ment, provide the more complete and accurate data that
DOJ sought. He thus could not, “consistent with” the kind
and quality of the “statistics required,” use administrative
records instead of asking about citizenship directly. Re-
spondents’ arguments to the contrary rehash their dis-
agreement with the Secretary’s policy judgment about
which approach would yield the most complete and accu-
22 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of the Court

rate citizenship data. For the reasons already discussed,
we may not substitute our judgment for that of the Secre-
tary here.
 We turn now to §141(f ), which requires the Secretary to
report to Congress about his plans for the census. Para-
graph (1) instructs him to submit, at least three years
before the census date, a report containing his “determina-
tion of the subjects proposed to be included, and the types
of information to be compiled,” in the census. Paragraph
(2) then tells him to submit, at least two years before the
census date, a report containing his “determination of the
questions proposed to be included” in the census. Para-
graph (3) provides:
 “[A]fter submission of a report under paragraph (1)
 or (2) of this subsection and before the appropriate
 census date, if the Secretary finds new circumstances
 exist which necessitate that the subjects, types of in-
 formation, or questions contained in reports so sub-
 mitted be modified, [he shall submit] a report contain-
 ing the Secretary’s determination of the subjects,
 types of information, or questions as proposed to be
 modified.”
 The Secretary timely submitted his paragraph (1) report
in March 2017. It did not mention citizenship. In Decem-
ber 2017, he received DOJ’s formal request. Three months
later, in March 2018, he timely submitted his para-
graph (2) report. It did propose asking a question about
citizenship.
 The District Court held that the Secretary’s failure to
mention citizenship in his March 2017 report violated
§141(f )(1) and provided an independent basis to set aside
his action. Assuming without deciding that the Secre-
tary’s compliance with the reporting requirement is for
courts—rather than Congress—to police, we disagree. The
Secretary’s March 2018 report satisfied the requirements
 Cite as: 588 U. S. ____ (2019) 23

 Opinion of the Court

of paragraph (3): By informing Congress that he proposed
to include a citizenship question, the Secretary necessarily
also informed Congress that he proposed to modify the
original list of subjects that he submitted in the March
2017 report. Nothing in §141(f ) suggests that the same
report cannot simultaneously fulfill the requirements of
paragraphs (2) and (3). And to the extent paragraph (3)
requires the Secretary to explain his finding of new cir-
cumstances, he did so in his March 2018 memo, which
described DOJ’s intervening request.
 In any event, even if we agreed with the District Court
that the Secretary technically violated §141(f ) by submit-
ting a paragraph (2) report that doubled as a paragraph
(3) report, the error would surely be harmless in these
circumstances, where the Secretary nonetheless fully
informed Congress of, and explained, his decision. See 5
U. S. C. §706 (in reviewing agency action, “due account
shall be taken of the rule of prejudicial error”).
 V
 We now consider the District Court’s determination that
the Secretary’s decision must be set aside because it rested
on a pretextual basis, which the Government conceded
below would warrant a remand to the agency.
 We start with settled propositions. First, in order to
permit meaningful judicial review, an agency must “dis-
close the basis” of its action. Burlington Truck Lines, Inc.
v. United States, 371 U. S. 156 , 167–169 (1962) (internal
quotation marks omitted); see also SEC v. Chenery Corp.,
 318 U. S. 80, 94 (1943) (“[T]he orderly functioning of the
process of review requires that the grounds upon which
the administrative agency acted be clearly disclosed and
adequately sustained.”).
 Second, in reviewing agency action, a court is ordinarily
limited to evaluating the agency’s contemporaneous ex-
planation in light of the existing administrative record.
24 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of the Court

Vermont Yankee Nuclear Power Corp. v. Natural Re-
sources Defense Council, Inc., 435 U. S. 519, 549 (1978);
Camp v. Pitts, 411 U. S. 138 , 142–143 (1973) (per curiam).
That principle reflects the recognition that further judicial
inquiry into “executive motivation” represents “a substan-
tial intrusion” into the workings of another branch of
Government and should normally be avoided. Arlington
Heights v. Metropolitan Housing Development Corp., 429
U. S. 252, 268, n. 18 (1977); see Overton Park, 401 U. S., at
420 .
 Third, a court may not reject an agency’s stated reasons
for acting simply because the agency might also have had
other unstated reasons. See Jagers v. Federal Crop Ins.
Corp., 758 F. 3d 1179 , 1185–1186 (CA10 2014) (rejecting
argument that “the agency’s subjective desire to reach a
particular result must necessarily invalidate the result,
regardless of the objective evidence supporting the
agency’s conclusion”). Relatedly, a court may not set aside
an agency’s policymaking decision solely because it might
have been influenced by political considerations or
prompted by an Administration’s priorities. Agency policy-
making is not a “rarified technocratic process, unaffected
by political considerations or the presence of Presidential
power.” Sierra Club v. Costle, 657 F. 2d 298, 408 (CADC
1981). Such decisions are routinely informed by unstated
considerations of politics, the legislative process, public
relations, interest group relations, foreign relations, and
national security concerns (among others).
 Finally, we have recognized a narrow exception to the
general rule against inquiring into “the mental processes
of administrative decisionmakers.” Overton Park, 401
U. S., at 420 . On a “strong showing of bad faith or im-
proper behavior,” such an inquiry may be warranted and
may justify extra-record discovery. Ibid.
 The District Court invoked that exception in ordering
extra-record discovery here. Although that order was
 Cite as: 588 U. S. ____ (2019) 25

 Opinion of the Court

premature, we think it was ultimately justified in light of
the expanded administrative record. Recall that shortly
after this litigation began, the Secretary, prodded by DOJ,
filed a supplemental memo that added new, pertinent
information to the administrative record. The memo
disclosed that the Secretary had been considering the
citizenship question for some time and that Commerce had
inquired whether DOJ would formally request reinstate-
ment of the question. That supplemental memo prompted
respondents to move for both completion of the adminis-
trative record and extra-record discovery. The District
Court granted both requests at the same hearing, agreeing
with respondents that the Government had submitted an
incomplete administrative record and that the existing
evidence supported a prima facie showing that the VRA
rationale was pretextual.
 The Government did not challenge the court’s conclusion
that the administrative record was incomplete, and the
parties stipulated to the inclusion of more than 12,000
pages of internal deliberative materials as part of the
administrative record, materials that the court later held
were sufficient on their own to demonstrate pretext. The
Government did, however, challenge the District Court’s
order authorizing extra-record discovery, as well as the
court’s later orders compelling depositions of the Secretary
and of the Acting Assistant Attorney General for DOJ’s
Civil Rights Division.
 We agree with the Government that the District Court
should not have ordered extra-record discovery when it
did. At that time, the most that was warranted was the
order to complete the administrative record. But the new
material that the parties stipulated should have been part
of the administrative record—which showed, among other
things, that the VRA played an insignificant role in the
decisionmaking process—largely justified such extra-
record discovery as occurred (which did not include the
26 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of the Court

deposition of the Secretary himself). We accordingly
review the District Court’s ruling on pretext in light of all
the evidence in the record before the court, including the
extra-record discovery.
 That evidence showed that the Secretary was deter-
mined to reinstate a citizenship question from the time he
entered office; instructed his staff to make it happen;
waited while Commerce officials explored whether another
agency would request census-based citizenship data;
subsequently contacted the Attorney General himself to
ask if DOJ would make the request; and adopted the
Voting Rights Act rationale late in the process. In the
District Court’s view, this evidence established that the
Secretary had made up his mind to reinstate a citizenship
question “well before” receiving DOJ’s request, and did so
for reasons unknown but unrelated to the VRA. 351 F.
Supp. 3d, at 660 .
 The Government, on the other hand, contends that there
was nothing objectionable or even surprising in this. And
we agree—to a point. It is hardly improper for an agency
head to come into office with policy preferences and ideas,
discuss them with affected parties, sound out other agen-
cies for support, and work with staff attorneys to substan-
tiate the legal basis for a preferred policy. The record here
reflects the sometimes involved nature of Executive
Branch decisionmaking, but no particular step in the
process stands out as inappropriate or defective.
 And yet, viewing the evidence as a whole, we share the
District Court’s conviction that the decision to reinstate a
citizenship question cannot be adequately explained in
terms of DOJ’s request for improved citizenship data to
better enforce the VRA. Several points, considered to-
gether, reveal a significant mismatch between the decision
the Secretary made and the rationale he provided.
 The record shows that the Secretary began taking steps
to reinstate a citizenship question about a week into his
 Cite as: 588 U. S. ____ (2019) 27

 Opinion of the Court

tenure, but it contains no hint that he was considering
VRA enforcement in connection with that project. The
Secretary’s Director of Policy did not know why the Secre-
tary wished to reinstate the question, but saw it as his
task to “find the best rationale.” Id., at 551 . The Director
initially attempted to elicit requests for citizenship data
from the Department of Homeland Security and DOJ’s
Executive Office for Immigration Review, neither of which
is responsible for enforcing the VRA. After those attempts
failed, he asked Commerce staff to look into whether the
Secretary could reinstate the question without receiving a
request from another agency. The possibility that DOJ’s
Civil Rights Division might be willing to request citizen-
ship data for VRA enforcement purposes was proposed by
Commerce staff along the way and eventually pursued.
 Even so, it was not until the Secretary contacted the
Attorney General directly that DOJ’s Civil Rights Division
expressed interest in acquiring census-based citizenship
data to better enforce the VRA. And even then, the record
suggests that DOJ’s interest was directed more to helping
the Commerce Department than to securing the data. The
December 2017 letter from DOJ drew heavily on contribu-
tions from Commerce staff and advisors. Their influence
may explain why the letter went beyond a simple entreaty
for better citizenship data—what one might expect of a
typical request from another agency—to a specific request
that Commerce collect the data by means of reinstating a
citizenship question on the census. Finally, after sending
the letter, DOJ declined the Census Bureau’s offer to
discuss alternative ways to meet DOJ’s stated need for
improved citizenship data, further suggesting a lack of
interest on DOJ’s part.
 Altogether, the evidence tells a story that does not
match the explanation the Secretary gave for his decision.
In the Secretary’s telling, Commerce was simply acting on
a routine data request from another agency. Yet the
28 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of the Court

materials before us indicate that Commerce went to great
lengths to elicit the request from DOJ (or any other willing
agency). And unlike a typical case in which an agency
may have both stated and unstated reasons for a decision,
here the VRA enforcement rationale—the sole stated rea-
son—seems to have been contrived.
 We are presented, in other words, with an explanation
for agency action that is incongruent with what the record
reveals about the agency’s priorities and decisionmaking
process. It is rare to review a record as extensive as the
one before us when evaluating informal agency action—
and it should be. But having done so for the sufficient
reasons we have explained, we cannot ignore the discon-
nect between the decision made and the explanation given.
Our review is deferential, but we are “not required to
exhibit a naiveté from which ordinary citizens are free.”
United States v. Stanchich, 550 F. 2d 1294, 1300 (CA2
1977) (Friendly, J.). The reasoned explanation require-
ment of administrative law, after all, is meant to ensure
that agencies offer genuine justifications for important
decisions, reasons that can be scrutinized by courts and
the interested public. Accepting contrived reasons would
defeat the purpose of the enterprise. If judicial review is to
be more than an empty ritual, it must demand something
better than the explanation offered for the action taken in
this case.
 In these unusual circumstances, the District Court was
warranted in remanding to the agency, and we affirm that
disposition. See Florida Power & Light Co. v. Lorion, 470
U. S. 729, 744 (1985). We do not hold that the agency
decision here was substantively invalid. But agencies
must pursue their goals reasonably. Reasoned deci-
sionmaking under the Administrative Procedure Act calls
for an explanation for agency action. What was provided
here was more of a distraction.
 Cite as: 588 U. S. ____ (2019) 29

 Opinion of the Court

 * * *
 The judgment of the United States District Court for the
Southern District of New York is affirmed in part and
reversed in part, and the case is remanded for further
proceedings consistent with this opinion.

 It is so ordered.
 Cite as: 588 U. S. ____ (2019) 1

 Opinion of THOMAS, J.

SUPREME COURT OF THE UNITED STATES
 _________________

 No. 18–966
 _________________

DEPARTMENT OF COMMERCE, ET AL., PETITIONERS
 v. NEW YORK, ET AL.
ON WRIT OF CERTIORARI BEFORE JUDGMENT TO THE UNITED
 STATES COURT OF APPEALS FOR THE SECOND CIRCUIT
 [June 27, 2019]

 JUSTICE THOMAS, with whom JUSTICE GORSUCH and
JUSTICE KAVANAUGH join, concurring in part and dissent-
ing in part.
 In March 2018, the Secretary of Commerce exercised his
broad discretion over the administration of the decennial
census to resume a nearly unbroken practice of asking a
question relating to citizenship. Our only role in this case
is to decide whether the Secretary complied with the law
and gave a reasoned explanation for his decision. The
Court correctly answers these questions in the affirmative.
Ante, at 11–23. That ought to end our inquiry.
 The Court, however, goes further. For the first time
ever, the Court invalidates an agency action solely because
it questions the sincerity of the agency’s otherwise ade-
quate rationale. Echoing the din of suspicion and distrust
that seems to typify modern discourse, the Court declares
the Secretary’s memorandum “pretextual” because, “view-
ing the evidence as a whole,” his explanation that includ-
ing a citizenship question on the census would help en-
force the Voting Rights Act (VRA) “seems to have been
contrived.” Ante, at 23, 26, 28. The Court does not hold
that the Secretary merely had additional, unstated rea-
sons for reinstating the citizenship question. Rather, it
holds that the Secretary’s stated rationale did not factor at
all into his decision.
2 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of THOMAS, J.

 The Court’s holding reflects an unprecedented departure
from our deferential review of discretionary agency deci-
sions. And, if taken seriously as a rule of decision, this
holding would transform administrative law. It is not
difficult for political opponents of executive actions to
generate controversy with accusations of pretext, deceit,
and illicit motives. Significant policy decisions are regu-
larly criticized as products of partisan influence, interest-
group pressure, corruption, and animus. Crediting these
accusations on evidence as thin as the evidence here could
lead judicial review of administrative proceedings to de-
volve into an endless morass of discovery and policy dis-
putes not contemplated by the Administrative Procedure
Act (APA).
 Unable to identify any legal problem with the Secre-
tary’s reasoning, the Court imputes one by concluding that
he must not be telling the truth. The Court therefore up-
holds the decision of the District Court—which, in turn, was
transparently based on the application of an administration-
specific standard. App. to Pet. for Cert. 527a (crediting
respondents’ allegations that “the current Depart-
ment of Justice has shown little interest in enforcing the”
VRA (emphasis added)).
 The law requires a more impartial approach. Even
assuming we are authorized to engage in the review un-
dertaken by the Court—which is far from clear—we have
often stated that courts reviewing agency action owe the
Executive a “presumption of regularity.” Citizens to Pre-
serve Overton Park, Inc. v. Volpe, 401 U. S. 402, 415
(1971). The Court pays only lipservice to this principle.
But, the evidence falls far short of supporting its decision.
The Court, I fear, will come to regret inventing the princi-
ples it uses to achieve today’s result. I respectfully dissent
from Part V of the opinion of the Court. 1
——————
 1 JUSTICE KAVANAUGH and I join Parts I, II, III, and IV of the opinion
 Cite as: 588 U. S. ____ (2019) 3

 Opinion of THOMAS, J.

 I
 As the Court explains, federal law directs the Secretary
of Commerce to “take a decennial census.” 13 U. S. C.
§141 (a); see U. S. Const., Art. I, §2, cl. 3; Amdt. XIV, §2;
ante, at 1–2. The discretion afforded the Secretary is
extremely broad. Subject only to constitutional limitations
and a handful of inapposite statutory requirements, the
Secretary is expressly authorized to “determine the inquir-
ies” on the census questionnaire and to conduct the census
“in such form and content as he may determine.” §§5,
141(a); see ante, at 14–16, 20–23. 2 Prior census question-
naires have included questions ranging from sex, age, and
race to commute, education, and radio ownership. And
between 1820 and 2010, every decennial census question-
naire but one asked some segment of the population a
question related to citizenship. The 2010 census was the
first since 1840 that did not include any such question.
 In March 2018, the Secretary issued a memorandum
reinstating a citizenship question on the 2020 census. He
explained that the Department of Justice (DOJ) had for-
mally requested reinstatement of the question because the
data obtained would help enforce §2 of the VRA. He fur-
ther explained that the question had been well tested in
light of its extensive previous use, that he had consulted
with the Census Bureau on the proposal, and that his final
——————
of the Court. JUSTICE GORSUCH joins Parts I, II, III, IV–B, and IV–C.
 2 JUSTICE ALITO has made a strong argument that the specific decision

at issue here—whether to include a citizenship question on the cen-
sus—is a matter “committed to agency discretion by law.” 5 U. S. C.
§701 (a)(2); see post, at 3 (opinion concurring in part and dissenting in
part). As he explains, the Secretary’s decision plainly falls within the
scope of the Secretary’s constitutional authority, does not implicate any
statutory prohibition, and is among the “inquiries” and “content[s]” of
the census that the Secretary is expressly directed to “determine” for
himself. §§5, 141(a); see post, at 5–15. Nevertheless, I assume, for the
purpose of this opinion, that the Secretary’s decision is subject to
judicial review.
4 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of THOMAS, J.

decision incorporated feedback from the Bureau. He
recognized that staff at the Bureau believed that better
data could be obtained through modeling and reliance on
existing records, but he disagreed with that assessment,
explaining that the data was inconclusive and that he
thought it preferable to ask the question directly of the
entire population. Respondents brought suit, seeking
judicial review of the Secretary’s decision under the APA,
 5 U. S. C. §706 .
 II
 As relevant here, the APA requires courts to “hold un-
lawful and set aside” agency action that is “arbitrary,
capricious, an abuse of discretion, or otherwise not in
accordance with law.” §706(2)(A). We have emphasized
that “[r]eview under the arbitrary and capricious standard
is deferential.” National Assn. of Home Builders v. De-
fenders of Wildlife, 551 U. S. 644, 658 (2007); see Glick-
man v. Wileman Brothers & Elliott, Inc., 521 U. S. 457,
466, n. 8 (1997). It requires the reviewing court to deter-
mine whether the agency “ ‘examine[d] the relevant data
and articulate[d] a satisfactory explanation for its action.’ ”
FCC v. Fox Television Stations, Inc., 556 U. S. 502, 513
(2009). We have described this as a “ ‘narrow’ standard of
review” under which the reviewing court cannot “ ‘substi-
tute its judgment for that of the agency,’ and should ‘up-
hold a decision of less than ideal clarity if the agency’s
path may reasonably be discerned.’ ” Id., at 513–514
(citation omitted); accord, Motor Vehicle Mfrs. Assn. of
United States, Inc. v. State Farm Mut. Automobile Ins. Co.,
 463 U. S. 29, 43 (1983). 3
——————
 3 Deferential review of the agency’s discretionary choices and reason-

ing under the arbitrary-and-capricious standard stands in marked
contrast to a court’s plenary review of the agency’s interpretation and
application of the law. See §§706(A)–(D) (court must review agency
action to ensure that it complies with all “constitutional,” “statutory,”
 Cite as: 588 U. S. ____ (2019) 5

 Opinion of THOMAS, J.

 Part IV–B of the opinion of the Court correctly applies
this standard to conclude that the Secretary’s decision
survives ordinary arbitrary-and-capricious review. That
holding should end our inquiry.
 But the opinion continues. Acknowledging that “no
particular step” in the proceedings here “stands out as
inappropriate or defective,” even after reviewing “all the
evidence in the record . . . , including the extra-record
discovery,” ante, at 26, the Court nevertheless agrees with
the District Court that the Secretary’s rationale for rein-
stating the citizenship question was “pretextual—that is,
that the real reason for his decision was something other
than the sole reason he put forward in his memorandum,
namely enhancement of DOJ’s VRA enforcement efforts.”
 351 F. Supp. 3d 502, 660 (SDNY 2019); see ante, at 28.
According to the Court, something just “seems” wrong.
 Ibid.
 This conclusion is extraordinary. The Court engages in
an unauthorized inquiry into evidence not properly before
us to reach an unsupported conclusion. Moreover, each
step of the inquiry offends the presumption of regularity
we owe the Executive. The judgment of the District Court
should be reversed.
 A
 Section 706(2) of the APA contemplates review of the
administrative “record” to determine whether an agency’s
“action, findings, and conclusions” satisfy six specified
standards. See §§706(2)(A)–(F). None instructs the Court
to inquire into pretext. Consistent with this statutory
text, we have held that a court is “ordinarily limited to
evaluating the agency’s contemporaneous explanation in
light of the existing administrative record.” Ante, at 23

——————
and “procedur[al]” requirements, and is otherwise “in accordance with
law”).
6 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of THOMAS, J.

(citing Vermont Yankee Nuclear Power Corp. v. Natural
Resources Defense Council, Inc., 435 U. S. 519, 549 (1978));
see SEC v. Chenery Corp., 318 U. S. 80, 87 (1943) (“The
grounds upon which an administrative order must be
judged are those upon which the record discloses that its
action was based”). If an agency’s stated findings and
conclusions withstand scrutiny, the APA does not permit a
court to set aside the decision solely because the agency
had “other unstated reasons” for its decision, such as
“political considerations” or the “Administration’s priori-
ties.” Ante, at 24.
 Unsurprisingly, then, this Court has never held an
agency decision arbitrary and capricious on the ground
that its supporting rationale was “pretextual.” Nor has it
previously suggested that this was even a possibility.
Under “settled propositions” of administrative law, ante,
at 23, pretext is virtually never an appropriate or relevant
inquiry for a reviewing court to undertake.
 Respondents conceptualize pretext as a subset of “arbi-
trary and capricious” review. It is far from clear that they
are correct. But even if they were, an agency action is not
arbitrary or capricious merely because the decisionmaker
has other, unstated reasons for the decision. Ante, at 24.
Nor is an agency action arbitrary and capricious merely
because the decisionmaker was “inclined” to accomplish it
before confirming that the law and facts supported that
inclination. In re Dept. of Commerce, 586 U. S. ___ , ___
(2018) (GORSUCH, J., concurring in part and dissenting in
part) (slip op., at 2).
 Accordingly, even under respondents’ approach, a show-
ing of pretext could render an agency action arbitrary and
capricious only in the infinitesimally small number of
cases in which the administrative record establishes that
an agency’s stated rationale did not factor at all into the
decision, thereby depriving the action of an adequate
 Cite as: 588 U. S. ____ (2019) 7

 Opinion of THOMAS, J.

supporting rationale. 4 This showing is extremely difficult
to make because the administrative record will rarely, if
ever, contain evidence sufficient to show that an agency’s
stated rationale did not actually factor into its decision.
And we have stated that a “strong showing of bad faith or
improper behavior” is necessary to venture beyond the
agency’s “administrative findings” and inquire into “the
mental processes of administrative decisionmakers.”
Overton Park, 401 U. S., at 420 . 5 We have never before
found Overton Park’s exception satisfied, much less invali-
dated an agency action based on “pretext.”
 Undergirding our arbitrary-and-capricious analysis is
our longstanding precedent affording the Executive a
“presumption of regularity.” Id., at 415 ; see United States
v. Chemical Foundation, Inc., 272 U. S. 1 , 14–15 (1926).
This presumption reflects respect for a coordinate branch
of government whose officers not only take an oath to
——————
 4 We do not have before us a claim that information outside the ad-

ministrative record calls into question the legality of an agency action
based on an unstated, unlawful bias or motivation (e.g., a claim of
religious discrimination under the Free Exercise Clause). But to the
extent such a claim is viable, the analysis would have nothing to do
with the arbitrary-and-capricious review pressed by respondents. See
§§706(2)(A)–(C) (addressing agency actions that violate “constitutional”
or “statutory” requirements, or that “otherwise [are] not in accordance
with law”).
 5 Insofar as Overton Park authorizes an exception to review on the

administrative record, it has been criticized as having “no textual
grounding in the APA” and as “created by the Court, without citation or
explanation, to facilitate Article III review.” Gavoor & Platt, Adminis-
trative Records and the Courts, 67 U. Kan. L. Rev. 1 , 44 (2018); see id.,
at 22 (further arguing that the exception was “neither presented by the
facts of the case nor briefed by the parties”). The legitimacy and scope
of the exception—which by its terms contemplates only “administrative
officials who participated in the decision . . . giv[ing] testimony explain-
ing their action,” Overton Park, 401 U. S., at 420—is an important
question that may warrant future consideration. But because the
Court’s holding is incorrect regardless of the validity of the Overton
Park exception, I will apply it here.
8 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of THOMAS, J.

support the Constitution, as we do, Art. VI, but also are
charged with “faithfully execut[ing]” our laws, Art. II, §3.
See United States v. Morgan, 313 U. S. 409, 422 (1941)
(presumption of regularity ensures that the “integrity of
the administrative process” is appropriately respected). In
practice, then, we give the benefit of the doubt to the
agency.
 B
 The Court errs at the outset by proceeding beyond the
administrative record to evaluate pretext. Respondents
have not made a “strong showing of bad faith or improper
behavior.” Overton Park, supra, at 420 .
 The District Court’s initial order granting extra-record
discovery relied on four categories of evidence:
 “evidence that [the Secretary] was predisposed to re-
 instate the citizenship question when he took office;
 that the [DOJ] hadn’t expressed a desire for more de-
 tailed citizenship data until the Secretary solicited its
 view; that he overruled the objections of his agency’s
 career staff; and that he declined to order more test-
 ing of the question given its long history.” Dept. of
 Commerce, 586 U. S., at ___ (slip op., at 2).
None of this comes close to showing bad faith or improper
behavior. Indeed, there is nothing even “unusual about a
new cabinet secretary coming to office inclined to favor a
different policy direction, soliciting support from other
agencies to bolster his views, disagreeing with staff, or
cutting through red tape.” Ibid. Today all Members of the
Court who reach the question agree that the District
Court abused its discretion in ordering extra-record dis-
covery based on this evidence. Ante, at 25 (“We agree with
the Government that the District Court should not have
ordered extra-record discovery when it did”).
 Nevertheless, the Court excuses the error because, in its
 Cite as: 588 U. S. ____ (2019) 9

 Opinion of THOMAS, J.

view, “the new material that the parties [later] stipulated
should have been part of the administrative record . . .
largely justified such extra-record discovery as occurred.”
 Ibid. Given the requirement that respondents make a
“strong showing” of bad faith, one would expect the Court
to identify which “new material” supported such a show-
ing. It does not. Nor does the Court square its suggestion
that some of the extra-record discovery was not “justified”
with its consideration of “all . . . the extra-record discov-
ery.” Ante, at 25–26. Regardless, I assume that the Court
has in mind the administrative-record materials that the
District Court would later rely on to establish pretext:
 “evidence that [the Secretary] had made the decision
 to add the citizenship question well before DOJ re-
 quested its addition in December 2017; the absence of
 any mention, at all, of VRA enforcement in the discus-
 sions of adding the question that preceded the [DOJ]
 Letter; unsuccessful attempts by Commerce Depart-
 ment staff to shop around for a request by another
 agency regarding citizenship data; and [the Secre-
 tary’s] personal outreach to Attorney General Ses-
 sions, followed by the [DOJ] Letter; not to mention the
 conspicuous procedural irregularities that accompa-
 nied the decision to add the question.” 351 F. Supp.
 3d, at 661 (citations omitted).
 This evidence fails to make a strong showing of bad
faith or improper behavior. Taken together, it proves at
most that the Secretary was predisposed to add a citizen-
ship question to the census and took steps to achieve that
end before settling on the VRA rationale he included in his
memorandum. Perhaps he had reasons for adding the
citizenship question other than the VRA, but by the
Court’s own telling, that does not amount to evidence of
bad faith or improper behavior. Ante, at 24; see Dept. of
Commerce, supra, at ___ (slip op., at 2).
10 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of THOMAS, J.

 The Court thus errs in relying on materials outside the
record to support its holding. And the Court does not
claim that the evidence in the administrative record alone
would prove that the March 2018 memorandum was a
pretext. Given the presumption of regularity, the evidence
discussed above falls far short of establishing that the
VRA rationale did not factor at all into the Secretary’s
decision.
 C
 Even if it were appropriate for the Court to rely on
evidence outside the administrative record, that evidence
still fails to establish pretext. None of the evidence cited
by the Court or the District Court comes close to showing
that the Secretary’s stated rationale—that adding a citi-
zenship question to the 2020 census questionnaire would
“provide . . . data that are not currently available” and
“permit more effective enforcement of the [VRA],” App. to
Pet. for Cert. 548a—did not factor at all into his decision.
 Once again, the evidence cited by the Court suggests at
most that the Secretary had “other unstated reasons” for
reinstating the citizenship question. Ante, at 24. For
example, the Court states that the Secretary’s Director of
Policy “initially attempted to elicit requests for citizenship
data from the Department of Homeland Security and
DOJ’s Executive Office for Immigration Review.” Ante, at
27. But this hardly shows pretext. It simply suggests that
the Director believed that citizenship information could be
useful in tackling problems related to national security
and illegal immigration—a view that would also explain
why the Secretary might not have been “considering VRA
enforcement” early on. Ibid.; see also American Commu-
nity Survey, Why We Ask: Place of Birth, Citizenship and
Year of Entry (2016) (explaining that inquiries about
“place of birth, citizenship, and year of entry” provide
statistics that are “essential for agencies and policy mak-
 Cite as: 588 U. S. ____ (2019) 11

 Opinion of THOMAS, J.

ers setting and evaluating immigration policies and laws,
understanding how different immigrant groups are assim-
ilated, and monitoring against discrimination”), https://
www2 . census . gov / programs - surveys / acs / about / qbyqfact /
2016/Citizenship.pdf (as last visited June 25, 2019).
 The Court emphasizes that the VRA rationale for the
citizenship question originated in the Department of
Commerce, and suggests that DOJ officials unthinkingly
fell in line after the Attorney General was looped into the
process. See ante, at 27. But the Court ignores that the
letter was drafted by the then-Acting Assistant Attorney
General for Civil Rights and reviewed by five other DOJ
attorneys, including the Chief of the DOJ’s Voting Section.
351 F. Supp. 3d, at 554–556. Given the DOJ’s multilayer
review process and its explanation for requesting citizen-
ship data, the Court’s suggestion that the DOJ’s letter was
inadequately vetted or improperly “influence[d]” by the
Department of Commerce is entirely unsupported. Ante,
at 27. In any event, none of this suggests, much less
proves, that the Secretary harbored an unstated belief
that adding the citizenship question would not help en-
force the VRA, or that the VRA rationale otherwise did not
factor at all into his decision. It simply suggests that a
number of executive officials agreed that adding a citizen-
ship question would support VRA enforcement.
 The Court’s other evidence is even further afield. The
Court thinks it telling that the DOJ’s letter included “a
specific request that Commerce collect the [citizenship]
data by means of reinstating a citizenship question on the
census,” rather than a more open-ended “entreaty for
better citizenship data.” Ibid. I do not understand how
the specificity of the DOJ’s letter bears on whether the
Secretary’s rationale was pretextual—particularly since
the letter specifically explained why “census questionnaire
data regarding citizenship, if available, would be more
appropriate for use in redistricting and in [VRA] litiga-
12 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of THOMAS, J.

tion” than existing data. App. to Pet. for Cert. 568a; see
 id., at 567a–568a. Unless the Court is now suggesting
that agency correspondence must comply with the Court’s
subjective, unsupported view of what “might” constitute a
“typical request from another agency,” ante, at 27, the
specificity of the DOJ’s letter is irrelevant. The Court also
points to the DOJ’s decision not to meet with the Census
Bureau “to discuss alternative ways to meet DOJ’s stated
need for improved citizenship data.” Ibid. But the Court
does not explain how the DOJ’s refusal bears on the Secre-
tary’s rationale. Besides, it is easy to understand why
DOJ officials would not be interested in meeting with the
Census Bureau. The meeting would have been with ca-
reer employees whose acknowledged purpose was to talk
the DOJ out of its request. See 351 F. Supp. 3d, at 557 .
Having already considered the issue and explained the
rationale behind the request, it seems at least plausible
that the DOJ officials believed such a meeting would be
unproductive.
 In short, the evidence cited by the Court establishes, at
most, that leadership at both the Department of Com-
merce and the DOJ believed it important—for a variety of
reasons—to include a citizenship question on the census.
 The Court also fails to give credit where it is due. The
Secretary initiated this process inclined to favor what he
called “Option B”—that is, simply “add[ing] a citizenship
question to the decennial census.” App. to Pet. for Cert.
552a. But the Census Bureau favored “Option C”—relying
solely on “administrative records” to supply the infor-
mation needed by the DOJ. Id., at 554a. The Secretary
considered this view and found it a “potentially appealing
solution,” ibid., but concluded that it had shortcomings.
Rather than revert to his original inclination, however, he
“asked the Census Bureau to develop a fourth alternative,
Option D, which would combine Options B and C.” Id., at
555a. And he settled on that solution. Whatever one
 Cite as: 588 U. S. ____ (2019) 13

 Opinion of THOMAS, J.

thinks of the Secretary’s choice, his willingness to change
his mind in light of the Bureau’s feedback belies the idea
that his rationale or decisionmaking process was a
pretext.
 The District Court’s lengthy opinion pointed to other
facts that, in its view, supported a finding of pretext. 351
F. Supp. 3d, at 567–572, 660–664 (discussing the state-
ments, e-mails, acts, and omissions of numerous people
involved in the process). I do not deny that a judge pre-
disposed to distrust the Secretary or the administration
could arrange those facts on a corkboard and—with a jar
of pins and a spool of string—create an eye-catching con-
spiracy web. Cf. id., at 662 (inferring “from the various
ways in which [the Secretary] and his aides acted like
people with something to hide that they did have some-
thing to hide”). But the Court does not rely on this evi-
dence, and rightly so: It casts no doubt on whether the
Secretary’s stated rationale factored into his decision. The
evidence suggests, at most, that the Secretary had multi-
ple reasons for wanting to include the citizenship question
on the census.
 Finally, if there could be any doubt about this conclu-
sion, the presumption of regularity resolves it. Where
there are equally plausible views of the evidence, one of
which involves attributing bad faith to an officer of a
coordinate branch of Government, the presumption com-
pels giving the benefit of the doubt to that officer.
 III
 The Court’s erroneous decision in this case is bad
enough, as it unjustifiably interferes with the 2020 census.
But the implications of today’s decision are broader. With
today’s decision, the Court has opened a Pandora’s box of
pretext-based challenges in administrative law.
 Today’s decision marks the first time the Court has ever
invalidated an agency action as “pretextual.” Having
14 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of THOMAS, J.

taken that step, one thing is certain: This will not be the
last time it is asked to do so. Virtually every significant
agency action is vulnerable to the kinds of allegations the
Court credits today. These decisions regularly involve
coordination with numerous stakeholders and agencies,
involvement at the highest levels of the Executive Branch,
opposition from reluctant agency staff, and—perhaps most
importantly—persons who stand to gain from the action’s
demise. Opponents of future executive actions can be
expected to make full use of the Court’s new approach.
 The 2015 “Open Internet Order” provides a case in
point. In 2015, the Federal Communications Commission
(FCC) adopted a controversial order reclassifying broad-
band Internet access service as a “telecommunications
service” subject to regulation under Title II of the Com-
munications Act. See In re Protecting and Promoting the
Open Internet, 30 FCC Rcd. 5601, 5618 (2015). According
to a dissenting Commissioner, the FCC “flip-flopp[ed]” on
its previous policy not because of a change in facts or legal
understanding, but based on “one reason and one reason
alone. President Obama told us to do so.” Id., at 5921
(statement of Comm’r Pai). His view was supported by a
2016 congressional Report in which Republican Senate
staff concluded that “the FCC bent to the political pres-
sure of the White House” and “failed to live up to stand-
ards of transparency.” Majority Staff Report, Senate
Committee on Homeland Security and Governmental
Affairs, Regulating the Internet: How the White House
Bowled Over FCC Independence, 114th Cong., 1st Sess.,
29 (Comm. Print 2016). The Report cited evidence strik-
ingly similar to that relied upon by the Court here—
including agency-initiated “meetings with certain outside
groups to support” the new result, id., at 3 ; “apparen[t] . . .
concern from the career staff that there was insufficient
notice to the public and affected stakeholders,” id., at 4 ;
and “regula[r] communicatio[n]” between the FCC Chair-
 Cite as: 588 U. S. ____ (2019) 15

 Opinion of THOMAS, J.

man and “presidential advisors,” id., at 25 .
 Under the malleable standard applied by the Court
today, a serious case could be made that the Open Internet
Order should have been invalidated as “pretextual,” re-
gardless of whether any “particular step in the process
stands out as inappropriate or defective.” Ante, at 26. It
is enough, according to the Court, that a judge believes
that the ultimate rationale “seems to have been contrived”
when the evidence is considered “as a whole.” Ante, at
26, 28.
 Now that the Court has opened up this avenue of attack,
opponents of executive actions have strong incentives to
craft narratives that would derail them. Moreover, even if
the effort to invalidate the action is ultimately unsuccess-
ful, the Court’s decision enables partisans to use the
courts to harangue executive officers through depositions,
discovery, delay, and distraction. The Court’s decision
could even implicate separation-of-powers concerns insofar
as it enables judicial interference with the enforcement of
the laws.
 In short, today’s decision is a departure from traditional
principles of administrative law. Hopefully it comes to be
understood as an aberration—a ticket good for this day
and this train only.
 * * *
 Because the Secretary’s decision to reinstate a citizen-
ship question on the 2020 census was legally sound and a
reasoned exercise of his broad discretion, I respectfully
dissent from Part V of the opinion of the Court.
 Cite as: 588 U. S. ____ (2019) 1

 Opinion of BREYER, J.

SUPREME COURT OF THE UNITED STATES
 _________________

 No. 18–966
 _________________

DEPARTMENT OF COMMERCE, ET AL., PETITIONERS
 v. NEW YORK, ET AL.
ON WRIT OF CERTIORARI BEFORE JUDGMENT TO THE UNITED
 STATES COURT OF APPEALS FOR THE SECOND CIRCUIT
 [June 27, 2019]

 JUSTICE BREYER, with whom JUSTICE GINSBURG,
JUSTICE SOTOMAYOR, and JUSTICE KAGAN join, concurring
in part and dissenting in part.
 I join Parts I, II, IV–A, and V of the Court’s opinion
(except as otherwise indicated in this opinion). I dissent,
however, from the conclusion the Court reaches in Part
IV–B. To be more specific, I agree with the Court that the
Secretary of Commerce provided a pretextual reason for
placing a question about citizenship on the short-form
census questionnaire and that a remand to the agency is
appropriate on that ground. But I write separately be-
cause I also believe that the Secretary’s decision to add the
citizenship question was arbitrary and capricious and
therefore violated the Administrative Procedure Act
(APA).
 There is no serious dispute that adding a citizenship
question would diminish the accuracy of the enumeration
of the population—the sole constitutional function of the
census and a task of great practical importance. The
record demonstrates that the question would likely cause
a disproportionate number of noncitizens and Hispanics to
go uncounted in the upcoming census. That, in turn,
would create a risk that some States would wrongfully
lose a congressional representative and funding for a host
of federal programs. And, the Secretary was told, the
2 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of BREYER, J.

adverse consequences would fall most heavily on minority
communities. The Secretary decided to ask the question
anyway, citing a need for more accurate citizenship data.
But the evidence indicated that asking the question would
produce citizenship data that is less accurate, not more.
And the reason the Secretary gave for needing better
citizenship data in the first place—to help enforce the
Voting Rights Act of 1965—was not convincing.
 In short, the Secretary’s decision to add a citizenship
question created a severe risk of harmful consequences,
yet he did not adequately consider whether the question
was necessary or whether it was an appropriate means of
achieving his stated goal. The Secretary thus failed to
“articulate a satisfactory explanation” for his decision,
“failed to consider . . . important aspect[s] of the problem,”
and “offered an explanation for [his] decision that runs
counter to the evidence,” all in violation of the APA. Motor
Vehicle Mfrs. Assn. of United States, Inc. v. State Farm
Mut. Automobile Ins. Co., 463 U. S. 29, 43 (1983). These
failures, in my view, risked undermining public confidence
in the integrity of our democratic system itself. I would
therefore hold that the Secretary’s decision—whether
pretextual or not—was arbitrary, capricious, and an abuse
of discretion.
 I
 A
 Three sets of laws determine the legal outcome of this
case. First, the Constitution requires an “actual Enumer-
ation” of the “whole number of persons in each State”
every 10 years. Art. I, §2, cl. 3; Amdt. 14, §2. It does so in
order to “provide a basis for apportioning representatives
among the states in the Congress.” Baldrige v. Shapiro,
 455 U. S. 345, 353 (1982); see also Art. I, §2, cl. 3. The
inclusion of this provision in the Constitution itself under-
scores the importance of conducting an accurate census.
 Cite as: 588 U. S. ____ (2019) 3

 Opinion of BREYER, J.

See Utah v. Evans, 536 U. S. 452, 478 (2002) (recognizing
“a strong constitutional interest in [the] accuracy” of the
enumeration).
 Second, the Census Act contains two directives that
constrain the Secretary’s ability to add questions to the
census. Section 195 says that the Secretary “shall, if he
considers it feasible,” authorize the use of statistical “sam-
pling” in collecting demographic information. That means
the Secretary must, if feasible, obtain demographic infor-
mation through a survey sent to a sample of households,
rather than through the short-form census questionnaire
to which every household must respond. The other rele-
vant provision, §6(c), says that “[t]o the maximum extent
possible and consistent with the kind, timeliness, quality
and scope of the statistics required, the Secretary shall
acquire and use information available” from administra-
tive sources “instead of conducting direct inquiries.”
(Emphasis added.) These provisions, taken together,
reflect a congressional preference for keeping the short
form short, so that it does not burden recipients and
thereby discourage them from responding.
 Third, the APA prohibits administrative agencies from
making choices that are “arbitrary, capricious, an abuse of
discretion, or otherwise not in accordance with law.” 5
U. S. C. §706 (2)(A). We have said that courts, in applying
this provision, must decide “whether the decision was
based on a consideration of the relevant factors and
whether there has been a clear error of judgment.” Citi-
zens to Preserve Overton Park, Inc. v. Volpe, 401 U. S. 402,
416 (1971). The agency must have “examine[d] the rele-
vant data and articulate[d] a satisfactory explanation for
its action[,] including a ‘rational connection between the
facts found and the choice made.’ ” State Farm, 463 U. S.,
at 43 . An agency ordinarily fails to meet this standard if
it has “failed to consider an important aspect of the prob-
lem, offered an explanation for its decision that runs
4 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of BREYER, J.

counter to the evidence before the agency, or is so implau-
sible that it could not be ascribed to a difference in view or
the product of agency expertise.” Ibid.
 Courts do not apply these principles of administrative
law mechanically. Rather, they take into account, for
example, the nature and importance of the particular
decision, the relevance and importance of missing infor-
mation, and the inadequacies of a particular explanation
in light of their importance. The Federal Government
makes tens of thousands, perhaps millions, of administra-
tive decisions each year. And courts would be wrong to
expect or insist upon administrative perfection. But here,
the Enumeration Clause, the Census Act, and the nature
of the risks created by the agency’s decision all make clear
that the decision before us is highly important to the
proper functioning of our democratic system. It is there-
fore particularly important that courts here not overlook
an agency’s (1) failure to consider serious risks of harm,
(2) failure to explain its refusal to minimize those risks, or
(3) failure to link its conclusion to available evidence. My
view, like that of the District Court, is that the agency
here failed on all three counts.
 B
 A brief history of how the census has worked over the
years will help the reader understand some of the short-
comings of the Secretary’s decisionmaking process. The
Framers wrote into the Constitution a mandate to conduct
an “actual Enumeration” of the population every 10 years.
Art. I, §2, cl. 3. They did so for good reason. The purpose
of the census is to “provide a basis for apportioning repre-
sentatives among the states in the Congress,” Baldrige,
 455 U. S., at 353 , ensuring that “comparative state politi-
cal power in the House . . . reflect[s] comparative popula-
tion,” Evans, 536 U. S., at 477 . The Framers required an
actual count of every resident to “limit political chicanery”
 Cite as: 588 U. S. ____ (2019) 5

 Opinion of BREYER, J.

and to prevent the census count from being “skewed for
political . . . purposes.” Id., at 500 (THOMAS, J., concurring
in part and dissenting in part).
 Throughout most of the Nation’s history, the Federal
Government used enumerators, often trained census
takers, to conduct the census by going door to door. The
enumerators would ask a host of questions, including
place of birth, citizenship, and others. But after the 1950
census, the Bureau began to change its approach. Post-
census studies revealed that the census had failed to count
more than 5 million people and that the undercount dis-
proportionately affected members of minority groups. See
M. Anderson, The American Census: A Social History
201−202 (1988); Brief for Historians and Social Scientists
as Amici Curiae 15. Studies showed that statistical
sampling would produce higher quality data. Anderson,
American Census, at 201.
 Beginning with the 1960 census, the Bureau conse-
quently divided its questioning into a short form and a
long form. The short form contained a list of questions—a
short list—that the census would ask of every household.
That list included basic demographic questions like sex,
age, race, and marital status. The short form did not
include, and has never included, a question about citizen-
ship. See ibid.; Dept. of Commerce, U. S. Census Bureau,
Measuring America: The Decennial Censuses From 1790
to 2000, p. 128 (2002). By way of contrast, the long form
set forth a host of questions that would be asked of only a
sample of households. In 1960, the long form was sent to
one in every four households; in subsequent years, it was
sent to approximately one in every six. See 351 F. Supp.
3d 502, 520 (SDNY 2019). And it was more recently re-
placed by the American Community Survey (ACS), which
is sent to approximately 1 in 38 households each year.
The long form (and now the ACS) has often included a
question about citizenship.
6 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of BREYER, J.

 In 1970, the Census Bureau made another important
change to the census. It significantly reduced its reliance
upon in-person enumerators. See Anderson, supra, at 206.
Instead, it sent nearly all households a questionnaire by
mail. Most households received the short form, and a
small sample received the long form. Instructions on the
form told each household to fill out the questionnaire and
return it to the Census Bureau by mail. Enumerators
would follow up with households that did not return the
questionnaire.
 To maximize accuracy and minimize cost, the Bureau
tried to bring about the highest possible “self-response”
rate, i.e., to encourage as many households as possible to
respond by mail. For that reason, it tried to keep the
short form as short as possible. And it consistently op-
posed placing a citizenship question on that form. It
feared that adding a question about citizenship would
“inevitably jeopardize the overall accuracy of the popula-
tion count,” partly because of added response burden but
also because, as it explained, noncitizens faced with a
citizenship question would be less likely to respond due to
fears of “the information being used against them.” Fed-
eration for Am. Immigration Reform v. Klutznick, 486
F. Supp. 564, 568 (DC 1980).
 Likely for similar reasons, Congress amended the Cen-
sus Act in 1976, enacting the two statutory provisions to
which I previously referred. These two provisions, 13
U. S. C. §6 (c) and §195, together encourage the Secretary
not to ask demographic questions on the short form if the
information can be obtained either through the long form
or through administrative records.
 II
 With this statutory and historical background, we can
more easily consider the agency decision directly under
review. That decision “reinstate[s] [a] citizenship question
 Cite as: 588 U. S. ____ (2019) 7

 Opinion of BREYER, J.

on the 2020 decennial census.” App. to Pet. for Cert.
549a−550a (Memorandum from Wilbur L. Ross, Jr., Secre-
tary of Commerce, to Karen Dunn Kelley, Under Secretary
for Economic Affairs (Mar. 26, 2018)). The agency’s deci-
sion memorandum provided one and only one reason for
making that decision—namely, that the question was
“necessary to provide complete and accurate data in re-
sponse to” a request from the Department of Justice
(DOJ). Id., at 562a. The DOJ had requested the citizen-
ship question for “use [in] . . . determining violations of
Section 2 of the Voting Rights Act.” Id., at 548a.
 The decision memorandum adds that the agency had not
been able to “determine definitively how inclusion of a
citizenship question on the decennial census will impact
responsiveness. However, even if there is some impact on
responses, the value of more complete and accurate data
derived from surveying the entire population outweighs
such concerns.” Id., at 562a. The Secretary’s decision
thus rests upon a weighing of potentially adverse conse-
quences (diminished responses and a less accurate census
count) against potentially offsetting advantages (better
citizenship data). In my view, however, the Secretary did
not make reasonable decisions about these potential costs
and benefits in light of the administrative record.
 A
 Consider first the Secretary’s conclusion that he was
“not able to determine definitively how inclusion of a
citizenship question on the decennial census will impact
responsiveness.” Ibid. Insofar as this statement implies
that adding the citizenship question is unlikely to affect
“responsiveness” very much (or perhaps at all), the evi-
dence in the record indicates the contrary.
 1
 The administrative record includes repeated Census
8 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of BREYER, J.

Bureau statements that adding the question would pro-
duce a less accurate count because noncitizens and His-
panics would be less likely to respond to the questionnaire.
See App. 105, 109–112, 158. The Census Bureau’s chief
scientist said specifically that adding the question would
have “an adverse impact on self-response and, as a result,
on the accuracy and quality of the 2020 Census.” Id., at
109 . And the chief scientist backed this statement up by
pointing to “[t]hree distinct analyses.” Ibid.
 The first analysis compared nonresponse rates for the
short-form census questionnaire (which did not include a
citizenship question) to nonresponse rates for the ACS
(which did). Obviously, more people fail to respond to the
ACS than to the short form. Yet taking into account the
fact that the nonresponse rate will be greater for the ACS
than for the short form, the Bureau found that the differ-
ence between the two is yet greater for noncitizen house-
holds than for citizen households (by 5.1%, according to
the Bureau). Id., at 111 . This led the Bureau to say that
it was a “reasonable inference” that the presence of the
citizenship question accounted for the difference. Ibid.
 The Bureau conducted two additional studies, both
analyzing data from the ACS. One study looked at re-
sponse rates for particular questions on the ACS. It
showed that the “no answer” rate for the citizenship ques-
tion was “much greater than the comparable rates” for
other census questions (for example, questions about age,
sex, race, and ethnicity). Id., at 110 . And it showed that
the “no answer” rate for the citizenship question was
significantly higher among Hispanics. Id., at 109−110.
The last study examined “break-off ” rates, i.e., the rate at
which respondents stopped answering the questionnaire
upon reaching a particular question. It found that His-
panics were significantly more likely than were non-
Hispanics to stop answering at the point they reached the
citizenship question. Id., at 112 . Together, these two
 Cite as: 588 U. S. ____ (2019) 9

 Opinion of BREYER, J.

studies provided additional support for the Census Bu-
reau’s determination that the citizenship question is likely
to mean disproportionately fewer responses from nonciti-
zens and Hispanics than from others. Ibid.
 Putting numbers upon these study results, the Census
Bureau estimated that adding the question to the short
form would lead to 630,000 additional nonresponding
households. Id., at 114 . That is to say, the question would
cause households covering more than 1 million additional
people to decline to respond to the census. When the
Bureau does not receive a response, it follows up with in-
person interviews in an effort to obtain the missing infor-
mation. The Bureau often interviews what it calls “prox-
ies,” such as family members and neighbors. But this
followup process is subject to error; and the error rate is
much greater than the error rate for self-responses. Ibid.
The Bureau thus explained that lower self-response rates
“degrade data quality” by increasing the risk of error and
leading to hundreds of thousands of fewer correct enumer-
ations. Id., at 113−115. The Bureau added that its esti-
mate was “conservative.” Id., at 115 . It expected “differ-
ences between citizen and noncitizen response rates and
data quality” to be “amplified” in the 2020 census “com-
pared to historical levels.” Ibid. Thus, it explained, “the
decrease in self-response for citizen households in 2020
could be much greater than the 5.1 percentage points [it]
observed during the 2010 Census.” Id., at 115−116. Its
conclusion in light of this evidence was clear. Adding the
citizenship question to the short form was “very likely to
reduce the self-response rate” and thereby “har[m] the
quality of the census count.” Id., at 105, 158 .
 The Census Bureau’s analysis received support from
other submissions. Several States pointed out that noncit-
izens and racial minorities had been undercounted in
every prior census. Administrative Record 1091−1092.
They also drew attention to recent surveys indicating that
10 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of BREYER, J.

noncitizens had significant concerns about the confidenti-
ality of census responses. Ibid. Former directors of the
Census Bureau wrote that adding the citizenship question
so late in the process “would put the accuracy of the enu-
meration and success of the census in all communities at
grave risk.” Id., at 1057 . The American Sociological Asso-
ciation and Census Scientific Advisory Committee echoed
these warnings. See id., at 787 , 794−795. On the other
hand, the Secretary received submissions by other groups
that supported adding the question. See, e.g., id., at
1178−1179, 1206, 1276. But as far as I can tell (or as far
as the arguments made here and in the District Court
inform the matter), none of these latter submissions
significantly added to, or detracted from, the Census
Bureau’s submissions in respect to the question’s likely
impact on response rates.
 2
 The Secretary’s decision memorandum reached a quite
different conclusion from the Census Bureau. The memo-
randum conceded that “a lower response rate would lead
to . . . less accurate responses.” App. to Pet. for Cert. 556a.
But it concluded that neither the Census Bureau nor any
stakeholders had provided “definitive, empirical support”
for the proposition that the citizenship question would
reduce response rates. Id., at 554a. The memorandum
relied for that conclusion upon a number of considerations,
but each is contradicted by the record.
 The memorandum first pointed to perceived shortcom-
ings in the Census Bureau’s analysis of nonresponse rates.
It noted that response rates are generally lower overall for
the long form and ACS than they are for the short form.
 Id., at 552a−554a. But the Bureau explained that its
analysis accounted for this consideration, see App. 111,
and no one has given us reason to think the contrary. The
Secretary also noted that the Bureau “was not able to
 Cite as: 588 U. S. ____ (2019) 11

 Opinion of BREYER, J.

isolate what percentage of [the] decline was caused by the
inclusion of a citizenship question rather than some other
aspect of the long form survey.” App. to Pet. for Cert.
554a. But the Bureau said attributing the decline to the
citizenship question was a “reasonable inference,” App.
111, and again, nothing in the record contradicted the
Bureau’s judgment. And later analyses have borne out the
Bureau’s judgment that the citizenship question contrib-
utes to the decline in self-response. See, e.g., id., at
1002−1006, 1008 (August 2018 Census Bureau study).
 The memorandum next cast doubt on the Census Bu-
reau’s analysis of the rate at which people responded to
particular questions on the ACS. It noted that the “no
answer” rate to the citizenship question was comparable to
the “no answer” rate for other questions on the ACS, in-
cluding educational attainment, income, and property
insurance. App. to Pet. for Cert. 553a. But as discussed
above, the Bureau found it significant that the “no an-
swer” rate for the citizenship question was “much greater”
than the “no answer” rate for the other questions that
appear on the short form—that is, the form on which the
citizenship question would appear. App. 110, 124. The
Secretary offered no reason why the demographic varia-
bles to which he pointed provided a better point of
comparison.
 Finally, the memorandum relied on information provided
by two outside stakeholders. The first was a study con-
ducted by the private survey company Nielsen, in which
questions about place of birth and time of arrival had not
led to any appreciable decrease in the response rate. App.
to Pet. for Cert. 552a. But Nielsen, which in fact urged
the Secretary not to add the question, stated that its re-
spondents (unlike census respondents) were paid to re-
spond, and it is consequently not surprising that they did
so. Administrative Record 1276. The memo-
randum also cited statements by former Census Bureau
12 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of BREYER, J.

officials suggesting that empirical evidence about the
question’s potential impact on response rates was “lim-
ited.” App. to Pet. for Cert. 558a−559a; see also id., at
552a. But there was no reason to expect the former offi-
cials to provide more extensive empirical evidence as to a
citizenship question when they were not privy to the
internal Bureau analyses on this question. And, like
Nielsen, the former officials strongly urged the Secretary
not to ask the question. See Administrative Record 1057.
 The upshot is that the Secretary received evidence of a
likely drop in census accuracy by a number somewhere in
the hundreds of thousands, and he received nothing signif-
icant to the contrary. The Secretary pointed out that the
Census Bureau’s information was uncertain, i.e., not
“definitive.” But that is not a satisfactory answer. Few
public-policy-related statistical studies of risks (say, of
many health or safety matters) are definitive. As the
Court explained in State Farm, “[i]t is not infrequent that
the available data do not settle a regulatory issue, and the
agency must then exercise its judgment in moving from
the facts and probabilities on the record to a policy conclu-
sion.” 463 U. S., at 52 . But an agency confronted with
this situation cannot “merely recite the terms ‘substantial
uncertainty’ as a justification for its actions.” Ibid. In-
stead, it “must explain the evidence which is available”
and typically must offer a reasoned explanation for taking
action without “engaging in a search for further evidence.”
 Ibid.
 The Secretary did not do so here. He did not explain
why he made the decision to add the question without
following the Bureau’s ordinary practice of extensively
testing proposed changes to the census questionnaire. See
App. 624−630, 641 (discussing testing process); see also,
e.g., Brief for Former Census Bureau Directors as Amici
Curiae 17−21 (discussing prior examples of questions that
the Bureau decided not to add after many years of pretest-
 Cite as: 588 U. S. ____ (2019) 13

 Opinion of BREYER, J.

ing). Without that testing, the Secretary could not treat
the Bureau’s expert opinions and its experience with the
relevant surveys as worthless merely because its conclu-
sions were not precise. The Bureau’s opinions were
properly considered as evidence of likelihoods, probabili-
ties, or risks.
 As noted above, the consequences of mistakes in the
census count, of even a few hundred thousand, are grave.
Differences of a few thousand people, as between one State
and another, can mean a loss or gain of a congressional
seat—a matter of great consequence to a State. See 351
F. Supp. 3d, at 594 . And similar small differences can
make a large difference to the allocation of federal funds
among competing state programs. Id., at 596−597; see
also Baldrige, 455 U. S., at 353−354, n. 9. If near-absolute
certainty is what the Secretary meant by “definitive,” that
insistence would itself be arbitrary in light of the constitu-
tional and statutory consequences at stake. And if the
Secretary instead meant that the evidence does not indi-
cate a serious risk of a less accurate count, that conclusion
does not find support in the record.
 B
 Now consider the Secretary’s conclusion that, even if
adding a citizenship question diminishes the accuracy of
the enumeration, “the value of more complete and accu-
rate data derived from surveying the entire population
outweighs . . . concerns” about diminished accuracy. App.
to Pet. for Cert. 562a (emphasis added). That conclusion
was also arbitrary. The administrative record indicates
that adding a citizenship question to the short form would
produce less “complete and accurate data,” not more.
 1
 The Census Bureau informed the Secretary that, for
about 90% of the population, accurate citizenship data is
14 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of BREYER, J.

available from administrative records maintained by the
Social Security Administration and Internal Revenue
Service. App. 146. The Bureau further informed the
Secretary that it had “high confidence” that it could develop
a statistical model that would accurately impute citizen-
ship status for the remaining 10% of the population. Ibid.
The Bureau stated that these methods alone—using exist-
ing administrative records for 90% of the population and
statistical modeling for the remaining 10%—would yield
more accurate citizenship data than also asking a citizen-
ship question. Id., at 159. How could that be so? The
answer is somewhat technical but readily understandable.
 First, consider the 90% of the population (about 295
million people) as to whom administrative records are
available. The Government agrees that using these
administrative records would provide highly reliable
information about citizenship, because the records “re-
quire proof of citizenship.” Id., at 117. By contrast, if
responses to a citizenship question were used for this
group, the Census Bureau predicted without contradiction
that about one-third of the noncitizens in this group who
respond would answer the question untruthfully, claiming
to be citizens when they are not. Id., at 147. Those incor-
rect answers—about 9.5 million in total—would conflict
with the administrative records on file for those nonciti-
zens. And what would the Census Bureau do with the
conflicting data? If it accepts the answer to the citizenship
question as determinative, it will have less accurate data.
If it accepts the citizenship data from administrative
records as determinative, asking the question will have
served no purpose.
 Thus, as to 295 million people—the overwhelming ma-
jority of the population—asking the citizenship question
would at best add nothing at all. I say “at best” because,
for one thing, the Census Bureau informed the Secretary
that asking the question would produce 1 million more
 Cite as: 588 U. S. ____ (2019) 15

 Opinion of BREYER, J.

people who could not be linked to administrative records,
which in turn would require the Census Bureau to resort
to a less accurate source of citizenship data for these
people. See id., at 147−149; see also 351 F. Supp. 3d, at
538−539. For another, the policy of the Census Bureau
has always been to use census responses rather than
administrative records in cases where the two conflict.
App. 147. In this case, that practice would mean accepting
9.5 million inaccurate responses even though accurate
administrative records are available. See ibid. The Census
Bureau could perhaps change that practice, but the Secre-
tary’s decision memorandum said nothing about the
matter. It did not address the problem.
 Second, consider the remaining 10% of the population
(about 35 million people) for whom the Government lacks
administrative records. The question here is which ap-
proach would yield the most “complete and accurate”
citizenship data for this group—adding a citizenship
question or using statistical modeling alone? To answer
this question, we must further divide this group into two
categories—those who would respond to the citizenship
question if it were asked and those who would not.
 Start with the category of about 22 million people who
would answer a citizenship question if it were asked.
Would their answers regarding citizenship be more accu-
rate than citizenship data produced by statistical model-
ing? The Census Bureau said no. That is because many of
the noncitizens in this group would answer the question
falsely, resulting in an estimated 500,000 inaccurate
answers. See id., at 148 . And those who answer the
question falsely would be commingled, perhaps randomly,
with those who answer it correctly, thereby casting doubt
on the answers of all 22 million, with no way of knowing
which answers are correct and which are false. By con-
trast, the Bureau believed that it could develop a statisti-
cal model that would produce more accurate citizenship
16 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of BREYER, J.

data than these census responses. The Bureau therefore
informed the Secretary that it could do better. As the
Bureau’s chief scientist explained, although “[o]ne might
think” that asking the question “could help fill the . . .
gaps” in the administrative records, the data did not sup-
port that assumption. Id., at 157 . Instead, he explained,
responses to the citizenship question “may not be reliable,”
which “calls into question their ability to improve upon”
the Bureau’s statistical modeling process. Ibid.
 Next, turn to the more than 13 million remaining people
who would not answer the citizenship question even if it
were asked. As to this category, the Census Bureau would
still need to use statistical modeling to obtain citizenship
data, because there would be no census response to use
instead. Hence, asking the citizenship question would add
nothing at all as to this group. To the contrary, as the
Government concedes, asking the question would reduce
the accuracy of the citizenship data for this group, because
the relatively inaccurate answers to the citizenship ques-
tion would diminish the overall accuracy of the Census
Bureau’s statistical model. See Brief for Petitioners 34
(conceding that the Census Bureau model will be “highe[r]
quality” without the question than with it); 351 F. Supp.
3d, at 640 (explaining that asking the question would
“corrup[t] . . . the data generated by extrapolating from
self-responses through imputation”).
 In sum, in respect to the 295 million persons for whom
administrative records exist, asking the question on the
short form would, at best, be no improvement over using
administrative records alone. And in respect to the re-
maining 35 million people for whom no administrative
records exist, asking the question would be no better, and
in some respects would be worse, than using statistical
modeling. The Census Bureau therefore told the Secre-
tary that asking the citizenship question, even in addition
to using administrative records, “would result in poorer
 Cite as: 588 U. S. ____ (2019) 17

 Opinion of BREYER, J.

quality citizenship data” than using administrative rec-
ords alone, and would “still have all the negative cost and
quality implications” of asking the citizenship question.
App. 159. I could find no evidence contradicting that
prediction.
 2
 If my description of the record is correct, it raises a
serious legal problem. How can an agency support the
decision to add a question to the short form, thereby risk-
ing a significant undercount of the population, on the
ground that it will improve the accuracy of citizenship
data, when in fact the evidence indicates that adding the
question will harm the accuracy of citizenship data? Of
course it cannot. But, as I have just said, I have not been
able to find evidence to suggest that adding the question
would result in more accurate citizenship data. Neither
could the District Court. After reviewing the record in
detail, the District Court found that “all of the relevant
evidence before Secretary Ross—all of it—demonstrated
that using administrative records . . . would actually
produce more accurate [citizenship] data than adding a
citizenship question to the census.” 351 F. Supp. 3d,
at 650 .
 What consideration did the Secretary give to this prob-
lem? He stated simply that “[a]sking the citizenship
question of 100 percent of the population gives each re-
spondent the opportunity to provide an answer,” which
“may eliminate the need for the Census Bureau to have to
impute an answer for millions of people.” App. to Pet. for
Cert. 556a. He therefore must have assumed, sub silentio,
exactly what the Census Bureau experts urged him not to
assume—that answers to the citizenship question would
be more accurate than statistical modeling. And he ig-
nored the undisputed respects in which asking the ques-
tion would make the existing data less accurate. Other
18 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of BREYER, J.

than his assumption, the Secretary said nothing, absolutely
nothing, to suggest a reasoned basis for disagreeing with
the Bureau’s expert statistical judgment.
 The Government now maintains that the Secretary
reasonably discounted the Census Bureau’s recommenda-
tion because it was based on an untested prediction about
the accuracy of its model. But this is not a case in which
the Secretary was presented with a policy choice between
two reasonable but uncertain options. For one thing, the
record is much less uncertain than the Government
acknowledges. Although it is true that the Census Bureau
at one point told the Secretary that it could not “quantify
the relative magnitude of the errors across the alterna-
tives at this time,” App. 148, it unequivocally stated that
asking the question “would result in poorer quality citizen-
ship data” than omitting it, id., at 159 (emphasis added).
Thus, even if the Bureau could not “quantify” the relative
accuracy of the options, it could and did conclude that one
option was likely more accurate than the other. Even in
the face of some uncertainty, where all available evi-
dence indicates that one option is better than the other,
it is unreasonable to choose the worse option without
explanation.
 For another thing, to the extent the record reflects some
uncertainty regarding the accuracy of the Census Bureau’s
statistical model, that is because the model needed to be
“developed and tested” before it could be employed. Id., at
146. But the Secretary made his decision before any such
development or testing could be completed. Having decided
to make an immediate decision rather than wait for test-
ing, the Secretary could not dismiss the Bureau’s predic-
tion about the inadvisability of that decision on the ground
that the prediction reflected likelihoods, probabilities, and
risks rather than certainties.
 Finally, recall that the Census Act requires the Secre-
tary to use administrative records rather than direct
 Cite as: 588 U. S. ____ (2019) 19

 Opinion of BREYER, J.

inquiries to “the maximum extent possible.” 13 U. S. C.
§6 (c). That statutory requirement highlights what should
be obvious: Whether adding a citizenship question to the
short form would produce more accurate citizenship data
is a relevant factor—indeed, a critically important factor—
that the Secretary was required to consider. Here, the
Secretary did not adequately explain why he rejected the
evidence that adding the question would yield less accu-
rate data. He did not even acknowledge that the Census
Act obliged him to use administrative records rather than
asking a question to the extent possible. And he did not
explain how obtaining citizenship data that is no better or
worse than the data otherwise available could justify
jeopardizing the accuracy of the census count.
 In these respects, the Secretary failed to consider “im-
portant aspect[s] of the problem” and “offered an explana-
tion for [his] decision that runs counter to the evidence
before the agency.” State Farm, 463 U. S., at 43 .
 C
 The Secretary’s failure to consider this evidence—that
adding the question would harm the census count in the
interest of obtaining less accurate citizenship data—
provides a sufficient basis for setting the decision aside.
But there is more. The reason that the Secretary provided
for needing more accurate citizenship information in the
first place—to help the DOJ enforce the Voting Rights
Act—is unconvincing.
 The Secretary stated that adding the citizenship ques-
tion was “necessary to provide complete and accurate data
in response to the DOJ request.” App. to Pet. for Cert.
562a. The DOJ’s request in turn asserted that the citizen-
ship data currently available from the ACS was not “ideal”
for enforcing the Voting Rights Act. Id., at 567a. One of
the DOJ’s principal complaints was that ACS data is
reported for groups of census blocks rather than for each
20 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of BREYER, J.

census block itself. The DOJ letter stated that adding a
citizenship question could provide it with individual block-
by-block data which, the DOJ maintained, would allow it
to better enforce the Voting Rights Act’s protections for
minority voters. Id., at 568a.
 This rationale is difficult to accept. One obvious prob-
lem is that the DOJ provided no basis to believe that more
precise data would in fact help with Voting Rights Act
enforcement. Congress enacted the Voting Rights Act in
1965—15 years after the census last asked every house-
hold about citizenship. Actions to enforce the Act have
therefore always used citizenship data derived from sam-
pling. Yet I am aware of no one—not in the Department of
Commerce proceeding, in the District Court, or in this
Court—who has provided a single example in which en-
forcement of the Act has suffered due to lack of more
precise citizenship data. Organizations with expertise in
this area tell us that asking the citizenship question will
not help enforce the Act. See, e.g., Brief for NAACP Legal
Defense & Educational Fund, Inc., as Amicus Curiae
30−36. Rather, the question will, by depressing the count
of minority groups, hurt those whom the Act seeks to help.
See, e.g., Brief for Leadership Conference on Civil and
Human Rights et al. as Amici Curiae 21−29.
 Another problem with the Secretary’s rationale is that,
even assuming the DOJ needed more detailed citizenship
data, there were better ways of obtaining the needed data.
The Census Bureau offered to provide the DOJ with data
using administrative records, which, as I have pointed out,
are likely just as accurate, if not more accurate, than
responses to a citizenship question. The Census Bureau
offered to provide this data at the census block level,
which would resolve each of the DOJ’s complaints about
the existing ACS data. See Administrative Record 3289.
But the Secretary rejected this alternative without ex-
plaining why it would not fully respond to the DOJ’s re-
 Cite as: 588 U. S. ____ (2019) 21

 Opinion of BREYER, J.

quest. That failure was particularly problematic given
that the Census Act requires the Secretary to use other
methods of obtaining demographic information if at all
possible. See §§6(c), 195.
 Normally, the Secretary would be entitled to place
considerable weight upon the DOJ’s expertise in matters
involving the Voting Rights Act, but there are strong
reasons for discounting that expertise here. The adminis-
trative record shows that DOJ’s request to add a citizen-
ship question originated not with the DOJ, but with the
Secretary himself. See Administrative Record 3710. The
Voting Rights Act rationale was in fact first proposed by
Commerce Department officials. See ibid. DOJ officials,
for their part, were initially uninterested in obtaining
more detailed citizenship data, App. 414, and they agreed
to request the data only after the Secretary personally
spoke to the Attorney General about the matter, see Ad-
ministrative Record 2651. And when the acting director of
the Census Bureau proposed alternative means of obtain-
ing better citizenship data, DOJ officials declined to meet
to discuss the proposal. See id., at 3460.
 Taken as a whole, the evidence in the administrative
record indicates that the Voting Rights Act rationale
offered by the Secretary was not just unconvincing, but
pretextual. And, as the Court concludes, further evidence
outside the administrative record but present in the trial
record supports the finding of pretext. See Part V, ante.
Among other things, that evidence reveals that the DOJ
official who wrote the letter agreed that adding the ques-
tion “is not necessary for DOJ’s VRA enforcement efforts.”
App. 1113. And that official further acknowledged that he
did not “know whether or not [citizenship] data produced
from responses to the citizenship question . . . will, in fact,
be more precise than the [citizenship] data on which the
DOJ is currently relying for purposes of VRA enforce-
ment.” Id., at 1102.
22 DEPARTMENT OF COMMERCE v. NEW YORK

 Opinion of BREYER, J.

 The Court explains, and I agree, that a court normally
should not “reject an agency’s stated reasons for acting
simply because the agency might also have had other
unstated reasons.” Ante, at 24. But in this case, “the
evidence tells a story that does not match the explanation
the Secretary gave for his decision.” Ante, at 27. This
evidence strongly suggests that the Secretary’s stated
rationale was pretextual. I consequently join Part V of the
Court’s opinion (except insofar as it concludes that the
Secretary’s decision was reasonable apart from the ques-
tion of pretext). And I agree that the pretextual nature of
the Secretary’s decision provides a sufficient basis to
affirm the District Court’s decision to send the matter
back to the agency.
 * * *
 I agree with the Court that the APA gives agencies
broad leeway to carry out their legislatively delegated
duties. And I recognize that Congress has specifically
delegated to the Secretary of Commerce the authority to
conduct a census of the population “in such form and
content as he may determine.” §141(a). But although this
delegation is broad, it is not without limits. The APA
supplies one such limit. In an effort to ensure rational
decisionmaking, the APA prohibits an agency from mak-
ing decisions that are “arbitrary, capricious, [or] an abuse
of discretion.” 5 U. S. C. §706 (2)(A).
 This provision, of course, does not insist that deci-
sionmakers think through every minor aspect of every
problem that they face. But here, the Secretary’s decision
was a major one, potentially affecting the proper workings
of our democratic government and the proper allocation of
hundreds of billions of dollars in federal funds. Cf. ante, at
10. Yet the decision was ill considered in a number of
critically important respects. The Secretary did not give
adequate consideration to issues that should have been
 Cite as: 588 U. S. ____ (2019) 23

 Opinion of BREYER, J.

central to his judgment, such as the high likelihood of an
undercount, the low likelihood that a question would yield
more accurate citizenship data, and the apparent lack of
any need for more accurate citizenship data to begin with.
The Secretary’s failures in considering those critical issues
make his decision unreasonable. They are the kinds of
failures for which, in my view, the APA’s arbitrary and
capricious provision was written.
 As I have said, I agree with the Court’s conclusion as to
pretext and with the decision to send the matter back to
the agency. I do not agree, however, with several of the
Court’s conclusions concerning application of the arbitrary
and capricious standard. In my view, the Secretary’s
decision—whether pretextual or not—was arbitrary,
capricious, and an abuse of his lawfully delegated discre-
tion. I consequently concur in the Court’s judgment to the
extent that it affirms the judgment of the District Court.
 Cite as: 588 U. S. ____ (2019) 1

 ALITO, J., concurring in part
 Opinion and,dissenting
 of ALITO J. in part

SUPREME COURT OF THE UNITED STATES
 _________________

 No. 18–966
 _________________

DEPARTMENT OF COMMERCE, ET AL., PETITIONERS
 v. NEW YORK, ET AL.
ON WRIT OF CERTIORARI BEFORE JUDGMENT TO THE UNITED
 STATES COURT OF APPEALS FOR THE SECOND CIRCUIT
 [June 27, 2019]

 JUSTICE ALITO, concurring in part and dissenting in
part.
 It is a sign of our time that the inclusion of a question
about citizenship on the census has become a subject of
bitter public controversy and has led to today’s regrettable
decision. While the decision to place such a question on
the 2020 census questionnaire is attacked as racist, there
is a broad international consensus that inquiring about
citizenship on a census is not just appropriate but advisa-
ble. No one disputes that it is important to know how
many inhabitants of this country are citizens. 1 And the
most direct way to gather this information is to ask for it
in a census. The United Nations recommends that
a census inquire about citizenship, 2 and many countries

——————
 1 As a 2016 Census Bureau guidance document explained, obtaining
citizenship statistics is “essential for agencies and policy makers setting
and evaluating immigration policies and laws, understanding how
different immigrant groups are assimilated, and monitoring against
discrimination.” Dept. of Commerce, Census Bureau, American Com-
munity Survey, Why We Ask: Place of Birth, Citizenship and Year of
Entry, www2.census. gov /programs - surveys / acs / about / qbyqfact / 2016 /
Citizenship.pdf (all Internet materials as last visited June 25, 2019).
 2 United Nations, Dept. of Economic and Social Affairs Statistics Div.,

Principles and Recommendations for Population and Housing Censuses
163, 191 (rev. 3, 2017).
2 DEPARTMENT OF COMMERCE v. NEW YORK

 ALITO, J., concurring in part
 Opinion and,dissenting
 of ALITO J. in part

do so. 3
 Asking about citizenship on the census also has a rich
history in our country. Every census, from the very first
one in 1790 to the most recent in 2010, has sought not just
a count of the number of inhabitants but also varying
amounts of additional demographic information. In 1800,
Thomas Jefferson, as president of the American Philo-
sophical Society, signed a letter to Congress asking for the
inclusion on the census of questions regarding “ ‘the re-
spective numbers of native citizens, citizens of foreign
birth, and of aliens’ ” “ ‘for the purpose . . . of more exactly
distinguishing the increase of population by birth and
immigration.’ ” C. Wright, History and Growth of the
United States Census (prepared for the Senate Committee
on the Census), S. Doc. No. 194, 56th Cong., 1st Sess., 19
(1900). In 1820, John Quincy Adams, as Secretary of
State, was responsible for conducting the census, and
consistent with the 1820 Census Act, he instructed the
marshals who were charged with gathering the infor-
mation to ask about citizenship. 4 In 1830, when Martin
Van Buren was Secretary of State, a question about citi-
zenship was again included. 5 With the exception of the
census of 1840, at least some portion of the population was
asked a question about citizenship as part of the census
through 2000, after which the question was moved to the
American Community Survey, which is sent to only a
small fraction of the population. All these census inquiries
——————
 3 See, e.g., Brief for Petitioners 29 (“ ‘[O]ther major democracies in-

quire about citizenship on their census, including Australia, Canada,
France, Germany, Indonesia, Ireland, Mexico, Spain, and the United
Kingdom, to name a few’ ” (quoting App. to Pet. for Cert. 561a)).
 4 See Act of Mar. 14, 1820, ch. 24, 3 Stat. 550 ; Wright, History and

Growth of the United States Census, S. Doc. No. 194, 56th Cong., 1st
Sess., 133–137.
 5 See Dept. of Commerce, Census Bureau, History: 1830 Census Ques-

tionnaire, https: / /www . census . gov /history /www /through _ the _ decades /
questionnaires/1830_2.html.
 Cite as: 588 U. S. ____ (2019) 3

 ALITO, J., concurring in part
 Opinion and,dissenting
 of ALITO J. in part

were made by the Executive pursuant to congressional
authorization. None were reviewed by the courts.
 Now, for the first time, this Court has seen fit to claim a
role with respect to the inclusion of a citizenship question
on the census, and in doing so, the Court has set a danger-
ous precedent, both with regard to the census itself and
with regard to judicial review of all other executive agency
actions. For the reasons ably stated by JUSTICE THOMAS,
see ante, p. ___ (opinion concurring in part and dissenting
in part), today’s decision is either an aberration or a li-
cense for widespread judicial inquiry into the motivations
of Executive Branch officials. If this case is taken as a
model, then any one of the approximately 1,000 district
court judges in this country, upon receiving information
that a controversial agency decision might have been
motivated by some unstated consideration, may order the
questioning of Cabinet officers and other high-ranking
Executive Branch officials, and the judge may then pass
judgment on whether the decision was pretextual. What
Bismarck is reputed to have said about laws and sausages
comes to mind. And that goes for decisionmaking by all
three branches.
 To put the point bluntly, the Federal Judiciary has no
authority to stick its nose into the question whether it is
good policy to include a citizenship question on the census
or whether the reasons given by Secretary Ross for that
decision were his only reasons or his real reasons. Of
course, we may determine whether the decision is consti-
tutional. But under the considerations that typically
guide this Court in the exercise of its power of judicial
review of agency action, we have no authority to decide
whether the Secretary’s decision was rendered in compli-
ance with the Administrative Procedure Act (APA).
 I
 The APA authorizes judicial review of “agency action”
4 DEPARTMENT OF COMMERCE v. NEW YORK

 ALITO, J., concurring in part
 Opinion and,dissenting
 of ALITO J. in part

taken in violation of law, 5 U. S. C. §§706 (2)(A)–(D), but
§701(a)(2) of the APA bars judicial review of agency ac-
tions that are “committed to agency discretion by law.”
Although we have characterized the scope of §701(a)(2) as
“ ‘narrow,’ ” Heckler v. Chaney, 470 U. S. 821, 830 (1985),
there are circumstances in which it applies. And while
our cases recognize a strong presumption in favor of judi-
cial review of agency action, see, e.g., Weyerhaeuser Co. v.
United States Fish and Wildlife Serv., 586 U. S. ___ , ___
(2018) (slip op., at 11), this “is ‘just’ a presumption,” and
like all real presumptions, it may be (and has been) rebut-
ted, Lincoln v. Vigil, 508 U. S. 182, 190 (1993). 6
 In considering whether the general presumption in
favor of judicial review has been rebutted in specific cases,
we have identified factors that are relevant to the inquiry:
whether the text and structure of the relevant statutes
leave a court with any “ ‘meaningful standard against
which to judge the agency’s exercise of discretion,’ ” Web-
ster v. Doe, 486 U. S. 592, 600 (1988) (quoting Heckler,
supra, at 830 ); whether the matter at hand has tradition-
ally been viewed as committed to agency discretion, see
ICC v. Locomotive Engineers, 482 U. S. 270, 282 (1987);
whether the challenged action manifests a “general un-
suitability” for judicial review because it involves a “com-
plicated balancing of a number of factors,” including
judgments regarding the allocation of agency resources or
matters otherwise committed to another branch, Heckler,
supra, at 831–832; and whether judicial review would
produce “disruptive practical consequences,” Southern R.
——————
 6 Because the §701(a)(2) analysis dictates whether APA review may be

had, JUSTICE BREYER’s assertion that the APA “supplies [a] limit” on
the Secretary’s otherwise “broad” delegation, ante, at 22 (opinion
concurring in part and dissenting in part), mistakenly assumes the
answer to the reviewability question. Cf. Heckler v. Chaney, 470 U. S.
821, 828 (1985) (“[B]efore any review at all may be had, a party must
first clear the hurdle of §701(a)”).
 Cite as: 588 U. S. ____ (2019) 5

 ALITO, J., concurring in part
 Opinion and,dissenting
 of ALITO J. in part

Co. v. Seaboard Allied Milling Corp., 442 U. S. 444, 457
(1979) (applying this factor to the reviewability inquiry
under §701(a)(1)).
 Applying those factors, I conclude that the decision of
the Secretary of Commerce to add core demographic ques-
tions to the decennial census questionnaire is committed
to agency discretion by law and therefore may not be
challenged under the APA. 7
 II
 A
 I start with the question whether the relevant statutory
provisions provide any standard that courts can apply in
reviewing the Secretary’s decision to restore a citizenship
question to the census. The provision that directly ad-
dresses this question is 13 U. S. C. §141 (a), the statute
that vests the Secretary with authority to administer the
decennial census. This provision gives the Secretary
unfettered discretion to include on the census questions
about basic demographic characteristics like citizenship.
It begins by providing that the Secretary
 “shall, in the year 1980 and every 10 years thereafter,
 take a decennial census of population . . . in such form
 and content as he may determine, including the use of
 sampling procedures and special surveys.” Ibid.
 (emphasis added).
 The two phrases I have highlighted—“census of popula-
tion” and “in such form and content as he may deter-
mine”—are of immediate importance. A “census of popu-

——————
 7 The Government concedes that courts may review constitutional

challenges to the Secretary’s actions. Cf. Webster v. Doe, 486 U. S. 592,
603 (1988). For the reasons given in the Court’s opinion, see ante, at
11–13, I agree that the only remaining constitutional claim at issue—
respondents’ Enumeration Clause claim—lacks merit and thus does not
constitute a basis for enjoining the addition of the citizenship question.
6 DEPARTMENT OF COMMERCE v. NEW YORK

 ALITO, J., concurring in part
 Opinion and,dissenting
 of ALITO J. in part

lation” is broader than a mere head count. The term is
defined as “a census of population . . . and matters relating
to population.” §141(g) (emphasis added). Because this
definition refers to both “a census of population” and
“matters relating to population,” the latter concept must
include more than a “census of population” in the strict
sense of a head count. And it seems obvious that what
this additional information must include is the sort of
basic demographic information that has long been sought
in the census. So the statute clearly authorizes the Secre-
tary to gather such information.
 The second phrase, “in such form and content as he may
determine,” specifies how this information is to be gath-
ered, namely, by a method having the “form and content”
that the Secretary “may determine.” In other words, this
is left purely to the Secretary’s discretion. A clearer and
less restricted conferral of discretion is hard to imagine.
 It is instructive to compare this delegation of authority
to the statutory language at issue in one of our most well-
known §701(a)(2) cases, Webster v. Doe, 486 U. S. 592 .
There, the relevant statute allowed termination of a Cen-
tral Intelligence Agency employee whenever the Director
“shall deem such termination necessary or advisable in
the interests of the United States.” Id., at 600 (internal
quotation marks omitted and emphasis deleted). Reason-
ing that the statute’s “shall deem” standard “fairly exudes
deference to the Director,” the Court concluded that the
text of the statute “appear[ed] . . . to foreclose the applica-
tion of any meaningful judicial standard of review.” Ibid.
 The §141(a) language discussed above is even more
sweeping than that of the statute in Webster. Unlike the
Census Act, the statute in Webster placed a condition on
the Director’s action—in particular, the requirement that
he terminate an employee only after concluding that doing
so would further the “interests of the United States.” No
such condition applies to the Secretary’s determination
 Cite as: 588 U. S. ____ (2019) 7

 ALITO, J., concurring in part
 Opinion and,dissenting
 of ALITO J. in part

about the form and content of the decennial census, a fact
that distinguishes the statute at issue here from others
this Court has found to fall outside §701(a)(2) and thus
within courts’ power to review. See, e.g., Weyerhaeuser
Co., 586 U. S., at ___ (slip op., at 10) (statute conditioning
agency power to exclude land from critical habitat desig-
nation on agency’s consideration of “ ‘economic impact’ ” of
designation and “ ‘determin[ation] that the benefits of such
exclusion outweigh the benefits of specifying such area as
part of the critical habitat’ ”).
 B
 Those arguing in favor of judicial review contend that
the §141(a) language that I have discussed so far is limited
by language that follows immediately after. That part of
§141(a) states:
 “In connection with any such census [i.e., the decennial
 “census of population”], the Secretary is authorized
 to obtain such other census information as necessary.”
 (Emphasis added.)
 This means, it is argued, that information about citizen-
ship may be obtained by means of the census only if that is
“necessary.” But this argument is clearly wrong. The
information that must be “necessary” (whatever that
means in this context) is “other census information.” That
refers to information other than that obtained in the
“census of population,” and as explained, the term “census
of population” includes not just a head count but other
“matters relating to population,” a category that encom-
passes basic demographic information such as citizenship.
Accordingly, this argument is definitively refuted by the
text of §141. And although it is not necessary to look
beyond that text, it is worth noting that this argument, if
accepted, would require that the term “necessary” be given
a less than strictly literal meaning; otherwise, it would
8 DEPARTMENT OF COMMERCE v. NEW YORK

 ALITO, J., concurring in part
 Opinion and,dissenting
 of ALITO J. in part

run contrary to the broad delegation effected by the first
portion of §141(a) by making it all but impossible for the
Secretary to include on the census anything other than
questions relating to the number of persons living at a
particular address. That would be so because it will often
not be “necessary” to obtain this information via the cen-
sus rather than by some other means.
 C
 Another argument in favor of review relies on 13
U. S. C. §195 , which states:
 “Except for the determination of population for pur-
 poses of apportionment of Representatives in Con-
 gress among the several States, the Secretary shall, if
 he considers it feasible, authorize the use of the statis-
 tical method known as ‘sampling’ in carrying out the
 provisions of this title.”
JUSTICE BREYER, for example, interprets this provision to
mean that “the Secretary must, if feasible, obtain demo-
graphic information through a survey sent to a sample of
households, rather than through the short-form census
questionnaire to which every household must respond.”
Ante, at 3 (opinion concurring in part and dissenting in
part). Under that reading of §195, it is asserted, the pro-
vision sets forth a judicially reviewable limit on the Secre-
tary’s authority to obtain information through direct
inquiries.
 This argument fails to take into account that the cur-
rent version of §195 was enacted as part of the same Act of
Congress that included the present version of §141 8 and
that the two provisions are both parts of a unified scheme
regarding the use of sampling. Section 141, a provision
concerned exclusively with the census, addresses the use

——————
 8 See 90 Stat. 2459 .
 Cite as: 588 U. S. ____ (2019) 9

 ALITO, J., concurring in part
 Opinion and,dissenting
 of ALITO J. in part

of sampling in that particular context. I previously
quoted the relevant language, but I repeat it now so that
it is clearly in mind. Section 141(a) provides that the
Secretary
 “shall, in the year 1980 and every 10 years thereafter,
 take a decennial census of population . . . in such form
 and content as he may determine, including the use of
 sampling procedures and special surveys.” (Emphasis
 added.)
What this means is that the Secretary, in conducting the
“census of population,” has discretion to choose the form
and content of the vehicles used in that project, and
among the methods that he may employ, if he sees fit, are
sampling and special surveys.
 Section 195 is not a census-specific provision, but it does
have one (important) thing to say specifically about the
census: It prohibits the use of sampling “for the determi-
nation of population for purposes of apportionment of
Representatives in Congress.” In this one way, it qualifies
the Secretary’s discretion regarding the “form and con-
tent” of the vehicles used in conducting the “census of
population.” And that is what we meant in Department of
Commerce v. United States House of Representatives, 525
U. S. 316, 338 (1999), when we said that §141(a)’s “broad
grant of authority . . . is informed . . . by the narrower and
more specific §195.” Otherwise, the text of §195 does not
deal specifically with the census. It addresses all the
many information-gathering activities conducted by the
Commerce Department, and as to these, it says that the
Secretary shall use sampling if he deems it “feasible.”
 If §195 were read to mean that no information other
than a head count can be sought by means of a census
questionnaire unless it is not “feasible” to get that infor-
mation by sampling, then there would be little if anything
left of the broad discretion “to use sampling techniques”
10 DEPARTMENT OF COMMERCE v. NEW YORK

 ALITO, J., concurring in part
 Opinion and,dissenting
 of ALITO J. in part

conferred on the Secretary by §141(a). “Feasible” means
“capable of being done, executed, or effected,” Webster’s
Third New International Dictionary 831 (1961), and it is
not clear that the gathering of any core demographic
information is not “capable of being done” by sampling. So
if that were what §195 means, then Congress, in the same
Act, would have given the Secretary discretion to use
sampling in the census “as he may determine” but also
compelled him to use sampling in almost all instances.
That is no way to read the provisions of a single Act. A
law’s provisions should be read to work together. See A.
Scalia & B. Garner, Reading Law 180 (2012) (“The provi-
sions of a text should be interpreted in a way that renders
them compatible, not contradictory”). See also, e.g., Par-
ker Drilling Management Services, Ltd. v. Newton, 587
U. S. ___ , ___–___ (2019) (slip op., at 5–6); Star Athletica,
L. L. C. v. Varsity Brands, Inc., 580 U. S. ___ , ___–___
(2017) (slip op., at 6–7); Kawasaki Kisen Kaisha Ltd. v.
Regal-Beloit Corp., 561 U. S. 89, 108 (2010). And if there is
tension between a specific provision, like §141’s instruc-
tion regarding the use of sampling in the decennial census,
and a general one, like §195’s directive regarding the use
of sampling in all data-collection activities, the specific
provision must take precedence. Cf. NLRB v. SW General,
Inc., 580 U. S. ___ , ___ (2017) (slip op., at 14).
 When §§141 and 195 are read in this way, it is easy to
see how they fit together. In using the census to gather
information “relating to population” for any use other than
the actual enumeration, the Secretary may use sampling
“as he may determine.” In conducting all the Depart-
ment’s efforts to collect data by other means, he may
authorize the use of sampling if he thinks that is “feasi-
ble.” The upshot for present purposes is that §195 does
not require the “counterintuitive resul[t]” of barring the
Secretary from including on the census questionnaire the
kinds of basic demographic questions that have been
 Cite as: 588 U. S. ____ (2019) 11

 ALITO, J., concurring in part
 Opinion and,dissenting
 of ALITO J. in part

asked as part of every census in U. S. history. RJR
Nabisco, Inc. v. European Community, 579 U. S. ___ , ___
(2016) (slip op., at 15).
 D
 One additional provision, 13 U. S. C. §6 (c), 9 requires
close consideration. This provision, which was enacted in
1976 in the same Act as §§141(a) and 195, has three sub-
sections. Subsection (a) provides that the Secretary may
call on other components of the Federal Government to
obtain information that is “pertinent to” the Department’s
work. Subsection (b) authorizes the Secretary to “acquire,
by purchase or otherwise” from state and local govern-
ments and private sources “such copies of records, reports,
and other material as may be required for the efficient and
economical conduct of the censuses and surveys provided
for in this title.” Finally, subsection (c) provides:
 “To the maximum extent possible and consistent with
 the kind, timeliness, quality and scope of the statistics
 required, the Secretary shall acquire and use infor-
 mation available from any source referred to in sub-
 section (a) or (b) of this section instead of conducting
——————
 9 Section 6 states:
 “(a) The Secretary, whenever he considers it advisable, may call upon
any other department, agency, or establishment of the Federal Gov-
ernment, or of the government of the District of Columbia, for infor-
mation pertinent to the work provided for in this title.
 “(b) The Secretary may acquire, by purchase or otherwise, from
States, counties, cities, or other units of government, or their instru-
mentalities, or from private persons and agencies, such copies of
records, reports, and other material as may be required for the efficient
and economical conduct of the censuses and surveys provided for in this
title.
 “(c) To the maximum extent possible and consistent with the kind,
timeliness, quality and scope of the statistics required, the Secretary
shall acquire and use information available from any source referred to
in subsection (a) or (b) of this section instead of conducting direct
inquiries.”
12 DEPARTMENT OF COMMERCE v. NEW YORK

 ALITO, J., concurring in part
 Opinion and,dissenting
 of ALITO J. in part

 direct inquiries.”
 The District Court interpreted subsection (c) to mean
that the Secretary must turn to another federal agency or
outside source for demographic information (rather than
seeking the information on the census) unless doing so
would not be “possible” or “consistent with the kind, time-
liness, quality and scope of the statistics required.” This
argument fails for reasons similar to those that sank the
§195 argument just discussed. Section 6(c) is not a census-
specific provision but instead applies generally to all the
Commerce Department’s information-gathering activities.
If it is read to apply to the “census of population,” it cannot
be reconciled with §141(a), which, as noted, broadly au-
thorizes the Secretary to use that vehicle for obtaining
information “relating to population,” i.e., core demographic
information. If §6(c) applied to the gathering of such
information, it would make it hard to justify the inclusion
of any demographic questions on the census, even though
this has been done since 1790. (Is it not possible to get
information about age and sex, for example, from any
outside source (or combination of sources), even if the
Department offers to acquire it from a private source by
purchase?) Reading §6(c) to mean what the District Court
thought would turn it into the proverbial elephant stuffed
into a mouse hole. Section 6(c), however, is a decidedly
mouse-like provision. It was enacted with no fanfare and
no real explanation, 10 and remained in the shadows, vir-
——————
 10 The most respondents can muster are snippets from the legislative

history of the 1976 Census Act indicating that §6(c) was enacted to
decrease the Secretary’s use of “direct inquiries” in the interest of
“reducing respondent burden.” H. R. Rep. No. 94–1719, p. 10 (1976).
Even accepting that premise, it simply raises the same question just
discussed—namely, whether Congress’s desire to reduce respondent
burden, as reflected by §6(c), yields to the Secretary’s broad authoriza-
tion in §141(a) to “determine” the “form and content” of any direct
inquiries on the census. Cf. id., at 11 (characterizing §141 as a “provi-
 Cite as: 588 U. S. ____ (2019) 13

 ALITO, J., concurring in part
 Opinion and,dissenting
 of ALITO J. in part

tually unused and unnoticed, for more than 40 years.
 E
 Respondents and the Court cite two other provisions in
support of reviewability, but neither has anything to do
with the issue of putting a citizenship question on the
census. In determining whether statutory provisions
include standards that could provide a basis for judicial
review, it is necessary to focus on the precise claims at
issue, see, e.g., Webster, 486 U. S., at 601–602 (distinguish-
ing between statutory and constitutional claims); Locomo-
tive Engineers, 482 U. S., at 277–279 (parsing claims
under different prongs of reopener statute); Heckler, 470
U. S., at 836 (rejecting as “irrelevant” to the agency deci-
sion at issue two statutory provisions that were argued to
provide “ ‘law to apply’ ”). And when viewed in this way,
the remaining statutory provisions cited in support of
reviewability are of no value.
 Respondents point to §141(b), which requires the Secre-
tary to complete the tabulation of total population by
States “within 9 months after the census date” and then to
report the results to the President. That provision sets
out an easily administered deadline, and it has nothing to
do with the content of the census questionnaire.
 Respondents also claim that §141(f) is relevant to the
question of judicial review, but that provision concerns
congressional review. It directs the Secretary to report to
Congress, at specified times, the subjects and questions
that he intends to include on the census. According to
respondents, the Secretary’s compliance with those re-
quirements is judicially reviewable, and that, they
contend, takes the Secretary’s decision to include a citizen-
ship question out from under §701(a)(2).
 Respondents fundamentally misunderstand the signifi-
——————
sio[n] directly related to decennial . . . census”).
14 DEPARTMENT OF COMMERCE v. NEW YORK

 ALITO, J., concurring in part
 Opinion and,dissenting
 of ALITO J. in part

cance of congressional reporting requirements in evaluat-
ing whether a particular agency action is subject to judi-
cial review. Congressional reporting requirements are
“legion in federal law,” Natural Resources Defense Council,
Inc. v. Hodel, 865 F. 2d 288, 317 (CADC 1988), and their
purpose is to permit Congress to monitor and, if it sees fit,
to correct Executive Branch actions to which it objects.
When a congressional reporting requirement “[l]ack[s] a
provision for judicial review,” compliance “by its nature
seems singularly committed to congressional discretion in
measuring the fidelity of the Executive Branch actor to
legislatively mandated requirements.” Id., at 318 . In
other words, it is Congress, not the Judiciary, that is best
situated to determine whether an agency’s responses to
Congress are sufficient and, if not, to “take what it deems
to be the appropriate action.” Id., at 319 .
 In that respect, §141(f) actually cuts against judicial
review. The Constitution gives Congress the authority to
“direct” the “Manner” in which the census is conducted,
and by imposing the §141(f) reporting requirements, Con-
gress retained some of that supervisory authority. It did
not transfer it to the courts. 11
 Respondents protest that congressional review may not
——————
 11 Itis notable that Congress, pursuant to its supervisory authority,
has in some cases limited the particular demographic characteristics
about which the Secretary may require information through census
questionnaires. In §221(c), for example, Congress has dictated that “no
person shall be compelled to disclose information relative to his reli-
gious beliefs or to membership in a religious body.” Similarly, in a
series of appropriation Acts, Congress has specified that “none of the
funds provided in this or any other Act for any fiscal year may be used
for the collection of census data on race identification that does not
include ‘some other race’ as a category.” 123 Stat. 3115 , note following
 13 U. S. C. §5 . Those examples highlight that when Congress wishes to
limit the Secretary’s authority to require responses to particular
demographic questions, it “knows precisely how to do so.” Limelight
Networks, Inc. v. Akamai Technologies, Inc., 572 U. S. 915, 923 (2014).
 Cite as: 588 U. S. ____ (2019) 15

 ALITO, J., concurring in part
 Opinion and,dissenting
 of ALITO J. in part

be enough to guard against a Secretary’s abuses, especially
when the party in control of Congress stands to benefit.
But that complaint simply expresses disagreement with
the Framers’ choice to vest power over the census in a
political body, cf. Baldrige v. Shapiro, 455 U. S. 345 , 347–
348 (1982) (“Under [the] Constitution, responsibility for
conducting the decennial census rests with Congress”),
and the manner in which Congress has chosen to exercise
that power, see Wisconsin v. City of New York, 517 U. S. 1,
19 (1996) (Congress has delegated its “virtually unlimited
discretion” in conducting the census to the Secretary). In
any event, the ability to press constitutional challenges to
the Secretary’s decisions, see n. 7, supra, answers many of
the examples in respondents’ parade of horribles.
 In short, the relevant text of §141(a) “fairly exudes
deference” to the Secretary. Webster, 486 U. S., at 600 .
And no other provision of law cited by respondents or my
colleagues provides any “meaningful judicial standard” for
reviewing the Secretary’s selection of demographic ques-
tions for inclusion on the census. Ibid.
 III
 In addition to requiring an examination of the text and
structure of the relevant statutes, our APA §701(a)(2)
cases look to whether the agency action in question is a
type that has traditionally been viewed as committed to
agency discretion or whether it is instead one that “federal
courts regularly review.” Weyerhaeuser Co., 586 U. S., at
___ (slip op., at 12). In cases where the Court has found
that agency action is committed to agency discretion by
law, an important factor has been the absence of an estab-
lished record of judicial review prior to the adoption of the
APA. See Heckler, 470 U. S., at 832–833 (agency nonen-
forcement); Locomotive Engineers, 482 U. S., at 282 (agency
decision not to reopen final decision based on material
error); Lincoln, 508 U. S., at 192 (agency use of lump-sum
16 DEPARTMENT OF COMMERCE v. NEW YORK

 ALITO, J., concurring in part
 Opinion and,dissenting
 of ALITO J. in part

appropriations).
 Here, there is no relevant record of judicial review. We
are confronted with a practice that reaches back two cen-
turies. The very first census went beyond a mere head
count and gathered additional demographic information,
and during virtually the entire period prior to the enact-
ment of the APA, a citizenship question was asked of
everyone. Notably absent from that long record is any
practice of judicial review of the content of the census.
Indeed, this Court has never before encountered a direct
challenge to a census question. App. to Pet. for Cert. 416a.
And litigation in the lower courts about the census is
sparse and generally of relatively recent vintage.
 Not only is this sort of history significant in all
§701(a)(2) cases, see Locomotive Engineers, supra, at 282 ,
but we have previously stressed the particular “im-
portance of historical practice” when it comes to evaluat-
ing the Secretary’s authority over the census. Wisconsin,
supra, at 21 ; see also ante, at 13 (opinion of the Court).
Moreover, where the relevant question is not whether
review may be had at all, but rather the branch with the
authority to exercise review, the absence of any substan-
tial record of judicial review is especially revealing. See,
e.g., NLRB v. Noel Canning, 573 U. S. 513 , 525 (2014) (it
is “neither new nor controversial” that “longstanding
practice of the government can inform our determination
of what the law is” (internal quotation marks and citation
omitted)); United States v. Midwest Oil Co., 236 U. S. 459,
473 (1915) (“in determining . . . the existence of a power,
weight [is] given to . . . usage”). Thus, the absence of any
real tradition of judicial review of decisions regarding the
content of the census counsels against review in this case.
 In an attempt to show that there is no relevant “tradi-
tion of nonreviewability,” Locomotive Engineers, supra, at
282 , respondents contend that this Court has recently
engaged in review of the “conduct of the census,” Brief for
 Cite as: 588 U. S. ____ (2019) 17

 ALITO, J., concurring in part
 Opinion and,dissenting
 of ALITO J. in part

Government Respondents 26–27. But in none of the cases
they cite did the Court address an APA challenge to the
content of census questions. 12 Some involved constitu-
tional claims about enumeration and apportionment. See
Franklin v. Massachusetts, 505 U. S. 788, 790, 801 (1992)
(constitutional challenge to “method used for counting
federal employees serving overseas” as part of “reappor-
tionment determination”); Wisconsin, 517 U. S., at 20
(constitutional challenge to Secretary’s decision not to
adjust count). Others concerned enforcement of statutes
with specific directives. See Department of Commerce, 525
U. S., at 343 (holding that §195 bars use of “sampling” to
reach actual enumeration for apportionment); Utah v.
Evans, 536 U. S. 452 , 464–465 (2002) (considering
whether statistical method violated §195’s bar on use of
“sampling” in apportionment enumeration). According
to respondents, these cases mean that all the Secretary’s
census-related decisions are suitable for judicial review
and thus fall outside of §701(a)(2), and the Court appar-
ently agrees, rejecting the Government’s §701(a)(2) argu-
ment in part because “[w]e and other courts have enter-
tained both constitutional and statutory challenges to
census-related decisionmaking.” Ante, at 15.
 This argument misses the point of §701(a)(2). The
question under that provision is whether the challenged
action “is committed to agency discretion by law,” not
whether a different action by the same agency is review-
able under the APA, much less whether an action taken by
the same agency can be challenged under the Constitu-
tion. Take the example of Heckler v. Chaney, supra, where
the Court considered whether a particular Food and Drug

——————
 12 The same can be said for the lower court cases on which respond-
ents rely. See, e.g., Brief for Government Respondents 26, and n. 6
(collecting cases, none of which “involved the census questionnaire” or
the Secretary’s selection of questions).
18 DEPARTMENT OF COMMERCE v. NEW YORK

 ALITO, J., concurring in part
 Opinion and,dissenting
 of ALITO J. in part

Administration (FDA) decision was reviewable under the
APA. Many FDA actions are subject to APA review, see,
e.g., Weinberger v. Hynson, Westcott & Dunning, Inc., 412
U. S. 609, 627 (1973), but that did not prevent the Heckler
Court from holding that the particular FDA decision at
issue there fell within §701(a)(2). See also, e.g., Heckler,
supra, at 836–837.
 Respondents and some of their amici contend that the
Secretary’s decision is at least amenable to judicial review
for consistency with the APA’s reasoned-explanation
requirement. See Motor Vehicle Mfrs. Assn. of United
States, Inc. v. State Farm Mut. Automobile Ins. Co., 463
U. S. 29, 43 (1983) (describing requirement). Thus, the
argument goes, even if no statute sets out a standard that
can be used in reviewing the particular agency action in
question, a court may review an agency’s explanation of
the reasons for its action and set it aside if the court finds
those reasons to be arbitrary or irrational.
 This argument would obliterate §701(a)(2). Even if a
statute expressly gave an agency absolute, unrestricted,
unfettered, unlimited, and unqualified discretion with
respect to a particular decision, a court could still review
the agency’s explanation of the reasons for its decision.
That is not what §701(a)(2) means. As we put it previously
in answering a similar argument against application of
§701(a)(2), it is “fals[e]” to suggest “that if the agency gives
a ‘reviewable’ reason for otherwise unreviewable action,
the action becomes reviewable.” Locomotive Engineers,
 482 U. S., at 283 . That is because when an action “is
committed to agency discretion by law,” the Judiciary has
no role to play, even when an agency sets forth “an emi-
nently ‘reviewable’ proposition.” Id., at 282–283.
 IV
 In sum, neither respondents nor my colleagues have
been able to identify any relevant, judicially manageable
 Cite as: 588 U. S. ____ (2019) 19

 ALITO, J., concurring in part
 Opinion and,dissenting
 of ALITO J. in part

limits on the Secretary’s decision to put a core demographic
question back on the census. And without an “adequate
standard of review for such agency action,” id., at 282 ,
courts reviewing decisions about the “form and content” of
the census would inevitably be drawn into second-
guessing the Secretary’s assessment of complicated policy
tradeoffs, 13 another indicator of “general unsuitability” for
judicial review. Heckler, supra, at 831 .
 Indeed, if this litigation is any indication, widespread
judicial review of the Secretary’s conduct of the census will
usher in an era of “disruptive practical consequences,” and
this too weighs against review. Seaboard Allied Milling
Corp., 442 U. S., at 457 . Cf. Tucker v. United States Dept.
of Commerce, 958 F. 2d 1411 , 1418 (CA7 1992) (expressing
doubt about “both the provenance and the practicability”
of allowing judicial review of census-related decisions).
 Respondents protest that the importance of the census
provides a compelling reason to allow APA review. See
also ante, at 22–23 (opinion of BREYER, J.). But this ar-
gument overlooks the fact that the Secretary is account-
able in other ways for census-related decisionmaking. 14 If
the Secretary violates the Constitution or any applicable
statutory provision related to the census, his action is
——————
 13 In determining how the census is to be conducted, the Secretary
must make decisions about a bevy of matters, such as the best way to
count particular persons or categories of persons with an adequate
degree of accuracy (e.g., by face-to-face interviews, telephone calls,
questionnaires to be mailed back, contacts with neighbors, or use of
existing records); the use of followup procedures and other quality
control measures; which persons should be included in which house-
holds; and issues concerning where a person should be enumerated.
These and countless other factors may affect whether an individual
receives or responds to the census questionnaire.
 14 Since the time Secretary Ross publicly announced his intent to add

the citizenship question, “Congress has questioned the Secretary about
his decision in public hearings on several occasions.” Brief for Petition-
ers 50 (collecting examples).
20 DEPARTMENT OF COMMERCE v. NEW YORK

 ALITO, J., concurring in part
 Opinion and,dissenting
 of ALITO J. in part

reviewable. The Secretary is also accountable to Congress
with respect to the administration of the census since he
has that power only because Congress has found it appro-
priate to entrust it to him. And the Secretary is always
answerable to the President, who is, in turn, accountable
to the people.
 * * *
 Throughout our Nation’s history, the Executive Branch
has decided without judicial supervision or interference
whether and, if so, in what form the decennial census
should inquire about the citizenship of the inhabitants of
this country. Whether to put a citizenship question on the
2020 census questionnaire is a question that is committed
by law to the discretion of the Secretary of Commerce and
is therefore exempt from APA review. The District Court
had the authority to decide respondents’ constitutional
claims, but the remainder of their complaint should have
been dismissed.
 I join Parts I, II, III, IV–B, and IV–C 15 of the opinion of
the Court. I do not join the remainder, and insofar as the
Court holds that the Secretary’s decision is reviewable
under the APA, I respectfully dissent.

——————
 15 Although I would hold that the Secretary’s decision is not review-
able under the APA, in the alternative I would conclude that the
decision survives review under the applicable standards. I join Parts
IV–B and IV–C on that understanding.

Chief Justice ROBERTS delivered the opinion of the Court.
 *2561 The Secretary of Commerce decided to reinstate a question about citizenship on the 2020 census questionnaire. A group of plaintiffs challenged that decision on constitutional and statutory grounds. We now decide whether the Secretary violated the Enumeration Clause of the Constitution, the Census Act, or otherwise abused his discretion.

 I

 A

 In order to apportion Members of the House of Representatives among the States, the Constitution requires an "Enumeration" of the population every 10 years, to be made "in such Manner" as Congress "shall by Law direct." Art. I, § 2, cl. 3; Amdt. 14, § 2. In the Census Act, Congress delegated to the Secretary of Commerce the task of conducting the decennial census "in such form and content as he may determine." 13 U. S. C. § 141 (a). The Secretary is aided in that task by the Census Bureau, a statistical agency housed within the Department of Commerce. See §§ 2, 21.

 The population count derived from the census is used not only to apportion representatives but also to allocate federal funds to the States and to draw electoral districts. Wisconsin v. City of New York , 517 U.S. 1 , 5-6, 116 S.Ct. 1091 , 134 L.Ed.2d 167 (1996). The census additionally serves as a means of collecting demographic information, which "is used for such varied purposes as computing federal grant-in-aid benefits, drafting of legislation, urban and regional planning, business planning, and academic and social studies." Baldrige v. Shapiro , 455 U.S. 345 , 353-354, n. 9, 102 S.Ct. 1103 , 71 L.Ed.2d 199 (1982). Over the years, the census has asked questions about (for example) race, sex, age, health, education, occupation, housing, and military service. It has also asked about radio ownership, age at first marriage, and native tongue. The Census Act obliges everyone to answer census questions truthfully and requires the Secretary to keep individual answers confidential, including from other Government agencies. §§ 221, 8(b), 9(a).

 There have been 23 decennial censuses from the first census in 1790 to the most recent in 2010. Every census between 1820 and 2000 (with the exception of 1840) asked at least some of the population about their citizenship or place of birth. Between 1820 and 1950, the question was asked of all households. Between 1960 and 2000, it was asked of about one-fourth to one-sixth of the population. That change was part of a larger effort to simplify the census by asking most people a few basic demographic questions (such as sex, age, race, and marital status) on a short-form questionnaire, while asking a sample of the population more detailed demographic questions on a long-form questionnaire. In explaining the decision to move the citizenship question to the long-form questionnaire, the Census Bureau opined that "general census information on citizenship had become of less importance compared with other possible questions to be included in the census, particularly in view of the *2562 recent statutory requirement for annual alien registration which could provide the Immigration and Naturalization Service, the principal user of such data, with the information it needed." Dept. of Commerce, Bureau of Census, 1960 Censuses of Population and Housing 194 (1966). 1

 In 2010, the year of the latest census, the format changed again. All households received the same questionnaire, which asked about sex, age, race, Hispanic origin, and living arrangements. The more detailed demographic questions previously asked on the long-form questionnaire, including the question about citizenship, were instead asked in the American Community Survey (or ACS), which is sent each year to a rotating sample of about 2.6% of households.

 The Census Bureau and former Bureau officials have resisted occasional proposals to resume asking a citizenship question of everyone, on the ground that doing so would discourage noncitizens from responding to the census and lead to a less accurate count of the total population. See, e.g. , Federation of Am. Immigration Reform v. Klutznick , 486 F.Supp. 564 , 568 (DDC 1980) ("[A]ccording to the Bureau[,] any effort to ascertain citizenship will inevitably jeopardize the overall accuracy of the population count"); Brief for Former Directors of the U. S. Census Bureau as Amici Curiae in Evenwel v. Abbott , O. T. 2014, No. 14-940, p. 25 (inquiring about citizenship would "invariably lead to a lower response rate").

 B

 In March 2018, Secretary of Commerce Wilbur Ross announced in a memo that he had decided to reinstate a question about citizenship on the 2020 decennial census questionnaire. The Secretary stated that he was acting at the request of the Department of Justice (DOJ), which sought improved data about citizen voting-age population for purposes of enforcing the Voting Rights Act (or VRA)-specifically the Act's ban on diluting the influence of minority voters by depriving them of single-member districts in which they can elect their preferred candidates. App. to Pet. for Cert. 548a. DOJ explained that federal courts determine whether a minority group could constitute a majority in a particular district by looking to the citizen voting-age population of the group. According to DOJ, the existing citizenship data from the American Community Survey was not ideal: It was not reported at the level of the census block, the basic component of legislative districting plans; it had substantial margins of error; and it did not align in time with the census-based population counts used to draw legislative districts. DOJ therefore formally requested reinstatement of the citizenship question on the census questionnaire. Id. , at 565a-569a.

 The Secretary's memo explained that the Census Bureau initially analyzed, and the Secretary considered, three possible courses of action. The first was to continue to collect citizenship information in the American Community Survey and attempt to develop a data model that would more accurately estimate citizenship at the census block level. The Secretary rejected that option because the Bureau "did not assert and could not confirm" that such ACS-based data modeling was possible "with a sufficient degree of accuracy." Id. , at 551a.

 *2563 The second option was to reinstate a citizenship question on the decennial census. The Bureau predicted that doing so would discourage some noncitizens from responding to the census. That would necessitate increased "non-response follow up" operations-procedures the Bureau uses to attempt to count people who have not responded to the census-and potentially lead to a less accurate count of the total population.

 Option three was to use administrative records from other agencies, such as the Social Security Administration and Citizenship and Immigration Services, to provide DOJ with citizenship data. The Census Bureau recommended this option, and the Secretary found it a "potentially appealing solution" because the Bureau has long used administrative records to supplement and improve census data. Id. , at 554a. But the Secretary concluded that administrative records alone were inadequate because they were missing for more than 10% of the population.

 The Secretary ultimately asked the Census Bureau to develop a fourth option that would combine options two and three: reinstate a citizenship question on the census questionnaire, and also use the time remaining until the 2020 census to "further enhance" the Bureau's "administrative record data sets, protocols, and statistical models." Id. , at 555a. The memo explained that, in the Secretary's judgment, the fourth option would provide DOJ with the "most complete and accurate" citizen voting-age population data in response to its request. Id. , at 556a.

 The Secretary "carefully considered" the possibility that reinstating a citizenship question would depress the response rate. Ibid . But after evaluating the Bureau's "limited empirical evidence" on the question-evidence drawn from estimated non-response rates to previous American Community Surveys and census questionnaires-the Secretary concluded that it was not possible to "determine definitively" whether inquiring about citizenship in the census would materially affect response rates. Id. , at 557a, 562a. He also noted the long history of the citizenship question on the census, as well as the facts that the United Nations recommends collecting census-based citizenship information, and other major democracies such as Australia, Canada, France, Indonesia, Ireland, Germany, Mexico, Spain, and the United Kingdom inquire about citizenship in their censuses. Altogether, the Secretary determined that "the need for accurate citizenship data and the limited burden that the reinstatement of the citizenship question would impose outweigh fears about a potentially lower response rate." Id. , at 557a.

 C

 Shortly after the Secretary announced his decision, two groups of plaintiffs filed suit in Federal District Court in New York, challenging the decision on several grounds. The first group of plaintiffs included 18 States, the District of Columbia, various counties and cities, and the United States Conference of Mayors. They alleged that the Secretary's decision violated the Enumeration Clause of the Constitution and the requirements of the Administrative Procedure Act. The second group of plaintiffs consisted of several non-governmental organizations that work with immigrant and minority communities. They added an equal protection claim. The District Court consolidated the two cases. Both groups of plaintiffs are respondents here.

 The Government moved to dismiss the lawsuits, arguing that the Secretary's decision was unreviewable and that respondents had failed to state cognizable claims *2564 under the Enumeration Clause and the Equal Protection Clause. The District Court dismissed the Enumeration Clause claim but allowed the other claims to proceed. 315 F.Supp.3d 766 (SDNY 2018).

 In June 2018, the Government submitted to the District Court the Commerce Department's "administrative record": the materials that Secretary Ross considered in making his decision. That record included DOJ's December 2017 letter requesting reinstatement of the citizenship question, as well as several memos from the Census Bureau analyzing the predicted effects of reinstating the question. Shortly thereafter, at DOJ's urging, the Government supplemented the record with a new memo from the Secretary, "intended to provide further background and context regarding" his March 2018 memo. App. to Pet. for Cert. 546a. The supplemental memo stated that the Secretary had begun considering whether to add the citizenship question in early 2017, and had inquired whether DOJ "would support, and if so would request, inclusion of a citizenship question as consistent with and useful for enforcement of the Voting Rights Act." Ibid. According to the Secretary, DOJ "formally" requested reinstatement of the citizenship question after that inquiry. Ibid.

 Respondents argued that the supplemental memo indicated that the Government had submitted an incomplete record of the materials considered by the Secretary. They asked the District Court to compel the Government to complete the administrative record. The court granted that request, and the parties jointly stipulated to the inclusion of more than 12,000 pages of additional materials in the administrative record. Among those materials were emails and other records confirming that the Secretary and his staff began exploring the possibility of reinstating a citizenship question shortly after he was confirmed in early 2017, attempted to elicit requests for citizenship data from other agencies, and eventually persuaded DOJ to request reinstatement of the question for VRA enforcement purposes.

 In addition, respondents asked the court to authorize discovery outside the administrative record. They claimed that such an unusual step was warranted because they had made a strong preliminary showing that the Secretary had acted in bad faith. See Citizens to Preserve Overton Park, Inc. v. Volpe , 401 U.S. 402 , 420, 91 S.Ct. 814 , 28 L.Ed.2d 136 (1971). The court also granted that request, authorizing expert discovery and depositions of certain DOJ and Commerce Department officials.

 In August and September 2018, the District Court issued orders compelling depositions of Secretary Ross and of the Acting Assistant Attorney General for DOJ's Civil Rights Division. We granted the Government's request to stay the Secretary's deposition pending further review, but we declined to stay the Acting AAG's deposition or the other extra-record discovery that the District Court had authorized.

 The District Court held a bench trial and issued findings of fact and conclusions of law on respondents' statutory and equal protection claims. After determining that respondents had standing to sue, the District Court ruled that the Secretary's action was arbitrary and capricious, based on a pretextual rationale, and violated certain provisions of the Census Act. On the equal protection claim, however, the District Court concluded that respondents had not met their burden of showing that the Secretary was motivated by discriminatory animus. The court granted judgment to respondents on their statutory claims, vacated the Secretary's decision, and enjoined him from reinstating the citizenship question until he cured the legal errors the *2565 court had identified. 351 F.Supp.3d 502 (SDNY 2019).

 The Government appealed to the Second Circuit, but also filed a petition for writ of certiorari before judgment, asking this Court to review the District Court's decision directly because the case involved an issue of imperative public importance, and the census questionnaire needed to be finalized for printing by the end of June 2019. We granted the petition. 586 U. S. ----, 139 S.Ct. 16 , 202 L.Ed.2d 306 (2019). At the Government's request, we later ordered the parties to address whether the Enumeration Clause provided an alternative basis to affirm. 586 U. S. ----, 139 S.Ct. 16 , 202 L.Ed.2d 306 (2019).

 II

 We begin with jurisdiction. Article III of the Constitution limits federal courts to deciding "Cases" and "Controversies." For a legal dispute to qualify as a genuine case or controversy, at least one plaintiff must have standing to sue. The doctrine of standing "limits the category of litigants empowered to maintain a lawsuit in federal court to seek redress for a legal wrong" and "confines the federal courts to a properly judicial role." Spokeo , Inc. v. Robins , 578 U. S. ----, ----, 136 S.Ct. 1540 , 1547, 194 L.Ed.2d 635 (2016). To have standing, a plaintiff 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. Federal Election Comm'n , 554 U.S. 724 , 733, 128 S.Ct. 2759 , 171 L.Ed.2d 737 (2008).

 Respondents assert a number of injuries-diminishment of political representation, loss of federal funds, degradation of census data, and diversion of resources-all of which turn on their expectation that reinstating a citizenship question will depress the census response rate and lead to an inaccurate population count. Several States with a disproportionate share of noncitizens, for example, anticipate losing a seat in Congress or qualifying for less federal funding if their populations are undercounted. These are primarily future injuries, which "may suffice if the threatened injury is certainly impending, or there is a substantial risk that the harm will occur." Susan B. Anthony List v. Driehaus , 573 U.S. 149 , 158, 134 S.Ct. 2334 , 189 L.Ed.2d 246 (2014) (internal quotation marks omitted).

 The District Court concluded that the evidence at trial established a sufficient likelihood that the reinstatement of a citizenship question would result in noncitizen households responding to the census at lower rates than other groups, which in turn would cause them to be undercounted and lead to many of respondents' asserted injuries. For purposes of standing, these findings of fact were not so suspect as to be clearly erroneous.

 We therefore agree that at least some respondents have Article III standing. Several state respondents here have shown that if noncitizen households are undercounted by as little as 2%-lower than the District Court's 5.8% prediction-they will lose out on federal funds that are distributed on the basis of state population. That is a sufficiently concrete and imminent injury to satisfy Article III, and there is no dispute that a ruling in favor of respondents would redress that harm.

 The Government contends, however, that any harm to respondents is not fairly traceable to the Secretary's decision, because such harm depends on the independent action of third parties choosing to violate their legal duty to respond to the census. The chain of causation is made *2566 even more tenuous, the Government argues, by the fact that such intervening, unlawful third-party action would be motivated by unfounded fears that the Federal Government will itself break the law by using noncitizens' answers against them for law enforcement purposes. The Government invokes our steady refusal to "endorse standing theories that rest on speculation about the decisions of independent actors," Clapper v. Amnesty Int'l USA , 568 U.S. 398 , 414, 133 S.Ct. 1138 , 185 L.Ed.2d 264 (2013), particularly speculation about future unlawful conduct, Los Angeles v. Lyons , 461 U.S. 95 , 105, 103 S.Ct. 1660 , 75 L.Ed.2d 675 (1983).

 But we are satisfied that, in these circumstances, respondents have met their burden of showing that third parties will likely react in predictable ways to the citizenship question, even if they do so unlawfully and despite the requirement that the Government keep individual answers confidential. The evidence at trial established that noncitizen households have historically responded to the census at lower rates than other groups, and the District Court did not clearly err in crediting the Census Bureau's theory that the discrepancy is likely attributable at least in part to noncitizens' reluctance to answer a citizenship question. Respondents' theory of standing thus does not rest on mere speculation about the decisions of third parties; it relies instead on the predictable effect of Government action on the decisions of third parties. Cf. Bennett v. Spear , 520 U.S. 154 , 169-170, 117 S.Ct. 1154 , 137 L.Ed.2d 281 (1997) ; Davis , 554 U.S. at 734 -735 , 128 S.Ct. 2759 . Because Article III "requires no more than de facto causality," Block v. Meese , 793 F.2d 1303 , 1309 (CADC 1986) (Scalia, J.), traceability is satisfied here. We may therefore consider the merits of respondents' claims, at least as far as the Constitution is concerned.

 III

 The Enumeration Clause of the Constitution does not provide a basis to set aside the Secretary's decision. The text of that clause "vests Congress with virtually unlimited discretion in conducting the decennial 'actual Enumeration,' " and Congress "has delegated its broad authority over the census to the Secretary." Wisconsin , 517 U.S. at 19 , 116 S.Ct. 1091 . Given that expansive grant of authority, we have rejected challenges to the conduct of the census where the Secretary's decisions bore a "reasonable relationship to the accomplishment of an actual enumeration." Id. , at 20, 116 S.Ct. 1091 .

 Respondents ask us to evaluate the Secretary's decision to reinstate a citizenship question under that "reasonable relationship" standard, but we agree with the District Court that a different analysis is needed here. Our cases applying that standard concerned decisions about the population count itself-such as a postcensus decision not to use a particular method to adjust an undercount, id. , at 4, 116 S.Ct. 1091 , and a decision to allocate overseas military personnel to their home States, Franklin v. Massachusetts , 505 U.S. 788 , 790-791, 112 S.Ct. 2767 , 120 L.Ed.2d 636 (1992). We have never applied the standard to decisions about what kinds of demographic information to collect in the course of taking the census. Indeed, as the District Court recognized, applying the "reasonable relationship" standard to every census-related decision "would lead to the conclusion that it is unconstitutional to ask any demographic question on the census" because "asking such questions bears no relationship whatsoever to the goal of an accurate headcount." 315 F.Supp.3d at 804 -805 . Yet demographic questions have been asked in every census since 1790, and questions about citizenship in particular *2567 have been asked for nearly as long. Like the District Court, we decline respondents' invitation to measure the constitutionality of the citizenship question by a standard that would seem to render every census since 1790 unconstitutional.

 We look instead to Congress's broad authority over the census, as informed by long and consistent historical practice. All three branches of Government have understood the Constitution to allow Congress, and by extension the Secretary, to use the census for more than simply counting the population. Since 1790, Congress has sought, or permitted the Secretary to seek, information about matters as varied as age, sex, marital status, health, trade, profession, literacy, and value of real estate owned. See id. , at 801. Since 1820, it has sought, or permitted the Secretary to seek, information about citizenship in particular. Federal courts have approved the practice of collecting demographic data in the census. See, e.g. , United States v. Moriarity , 106 F. 886 , 891 (CC SDNY 1901) (duty to take a census of population "does not prohibit the gathering of other statistics, if 'necessary and proper,' for the intelligent exercise of other powers enumerated in the constitution"). While we have never faced the question directly, we have assumed that Congress has the power to use the census for information-gathering purposes, see Legal Tender Cases , 12 Wall. 457 , 536, 20 L.Ed. 287 (1871), and we have recognized the role of the census as a "linchpin of the federal statistical system by collecting data on the characteristics of individuals, households, and housing units throughout the country," Department of Commerce v. United States House of Representatives , 525 U.S. 316 , 341, 119 S.Ct. 765 , 142 L.Ed.2d 797 (1999) (internal quotation marks omitted).

 That history matters. Here, as in other areas, our interpretation of the Constitution is guided by a Government practice that "has been open, widespread, and unchallenged since the early days of the Republic." NLRB v. Noel Canning , 573 U.S. 513 , 572, 134 S.Ct. 2550 , 189 L.Ed.2d 538 (2014) (Scalia, J., concurring in judgment); see also Wisconsin , 517 U.S. at 21 , 116 S.Ct. 1091 (noting "importance of historical practice" in census context). In light of the early understanding of and long practice under the Enumeration Clause, we conclude that it permits Congress, and by extension the Secretary, to inquire about citizenship on the census questionnaire. We need not, and do not, decide the constitutionality of any other question that Congress or the Secretary might decide to include in the census.

 IV

 The District Court set aside the Secretary's decision to reinstate a citizenship question on the grounds that the Secretary acted arbitrarily and violated certain provisions of the Census Act. The Government contests those rulings, but also argues that the Secretary's decision was not judicially reviewable under the Administrative Procedure Act in the first place. We begin with that contention.

 A

 The Administrative Procedure Act embodies a "basic presumption of judicial review," Abbott Laboratories v. Gardner , 387 U.S. 136 , 140, 87 S.Ct. 1507 , 18 L.Ed.2d 681 (1967), and instructs reviewing courts to set aside agency action that is "arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law," 5 U. S. C. § 706 (2)(A). Review is not available, however, "to the extent that" a relevant statute precludes it, § 701(a)(1), or the agency action is "committed to agency discretion by law," § 701(a)(2). The Government argues that the Census Act *2568 commits to the Secretary's unreviewable discretion decisions about what questions to include on the decennial census questionnaire.

 We disagree. To be sure, the Act confers broad authority on the Secretary. Section 141(a) instructs him to take "a decennial census of population" in "such form and content as he may determine, including the use of sampling procedures and special surveys." 13 U. S. C. § 141 . The Act defines "census of population" to mean "a census of population, housing, and matters relating to population and housing," § 141(g), and it authorizes the Secretary, in "connection with any such census," to "obtain such other census information as necessary," § 141(a). It also states that the "Secretary shall prepare questionnaires, and shall determine the inquiries, and the number, form, and subdivisions thereof, for the statistics, surveys, and censuses provided for in this title." § 5. And it authorizes him to acquire materials, such as administrative records, from other federal, state, and local agencies in aid of conducting the census. § 6. Those provisions leave much to the Secretary's discretion. See Wisconsin , 517 U.S. at 19 , 116 S.Ct. 1091 ("Through the Census Act, Congress has delegated its broad authority over the census to the Secretary.").

 But they do not leave his discretion unbounded. In order to give effect to the command that courts set aside agency action that is an abuse of discretion, and to honor the presumption of judicial review, we have read the § 701(a)(2) exception for action committed to agency discretion "quite narrowly, restricting it to 'those rare circumstances where the relevant statute is drawn so that a court would have no meaningful standard against which to judge the agency's exercise of discretion.' " Weyerhaeuser Co. v. United States Fish and Wildlife Serv. , 586 U. S. ----, ----, 139 S.Ct. 361 , 370, 202 L.Ed.2d 269 (2018) (quoting Lincoln v. Vigil , 508 U.S. 182 , 191, 113 S.Ct. 2024 , 124 L.Ed.2d 101 (1993) ). And we have generally limited the exception to "certain categories of administrative decisions that courts traditionally have regarded as 'committed to agency discretion,' " id. , at 191, 113 S.Ct. 2024 , such as a decision not to institute enforcement proceedings, Heckler v. Chaney , 470 U.S. 821 , 831-832, 105 S.Ct. 1649 , 84 L.Ed.2d 714 (1985), or a decision by an intelligence agency to terminate an employee in the interest of national security, Webster v. Doe , 486 U.S. 592 , 600-601, 108 S.Ct. 2047 , 100 L.Ed.2d 632 (1988).

 The taking of the census is not one of those areas traditionally committed to agency discretion. We and other courts have entertained both constitutional and statutory challenges to census-related decisionmaking. See, e.g. , Department of Commerce , 525 U.S. 316 , 119 S.Ct. 765 , 142 L.Ed.2d 797 ; Wisconsin , 517 U.S. 1 , 116 S.Ct. 1091 , 134 L.Ed.2d 167 ; Carey v. Klutznick , 637 F.2d 834 (CA2 1980).

 Nor is the statute here drawn so that it furnishes no meaningful standard by which to judge the Secretary's action. In contrast to the National Security Act in Webster , which gave the Director of Central Intelligence discretion to terminate employees whenever he "deem[ed]" it "advisable," 486 U.S. at 594 , 108 S.Ct. 2047 , the Census Act constrains the Secretary's authority to determine the form and content of the census in a number of ways. Section 195, for example, governs the extent to which he can use statistical sampling. Section 6(c), which will be considered in more detail below, circumscribes his power in certain circumstances to collect information through direct inquiries when administrative records are available. More generally, by mandating a population count that will be used to apportion representatives, *2569 see § 141(b), 2 U. S. C. § 2a, the Act imposes "a duty to conduct a census that is accurate and that fairly accounts for the crucial representational rights that depend on the census and the apportionment." Franklin , 505 U.S. at 819 -820 , 112 S.Ct. 2767 (Stevens, J., concurring in part and concurring in judgment).

 The Secretary's decision to reinstate a citizenship question is amenable to review for compliance with those and other provisions of the Census Act, according to the general requirements of reasoned agency decisionmaking. Because this is not a case in which there is "no law to apply," Overton Park , 401 U.S. at 410 , 91 S.Ct. 814 , the Secretary's decision is subject to judicial review.

 B

 At the heart of this suit is respondents' claim that the Secretary abused his discretion in deciding to reinstate a citizenship question. We review the Secretary's exercise of discretion under the deferential "arbitrary and capricious" standard. See 5 U. S. C. § 706 (2)(A). Our scope of review is "narrow": we determine only whether the Secretary examined "the relevant data" and articulated "a satisfactory explanation" for his decision, "including a rational connection between the facts found and the choice made." Motor Vehicle Mfrs. Assn. of United States, Inc. v. State Farm Mut. Automobile Ins. Co. , 463 U.S. 29 , 43, 103 S.Ct. 2856 , 77 L.Ed.2d 443 (1983) (internal quotation marks omitted). We may not substitute our judgment for that of the Secretary, ibid. , but instead must confine ourselves to ensuring that he remained "within the bounds of reasoned decisionmaking," Baltimore Gas & Elec. Co. v. Natural Resources Defense Council, Inc. , 462 U.S. 87 , 105, 103 S.Ct. 2246 , 76 L.Ed.2d 437 (1983).

 The District Court set aside the Secretary's decision for two independent reasons: His course of action was not supported by the evidence before him, and his stated rationale was pretextual. We focus on the first point here and take up the question of pretext later.

 The Secretary examined the Bureau's analysis of various ways to collect improved citizenship data and explained why he thought the best course was to both reinstate a citizenship question and use citizenship data from administrative records to fill in the gaps. He considered but rejected the Bureau's recommendation to use administrative records alone. As he explained, records are lacking for about 10% of the population, so the Bureau would still need to estimate citizenship for millions of voting-age people. Asking a citizenship question of everyone, the Secretary reasoned, would eliminate the need to estimate citizenship for many of those people. And supplementing census responses with administrative record data would help complete the picture and allow the Bureau to better estimate citizenship for the smaller set of cases where it was still necessary to do so.

 The evidence before the Secretary supported that decision. As the Bureau acknowledged, each approach-using administrative records alone, or asking about citizenship and using records to fill in the gaps-entailed tradeoffs between accuracy and completeness. Without a citizenship question, the Bureau would need to estimate the citizenship of about 35 million people; with a citizenship question, it would need to estimate the citizenship of only 13.8 million. Under either approach, there would be some errors in both the administrative records and the Bureau's estimates. With a citizenship question, there would also be some erroneous self-responses (about 500,000) and some conflicts *2570 between responses and administrative record data (about 9.5 million).

 The Bureau explained that the "relative quality" of the citizenship data generated by each approach would depend on the "relative importance of the errors" in each, but it was not able to "quantify the relative magnitude of the errors across the alternatives." App. 148. The Bureau nonetheless recommended using administrative records alone because it had "high confidence" that it could develop an accurate model for estimating the citizenship of the 35 million people for whom administrative records were not available, and it thought the resulting citizenship data would be of superior quality. Id. , at 146, 158-159. But when the time came for the Secretary to make a decision, the model did not yet exist, and even if it had, there was no way to gauge its relative accuracy. As the Bureau put it, "we will most likely never possess a fully adequate truth deck to benchmark" the model-which appears to be bureaucratese for "maybe, maybe not." Id. , at 146. The Secretary opted instead for the approach that would yield a more complete set of data at an acceptable rate of accuracy, and would require estimating the citizenship of fewer people.

 The District Court overruled that choice, agreeing with the Bureau's assessment that its recommended approach would yield higher quality citizenship data on the whole. But the choice between reasonable policy alternatives in the face of uncertainty was the Secretary's to make. He considered the relevant factors, weighed risks and benefits, and articulated a satisfactory explanation for his decision. In overriding that reasonable exercise of discretion, the court improperly substituted its judgment for that of the agency.

 The Secretary then weighed the benefit of collecting more complete and accurate citizenship data against the risk that inquiring about citizenship would depress census response rates, particularly among noncitizen households. In the Secretary's view, that risk was difficult to assess. The Bureau predicted a 5.1% decline in response rates among noncitizen households if the citizenship question were reinstated. 2 It relied for that prediction primarily on studies showing that, while noncitizens had responded at lower rates than citizens to the 2000 short-form and 2010 censuses, which did not ask about citizenship, they responded at even lower rates than citizens to the 2000 long-form census and the 2010 American Community Survey, which did ask about citizenship. The Bureau thought it was reasonable to infer that the citizenship question accounted for the differential decline in noncitizen responses. But, the Secretary explained, the Bureau was unable to rule out other causes. For one thing, the evidence before the Secretary suggested that noncitizen households tend to be more distrustful of, and less likely to respond to, any government effort to collect information. For another, both the 2000 long-form census and 2010 ACS asked over 45 questions on a range of topics, including employment, income, and housing characteristics. Noncitizen households might disproportionately fail to respond to a lengthy and intrusive Government questionnaire for a number of reasons besides reluctance to answer a citizenship question-reasons relating to education level, socioeconomic status, and less exposure to Government outreach efforts. See App. to Pet. for Cert. 553a-554a, 557a-558a.

 *2571 The Secretary justifiably found the Bureau's analysis inconclusive. Weighing that uncertainty against the value of obtaining more complete and accurate citizenship data, he determined that reinstating a citizenship question was worth the risk of a potentially lower response rate. That decision was reasonable and reasonably explained, particularly in light of the long history of the citizenship question on the census.

 Justice BREYER would conclude otherwise, but only by subordinating the Secretary's policymaking discretion to the Bureau's technocratic expertise. Justice BREYER's analysis treats the Bureau's (pessimistic) prediction about response rates and (optimistic) assumptions about its data modeling abilities as touchstones of substantive reasonableness rather than simply evidence for the Secretary to consider. He suggests that the Secretary should have deferred to the Bureau or at least offered some special justification for drawing his own inferences and adopting his own assumptions. But the Census Act authorizes the Secretary, not the Bureau, to make policy choices within the range of reasonable options. And the evidence before the Secretary hardly led ineluctably to just one reasonable course of action. It called for value-laden decisionmaking and the weighing of incommensurables under conditions of uncertainty. The Secretary was required to consider the evidence and give reasons for his chosen course of action. He did so. It is not for us to ask whether his decision was "the best one possible" or even whether it was "better than the alternatives." FERC v. Electric Power Supply Assn. , 577 U. S. ----, ----, 136 S.Ct. 760 , 782, 193 L.Ed.2d 661 (2016). By second-guessing the Secretary's weighing of risks and benefits and penalizing him for departing from the Bureau's inferences and assumptions, Justice BREYER-like the District Court-substitutes his judgment for that of the agency.

 C

 The District Court also ruled that the Secretary violated two particular provisions of the Census Act, § 6(c) and § 141(f).

 Section 6 has three subsections. Subsections (a) and (b) authorize the Secretary to acquire administrative records from other federal agencies and from state and local governments. 3 Subsection (c) states:

 "To the maximum extent possible and consistent with the kind, timeliness, quality and scope of the statistics required, the Secretary shall acquire and use information available from any source referred to in subsection (a) or (b) of this section instead of conducting direct inquiries." 13 U. S. C. § 6 (c).
 The District Court held, and respondents argue, that the Secretary failed to comply with § 6(c) because he opted to collect citizenship data using direct inquiries when it was possible to provide DOJ with data from administrative records alone.

 *2572 At the outset, § 6(c) may not even apply here. It governs the Secretary's choices with respect to "statistics required." The parties have assumed that phrase refers to census-related data that the Secretary wishes to acquire, but it may instead refer to particular kinds of statistics that other provisions of the Census Act actually do require the Secretary to collect and publish. See, e.g. , § 41 ("The Secretary shall collect and publish statistics concerning [cotton and cotton production]."); § 61 ("The Secretary shall collect, collate, and publish monthly statistics concerning [vegetable and animal oils and the like]."); § 91 ("The Secretary shall collect and publish quarterly financial statistics of business operations, organization, practices, management, and relation to other businesses."). If so, § 6(c) would seem to have nothing to say about the Secretary's collection of census-related citizenship data, which is not a "statistic" he is "required" to collect.

 Regardless, assuming the provision applies, the Secretary complied with it, for essentially the same reasons that his decision was not arbitrary and capricious. As he explained, administrative records would not, in his judgment, provide the more complete and accurate data that DOJ sought. He thus could not, "consistent with" the kind and quality of the "statistics required," use administrative records instead of asking about citizenship directly. Respondents' arguments to the contrary rehash their disagreement with the Secretary's policy judgment about which approach would yield the most complete and accurate citizenship data. For the reasons already discussed, we may not substitute our judgment for that of the Secretary here.

 We turn now to § 141(f), which requires the Secretary to report to Congress about his plans for the census. Paragraph (1) instructs him to submit, at least three years before the census date, a report containing his "determination of the subjects proposed to be included, and the types of information to be compiled," in the census. Paragraph (2) then tells him to submit, at least two years before the census date, a report containing his "determination of the questions proposed to be included" in the census. Paragraph (3) provides:

 "[A]fter submission of a report under paragraph (1) or (2) of this subsection and before the appropriate census date, if the Secretary finds new circumstances exist which necessitate that the subjects, types of information, or questions contained in reports so submitted be modified, [he shall submit] a report containing the Secretary's determination of the subjects, types of information, or questions as proposed to be modified."
 The Secretary timely submitted his paragraph (1) report in March 2017. It did not mention citizenship. In December 2017, he received DOJ's formal request. Three months later, in March 2018, he timely submitted his paragraph (2) report. It did propose asking a question about citizenship.

 The District Court held that the Secretary's failure to mention citizenship in his March 2017 report violated § 141(f)(1) and provided an independent basis to set aside his action. Assuming without deciding that the Secretary's compliance with the reporting requirement is for courts-rather than Congress-to police, we disagree. The Secretary's March 2018 report satisfied the requirements of paragraph (3): By informing Congress that he proposed to include a citizenship question, the Secretary necessarily also informed Congress that he proposed to modify the original list of subjects that he submitted in the March 2017 report. Nothing *2573 in § 141(f) suggests that the same report cannot simultaneously fulfill the requirements of paragraphs (2) and (3). And to the extent paragraph (3) requires the Secretary to explain his finding of new circumstances, he did so in his March 2018 memo, which described DOJ's intervening request.

 In any event, even if we agreed with the District Court that the Secretary technically violated § 141(f) by submitting a paragraph (2) report that doubled as a paragraph (3) report, the error would surely be harmless in these circumstances, where the Secretary nonetheless fully informed Congress of, and explained, his decision. See 5 U. S. C. § 706 (in reviewing agency action, "due account shall be taken of the rule of prejudicial error").

 V

 We now consider the District Court's determination that the Secretary's decision must be set aside because it rested on a pretextual basis, which the Government conceded below would warrant a remand to the agency.

 We start with settled propositions. First, in order to permit meaningful judicial review, an agency must "disclose the basis" of its action. Burlington Truck Lines, Inc. v. United States , 371 U.S. 156 , 167-169, 83 S.Ct. 239 , 9 L.Ed.2d 207 (1962) (internal quotation marks omitted); see also SEC v. Chenery Corp. , 318 U.S. 80 , 94, 63 S.Ct. 454 , 87 L.Ed. 626 (1943) ("[T]he orderly functioning of the process of review requires that the grounds upon which the administrative agency acted be clearly disclosed and adequately sustained.").

 Second, in reviewing agency action, a court is ordinarily limited to evaluating the agency's contemporaneous explanation in light of the existing administrative record. Vermont Yankee Nuclear Power Corp. v. Natural Resources Defense Council, Inc. , 435 U.S. 519 , 549, 98 S.Ct. 1197 , 55 L.Ed.2d 460 (1978) ; Camp v. Pitts , 411 U.S. 138 , 142-143, 93 S.Ct. 1241 , 36 L.Ed.2d 106 (1973) ( per curiam ). That principle reflects the recognition that further judicial inquiry into "executive motivation" represents "a substantial intrusion" into the workings of another branch of Government and should normally be avoided. Arlington Heights v. Metropolitan Housing Development Corp. , 429 U.S. 252 , 268, n. 18, 97 S.Ct. 555 , 50 L.Ed.2d 450 (1977) ; see Overton Park , 401 U.S. at 420 , 91 S.Ct. 814 .

 Third, a court may not reject an agency's stated reasons for acting simply because the agency might also have had other unstated reasons. See Jagers v. Federal Crop Ins. Corp. , 758 F.3d 1179 , 1185-1186 (CA10 2014) (rejecting argument that "the agency's subjective desire to reach a particular result must necessarily invalidate the result, regardless of the objective evidence supporting the agency's conclusion"). Relatedly, a court may not set aside an agency's policymaking decision solely because it might have been influenced by political considerations or prompted by an Administration's priorities. Agency policymaking is not a "rarified technocratic process, unaffected by political considerations or the presence of Presidential power." Sierra Club v. Costle , 657 F.2d 298 , 408 (CADC 1981). Such decisions are routinely informed by unstated considerations of politics, the legislative process, public relations, interest group relations, foreign relations, and national security concerns (among others).

 Finally, we have recognized a narrow exception to the general rule against inquiring into "the mental processes of administrative decisionmakers." Overton Park , 401 U.S. at 420 , 91 S.Ct. 814 . On a *2574 "strong showing of bad faith or improper behavior," such an inquiry may be warranted and may justify extra-record discovery. Ibid.

 The District Court invoked that exception in ordering extra-record discovery here. Although that order was premature, we think it was ultimately justified in light of the expanded administrative record. Recall that shortly after this litigation began, the Secretary, prodded by DOJ, filed a supplemental memo that added new, pertinent information to the administrative record. The memo disclosed that the Secretary had been considering the citizenship question for some time and that Commerce had inquired whether DOJ would formally request reinstatement of the question. That supplemental memo prompted respondents to move for both completion of the administrative record and extra-record discovery. The District Court granted both requests at the same hearing, agreeing with respondents that the Government had submitted an incomplete administrative record and that the existing evidence supported a prima facie showing that the VRA rationale was pretextual.

 The Government did not challenge the court's conclusion that the administrative record was incomplete, and the parties stipulated to the inclusion of more than 12,000 pages of internal deliberative materials as part of the administrative record, materials that the court later held were sufficient on their own to demonstrate pretext. The Government did, however, challenge the District Court's order authorizing extra-record discovery, as well as the court's later orders compelling depositions of the Secretary and of the Acting Assistant Attorney General for DOJ's Civil Rights Division.

 We agree with the Government that the District Court should not have ordered extra-record discovery when it did. At that time, the most that was warranted was the order to complete the administrative record. But the new material that the parties stipulated should have been part of the administrative record-which showed, among other things, that the VRA played an insignificant role in the decisionmaking process-largely justified such extra-record discovery as occurred (which did not include the deposition of the Secretary himself). We accordingly review the District Court's ruling on pretext in light of all the evidence in the record before the court, including the extra-record discovery.

 That evidence showed that the Secretary was determined to reinstate a citizenship question from the time he entered office; instructed his staff to make it happen; waited while Commerce officials explored whether another agency would request census-based citizenship data; subsequently contacted the Attorney General himself to ask if DOJ would make the request; and adopted the Voting Rights Act rationale late in the process. In the District Court's view, this evidence established that the Secretary had made up his mind to reinstate a citizenship question "well before" receiving DOJ's request, and did so for reasons unknown but unrelated to the VRA. 351 F.Supp.3d at 660 .

 The Government, on the other hand, contends that there was nothing objectionable or even surprising in this. And we agree-to a point. It is hardly improper for an agency head to come into office with policy preferences and ideas, discuss them with affected parties, sound out other agencies for support, and work with staff attorneys to substantiate the legal basis for a preferred policy. The record here reflects the sometimes involved nature of Executive Branch decisionmaking, but no *2575 particular step in the process stands out as inappropriate or defective.

 And yet, viewing the evidence as a whole, we share the District Court's conviction that the decision to reinstate a citizenship question cannot be adequately explained in terms of DOJ's request for improved citizenship data to better enforce the VRA. Several points, considered together, reveal a significant mismatch between the decision the Secretary made and the rationale he provided.

 The record shows that the Secretary began taking steps to reinstate a citizenship question about a week into his tenure, but it contains no hint that he was considering VRA enforcement in connection with that project. The Secretary's Director of Policy did not know why the Secretary wished to reinstate the question, but saw it as his task to "find the best rationale." Id., at 551 . The Director initially attempted to elicit requests for citizenship data from the Department of Homeland Security and DOJ's Executive Office for Immigration Review, neither of which is responsible for enforcing the VRA. After those attempts failed, he asked Commerce staff to look into whether the Secretary could reinstate the question without receiving a request from another agency. The possibility that DOJ's Civil Rights Division might be willing to request citizenship data for VRA enforcement purposes was proposed by Commerce staff along the way and eventually pursued.

 Even so, it was not until the Secretary contacted the Attorney General directly that DOJ's Civil Rights Division expressed interest in acquiring census-based citizenship data to better enforce the VRA. And even then, the record suggests that DOJ's interest was directed more to helping the Commerce Department than to securing the data. The December 2017 letter from DOJ drew heavily on contributions from Commerce staff and advisors. Their influence may explain why the letter went beyond a simple entreaty for better citizenship data-what one might expect of a typical request from another agency-to a specific request that Commerce collect the data by means of reinstating a citizenship question on the census. Finally, after sending the letter, DOJ declined the Census Bureau's offer to discuss alternative ways to meet DOJ's stated need for improved citizenship data, further suggesting a lack of interest on DOJ's part.

 Altogether, the evidence tells a story that does not match the explanation the Secretary gave for his decision. In the Secretary's telling, Commerce was simply acting on a routine data request from another agency. Yet the materials before us indicate that Commerce went to great lengths to elicit the request from DOJ (or any other willing agency). And unlike a typical case in which an agency may have both stated and unstated reasons for a decision, here the VRA enforcement rationale-the sole stated reason-seems to have been contrived.

 We are presented, in other words, with an explanation for agency action that is incongruent with what the record reveals about the agency's priorities and decisionmaking process. It is rare to review a record as extensive as the one before us when evaluating informal agency action-and it should be. But having done so for the sufficient reasons we have explained, we cannot ignore the disconnect between the decision made and the explanation given. Our review is deferential, but we are "not required to exhibit a naiveté from which ordinary citizens are free." United States v. Stanchich , 550 F.2d 1294 , 1300 (CA2 1977) (Friendly, J.). The reasoned explanation requirement of administrative law, after all, is meant to ensure that agencies offer genuine justifications for important *2576 decisions, reasons that can be scrutinized by courts and the interested public. Accepting contrived reasons would defeat the purpose of the enterprise. If judicial review is to be more than an empty ritual, it must demand something better than the explanation offered for the action taken in this case.

 In these unusual circumstances, the District Court was warranted in remanding to the agency, and we affirm that disposition. See Florida Power & Light Co. v. Lorion , 470 U.S. 729 , 744, 105 S.Ct. 1598 , 84 L.Ed.2d 643 (1985). We do not hold that the agency decision here was substantively invalid. But agencies must pursue their goals reasonably. Reasoned decisionmaking under the Administrative Procedure Act calls for an explanation for agency action. What was provided here was more of a distraction.

 * * *

 The judgment of the United States District Court for the Southern District of New York is affirmed in part and reversed in part, and the case is remanded for further proceedings consistent with this opinion.

 It is so ordered.

 Justice THOMAS, with whom Justice GORSUCH and Justice KAVANAUGH join, concurring in part and dissenting in part.

 In March 2018, the Secretary of Commerce exercised his broad discretion over the administration of the decennial census to resume a nearly unbroken practice of asking a question relating to citizenship. Our only role in this case is to decide whether the Secretary complied with the law and gave a reasoned explanation for his decision. The Court correctly answers these questions in the affirmative. Ante , at 2566 - 2573. That ought to end our inquiry.

 The Court, however, goes further. For the first time ever, the Court invalidates an agency action solely because it questions the sincerity of the agency's otherwise adequate rationale. Echoing the din of suspicion and distrust that seems to typify modern discourse, the Court declares the Secretary's memorandum "pretextual" because, "viewing the evidence as a whole," his explanation that including a citizenship question on the census would help enforce the Voting Rights Act (VRA) "seems to have been contrived." Ante , at 2572 - 2573, 2574 - 2575, 2575 - 2576. The Court does not hold that the Secretary merely had additional , unstated reasons for reinstating the citizenship question. Rather, it holds that the Secretary's stated rationale did not factor at all into his decision.

 The Court's holding reflects an unprecedented departure from our deferential review of discretionary agency decisions. And, if taken seriously as a rule of decision, this holding would transform administrative law. It is not difficult for political opponents of executive actions to generate controversy with accusations of pretext, deceit, and illicit motives. Significant policy decisions are regularly criticized as products of partisan influence, interest-group pressure, corruption, and animus. Crediting these accusations on evidence as thin as the evidence here could lead judicial review of administrative proceedings to devolve into an endless morass of discovery and policy disputes not contemplated by the Administrative Procedure Act (APA).

 Unable to identify any legal problem with the Secretary's reasoning, the Court imputes one by concluding that he must not be telling the truth. The Court therefore upholds the decision of the District Court-which, in turn, was transparently based on the application of an administration-specific standard. App. to Pet. for Cert. 527a (crediting respondents' allegations *2577 that "the current Department of Justice has shown little interest in enforcing the" VRA (emphasis added)).

 The law requires a more impartial approach. Even assuming we are authorized to engage in the review undertaken by the Court-which is far from clear-we have often stated that courts reviewing agency action owe the Executive a "presumption of regularity." Citizens to Preserve Overton Park, Inc. v. Volpe , 401 U.S. 402 , 415, 91 S.Ct. 814 , 28 L.Ed.2d 136 (1971). The Court pays only lipservice to this principle. But, the evidence falls far short of supporting its decision. The Court, I fear, will come to regret inventing the principles it uses to achieve today's result. I respectfully dissent from Part V of the opinion of the Court. 1

 I

 As the Court explains, federal law directs the Secretary of Commerce to "take a decennial census." 13 U. S. C. § 141 (a) ; see U. S. Const., Art. I, § 2, cl. 3 ; Amdt. XIV, § 2; ante , at 2561 - 2562. The discretion afforded the Secretary is extremely broad. Subject only to constitutional limitations and a handful of inapposite statutory requirements, the Secretary is expressly authorized to "determine the inquiries" on the census questionnaire and to conduct the census "in such form and content as he may determine." §§ 5, 141(a) ; see ante , at 2567 - 2569, 2571 - 2573. 2 Prior census questionnaires have included questions ranging from sex, age, and race to commute, education, and radio ownership. And between 1820 and 2010, every decennial census questionnaire but one asked some segment of the population a question related to citizenship. The 2010 census was the first since 1840 that did not include any such question.

 In March 2018, the Secretary issued a memorandum reinstating a citizenship question on the 2020 census. He explained that the Department of Justice (DOJ) had formally requested reinstatement of the question because the data obtained would help enforce § 2 of the VRA. He further explained that the question had been well tested in light of its extensive previous use, that he had consulted with the Census Bureau on the proposal, and that his final decision incorporated feedback from the Bureau. He recognized that staff at the Bureau believed that better data could be obtained through modeling and reliance on existing records, but he disagreed with that assessment, explaining that the data was inconclusive and that he thought it preferable to ask the question directly of the entire population. Respondents brought suit, seeking judicial review of the Secretary's decision under the APA, 5 U. S. C. § 706 .

 II

 As relevant here, the APA requires courts to "hold unlawful and set aside" agency action that is "arbitrary, capricious, an abuse of discretion, or otherwise not in *2578 accordance with law." § 706(2)(A). We have emphasized that "[r]eview under the arbitrary and capricious standard is deferential." National Assn. of Home Builders v. Defenders of Wildlife , 551 U.S. 644 , 658, 127 S.Ct. 2518 , 168 L.Ed.2d 467 (2007) ; see Glickman v. Wileman Brothers & Elliott, Inc. , 521 U.S. 457 , 466, n. 8, 117 S.Ct. 2130 , 138 L.Ed.2d 585 (1997). It requires the reviewing court to determine whether the agency " 'examine[d] the relevant data and articulate[d] a satisfactory explanation for its action.' " FCC v. Fox Television Stations, Inc. , 556 U.S. 502 , 513, 129 S.Ct. 1800 , 173 L.Ed.2d 738 (2009). We have described this as a " 'narrow' standard of review" under which the reviewing court cannot " 'substitute its judgment for that of the agency,' and should 'uphold a decision of less than ideal clarity if the agency's path may reasonably be discerned.' " Id. , at 513-514, 129 S.Ct. 1800 (citation omitted); accord, Motor Vehicle Mfrs. Assn. of United States, Inc. v. State Farm Mut. Automobile Ins. Co. , 463 U.S. 29 , 43, 103 S.Ct. 2856 , 77 L.Ed.2d 443 (1983). 3

 Part IV-B of the opinion of the Court correctly applies this standard to conclude that the Secretary's decision survives ordinary arbitrary-and-capricious review. That holding should end our inquiry.

 But the opinion continues. Acknowledging that "no particular step" in the proceedings here "stands out as inappropriate or defective," even after reviewing "all the evidence in the record ..., including the extra-record discovery," ante , at 2574, the Court nevertheless agrees with the District Court that the Secretary's rationale for reinstating the citizenship question was "pretextual-that is, that the real reason for his decision was something other than the sole reason he put forward in his memorandum, namely enhancement of DOJ's VRA enforcement efforts." 351 F.Supp.3d 502 , 660 (SDNY 2019) ; see ante , at 2575 - 2576. According to the Court, something just "seems" wrong. Ibid.

 This conclusion is extraordinary. The Court engages in an unauthorized inquiry into evidence not properly before us to reach an unsupported conclusion. Moreover, each step of the inquiry offends the presumption of regularity we owe the Executive. The judgment of the District Court should be reversed.

 A

 Section 706(2) of the APA contemplates review of the administrative "record" to determine whether an agency's "action, findings, and conclusions" satisfy six specified standards. See §§ 706(2)(A)-(F). None instructs the Court to inquire into pretext. Consistent with this statutory text, we have held that a court is "ordinarily limited to evaluating the agency's contemporaneous explanation in light of the existing administrative record." Ante , at 2573 (citing Vermont Yankee Nuclear Power Corp. v. Natural Resources Defense Council, Inc. , 435 U.S. 519 , 549, 98 S.Ct. 1197 , 55 L.Ed.2d 460 (1978) ); see SEC v. Chenery Corp. , 318 U.S. 80 , 87, 63 S.Ct. 454 , 87 L.Ed. 626 (1943) ("The grounds upon which an administrative order must be judged are those upon which the record discloses that its action was based"). If an agency's stated findings and conclusions withstand scrutiny, the APA does not permit a court to set aside the decision solely because the agency had "other unstated *2579 reasons" for its decision, such as "political considerations" or the "Administration's priorities." Ante , at 2573 - 2574.

 Unsurprisingly, then, this Court has never held an agency decision arbitrary and capricious on the ground that its supporting rationale was "pretextual." Nor has it previously suggested that this was even a possibility. Under "settled propositions" of administrative law, ante , at 2572 - 2573, pretext is virtually never an appropriate or relevant inquiry for a reviewing court to undertake.

 Respondents conceptualize pretext as a subset of "arbitrary and capricious" review. It is far from clear that they are correct. But even if they were, an agency action is not arbitrary or capricious merely because the decisionmaker has other, unstated reasons for the decision. Ante , at 2573 - 2574. Nor is an agency action arbitrary and capricious merely because the decisionmaker was "inclined" to accomplish it before confirming that the law and facts supported that inclination. In re Dept. of Commerce , 586 U. S. ----, ----, 139 S.Ct. 16 , 17, 202 L.Ed.2d 306 (2018) (GORSUCH, J., concurring in part and dissenting in part).

 Accordingly, even under respondents' approach, a showing of pretext could render an agency action arbitrary and capricious only in the infinitesimally small number of cases in which the administrative record establishes that an agency's stated rationale did not factor at all into the decision, thereby depriving the action of an adequate supporting rationale. 4 This showing is extremely difficult to make because the administrative record will rarely, if ever, contain evidence sufficient to show that an agency's stated rationale did not actually factor into its decision. And we have stated that a "strong showing of bad faith or improper behavior" is necessary to venture beyond the agency's "administrative findings" and inquire into "the mental processes of administrative decisionmakers." Overton Park , 401 U.S. at 420 , 91 S.Ct. 814 . 5 We have never before found Overton Park 's exception satisfied, much less invalidated an agency action based on "pretext."

 Undergirding our arbitrary-and-capricious analysis is our longstanding precedent affording the Executive a "presumption of regularity." Id., at 415 , 91 S.Ct. 814 ; see United States v. Chemical Foundation, Inc. , 272 U.S. 1 , 14-15, 47 S.Ct. 1 , 71 L.Ed. 131 (1926). This presumption reflects respect for a coordinate branch of government whose officers not only take *2580 an oath to support the Constitution, as we do, Art. VI, but also are charged with "faithfully execut[ing]" our laws, Art. II, § 3. See United States v. Morgan , 313 U.S. 409 , 422, 61 S.Ct. 999 , 85 L.Ed. 1429 (1941) (presumption of regularity ensures that the "integrity of the administrative process" is appropriately respected). In practice, then, we give the benefit of the doubt to the agency.

 B

 The Court errs at the outset by proceeding beyond the administrative record to evaluate pretext. Respondents have not made a "strong showing of bad faith or improper behavior." Overton Park , supra , at 420, 91 S.Ct. 814 .

 The District Court's initial order granting extra-record discovery relied on four categories of evidence:

 "evidence that [the Secretary] was predisposed to reinstate the citizenship question when he took office; that the [DOJ] hadn't expressed a desire for more detailed citizenship data until the Secretary solicited its view; that he overruled the objections of his agency's career staff; and that he declined to order more testing of the question given its long history." Dept. of Commerce , 586 U. S., at ----, 139 S.Ct., at 18.
 None of this comes close to showing bad faith or improper behavior. Indeed, there is nothing even "unusual about a new cabinet secretary coming to office inclined to favor a different policy direction, soliciting support from other agencies to bolster his views, disagreeing with staff, or cutting through red tape." Ibid . Today all Members of the Court who reach the question agree that the District Court abused its discretion in ordering extra-record discovery based on this evidence. Ante , at 2574 ("We agree with the Government that the District Court should not have ordered extra-record discovery when it did").

 Nevertheless, the Court excuses the error because, in its view, "the new material that the parties [later] stipulated should have been part of the administrative record ... largely justified such extra-record discovery as occurred." Ibid. Given the requirement that respondents make a "strong showing" of bad faith, one would expect the Court to identify which "new material" supported such a showing. It does not. Nor does the Court square its suggestion that some of the extra-record discovery was not "justified" with its consideration of "all ... the extra-record discovery." Ante , at 2574 - 2575 . Regardless, I assume that the Court has in mind the administrative-record materials that the District Court would later rely on to establish pretext:

 "evidence that [the Secretary] had made the decision to add the citizenship question well before DOJ requested its addition in December 2017; the absence of any mention, at all , of VRA enforcement in the discussions of adding the question that preceded the [DOJ] Letter; unsuccessful attempts by Commerce Department staff to shop around for a request by another agency regarding citizenship data; and [the Secretary's] personal outreach to Attorney General Sessions, followed by the [DOJ] Letter; not to mention the conspicuous procedural irregularities that accompanied the decision to add the question." 351 F.Supp.3d at 661 (citations omitted).
 This evidence fails to make a strong showing of bad faith or improper behavior. Taken together, it proves at most that the Secretary was predisposed to add a citizenship question to the census and took steps to achieve that end before settling on the VRA rationale he included in his memorandum. Perhaps he had reasons for adding *2581 the citizenship question other than the VRA, but by the Court's own telling, that does not amount to evidence of bad faith or improper behavior. Ante , at 2573 - 2574; see Dept. of Commerce , supra , at 17.

 The Court thus errs in relying on materials outside the record to support its holding. And the Court does not claim that the evidence in the administrative record alone would prove that the March 2018 memorandum was a pretext. Given the presumption of regularity, the evidence discussed above falls far short of establishing that the VRA rationale did not factor at all into the Secretary's decision.

 C

 Even if it were appropriate for the Court to rely on evidence outside the administrative record, that evidence still fails to establish pretext. None of the evidence cited by the Court or the District Court comes close to showing that the Secretary's stated rationale-that adding a citizenship question to the 2020 census questionnaire would "provide ... data that are not currently available" and "permit more effective enforcement of the [VRA]," App. to Pet. for Cert. 548a-did not factor at all into his decision.

 Once again, the evidence cited by the Court suggests at most that the Secretary had "other unstated reasons" for reinstating the citizenship question. Ante , at 2573 - 2574. For example, the Court states that the Secretary's Director of Policy "initially attempted to elicit requests for citizenship data from the Department of Homeland Security and DOJ's Executive Office for Immigration Review." Ante , at 2575. But this hardly shows pretext. It simply suggests that the Director believed that citizenship information could be useful in tackling problems related to national security and illegal immigration-a view that would also explain why the Secretary might not have been "considering VRA enforcement" early on. Ibid. ; see also American Community Survey, Why We Ask: Place of Birth, Citizenship and Year of Entry (2016) (explaining that inquiries about "place of birth, citizenship, and year of entry" provide statistics that are "essential for agencies and policy makers setting and evaluating immigration policies and laws, understanding how different immigrant groups are assimilated, and monitoring against discrimination"), https://www2.census.gov/programs-surveys/acs/about/qbyqfact/2016/Citizenship.pdf (as last visited June 25, 2019).

 The Court emphasizes that the VRA rationale for the citizenship question originated in the Department of Commerce, and suggests that DOJ officials unthinkingly fell in line after the Attorney General was looped into the process. See ante , at 2575. But the Court ignores that the letter was drafted by the then-Acting Assistant Attorney General for Civil Rights and reviewed by five other DOJ attorneys, including the Chief of the DOJ's Voting Section. 351 F.Supp.3d at 554 -556 . Given the DOJ's multilayer review process and its explanation for requesting citizenship data, the Court's suggestion that the DOJ's letter was inadequately vetted or improperly "influence[d]" by the Department of Commerce is entirely unsupported. Ante , at 2575. In any event, none of this suggests, much less proves, that the Secretary harbored an unstated belief that adding the citizenship question would not help enforce the VRA, or that the VRA rationale otherwise did not factor at all into his decision. It simply suggests that a number of executive officials agreed that adding a citizenship question would support VRA enforcement.

 The Court's other evidence is even further afield. The Court thinks it telling that the DOJ's letter included "a specific request *2582 that Commerce collect the [citizenship] data by means of reinstating a citizenship question on the census," rather than a more open-ended "entreaty for better citizenship data." Ibid. I do not understand how the specificity of the DOJ's letter bears on whether the Secretary's rationale was pretextual-particularly since the letter specifically explained why "census questionnaire data regarding citizenship, if available, would be more appropriate for use in redistricting and in [VRA] litigation" than existing data. App. to Pet. for Cert. 568a; see id. , at 567a-568a. Unless the Court is now suggesting that agency correspondence must comply with the Court's subjective, unsupported view of what "might" constitute a "typical request from another agency," ante , at 2575, the specificity of the DOJ's letter is irrelevant. The Court also points to the DOJ's decision not to meet with the Census Bureau "to discuss alternative ways to meet DOJ's stated need for improved citizenship data." Ibid. But the Court does not explain how the DOJ's refusal bears on the Secretary's rationale. Besides, it is easy to understand why DOJ officials would not be interested in meeting with the Census Bureau. The meeting would have been with career employees whose acknowledged purpose was to talk the DOJ out of its request. See 351 F.Supp.3d at 557 . Having already considered the issue and explained the rationale behind the request, it seems at least plausible that the DOJ officials believed such a meeting would be unproductive.

 In short, the evidence cited by the Court establishes, at most, that leadership at both the Department of Commerce and the DOJ believed it important-for a variety of reasons-to include a citizenship question on the census.

 The Court also fails to give credit where it is due. The Secretary initiated this process inclined to favor what he called "Option B"-that is, simply "add[ing] a citizenship question to the decennial census." App. to Pet. for Cert. 552a. But the Census Bureau favored "Option C"-relying solely on "administrative records" to supply the information needed by the DOJ. Id. , at 554a. The Secretary considered this view and found it a "potentially appealing solution," ibid. , but concluded that it had shortcomings. Rather than revert to his original inclination, however, he "asked the Census Bureau to develop a fourth alternative, Option D, which would combine Options B and C." Id. , at 555a. And he settled on that solution. Whatever one thinks of the Secretary's choice, his willingness to change his mind in light of the Bureau's feedback belies the idea that his rationale or decisionmaking process was a pretext.

 The District Court's lengthy opinion pointed to other facts that, in its view, supported a finding of pretext. 351 F.Supp.3d at 567-572, 660-664 (discussing the statements, e-mails, acts, and omissions of numerous people involved in the process). I do not deny that a judge predisposed to distrust the Secretary or the administration could arrange those facts on a corkboard and-with a jar of pins and a spool of string-create an eye-catching conspiracy web. Cf. id. , at 662 (inferring "from the various ways in which [the Secretary] and his aides acted like people with something to hide that they did have something to hide"). But the Court does not rely on this evidence, and rightly so: It casts no doubt on whether the Secretary's stated rationale factored into his decision. The evidence suggests, at most, that the Secretary had multiple reasons for wanting to include the citizenship question on the census.

 Finally, if there could be any doubt about this conclusion, the presumption of *2583 regularity resolves it. Where there are equally plausible views of the evidence, one of which involves attributing bad faith to an officer of a coordinate branch of Government, the presumption compels giving the benefit of the doubt to that officer.

 III

 The Court's erroneous decision in this case is bad enough, as it unjustifiably interferes with the 2020 census. But the implications of today's decision are broader. With today's decision, the Court has opened a Pandora's box of pretext-based challenges in administrative law.

 Today's decision marks the first time the Court has ever invalidated an agency action as "pretextual." Having taken that step, one thing is certain: This will not be the last time it is asked to do so. Virtually every significant agency action is vulnerable to the kinds of allegations the Court credits today. These decisions regularly involve coordination with numerous stakeholders and agencies, involvement at the highest levels of the Executive Branch, opposition from reluctant agency staff, and-perhaps most importantly-persons who stand to gain from the action's demise. Opponents of future executive actions can be expected to make full use of the Court's new approach.

 The 2015 "Open Internet Order" provides a case in point. In 2015, the Federal Communications Commission (FCC) adopted a controversial order reclassifying broadband Internet access service as a "telecommunications service" subject to regulation under Title II of the Communications Act. See In re Protecting and Promoting the Open Internet , 30 FCC Rcd. 5601 , 5618 (2015). According to a dissenting Commissioner, the FCC "flip-flopp[ed]" on its previous policy not because of a change in facts or legal understanding, but based on "one reason and one reason alone. President Obama told us to do so." Id., at 5921 (statement of Comm'r Pai). His view was supported by a 2016 congressional Report in which Republican Senate staff concluded that "the FCC bent to the political pressure of the White House" and "failed to live up to standards of transparency." Majority Staff Report, Senate Committee on Homeland Security and Governmental Affairs, Regulating the Internet: How the White House Bowled Over FCC Independence, 114th Cong., 1st Sess., 29 (Comm. Print 2016). The Report cited evidence strikingly similar to that relied upon by the Court here-including agency-initiated "meetings with certain outside groups to support" the new result, id. , at 3; "apparen[t] ... concern from the career staff that there was insufficient notice to the public and affected stakeholders," id., at 4; and "regula[r] communicatio[n]" between the FCC Chairman and "presidential advisors," id. , at 25.

 Under the malleable standard applied by the Court today, a serious case could be made that the Open Internet Order should have been invalidated as "pretextual," regardless of whether any "particular step in the process stands out as inappropriate or defective." Ante , at 2575. It is enough, according to the Court, that a judge believes that the ultimate rationale "seems to have been contrived" when the evidence is considered "as a whole." Ante , at 2574, 2575 - 2576.

 Now that the Court has opened up this avenue of attack, opponents of executive actions have strong incentives to craft narratives that would derail them. Moreover, even if the effort to invalidate the action is ultimately unsuccessful, the Court's decision enables partisans to use the courts to harangue executive officers through depositions, discovery, delay, and distraction. The Court's decision could even implicate separation-of-powers concerns insofar as it *2584 enables judicial interference with the enforcement of the laws.

 In short, today's decision is a departure from traditional principles of administrative law. Hopefully it comes to be understood as an aberration-a ticket good for this day and this train only.

 * * *

 Because the Secretary's decision to reinstate a citizenship question on the 2020 census was legally sound and a reasoned exercise of his broad discretion, I respectfully dissent from Part V of the opinion of the Court.

 Justice BREYER, with whom Justice GINSBURG, Justice SOTOMAYOR, and Justice KAGAN join, concurring in part and dissenting in part.

 I join Parts I, II, IV-A, and V of the Court's opinion (except as otherwise indicated in this opinion). I dissent, however, from the conclusion the Court reaches in Part IV-B. To be more specific, I agree with the Court that the Secretary of Commerce provided a pretextual reason for placing a question about citizenship on the short-form census questionnaire and that a remand to the agency is appropriate on that ground. But I write separately because I also believe that the Secretary's decision to add the citizenship question was arbitrary and capricious and therefore violated the Administrative Procedure Act (APA).

 There is no serious dispute that adding a citizenship question would diminish the accuracy of the enumeration of the population-the sole constitutional function of the census and a task of great practical importance. The record demonstrates that the question would likely cause a disproportionate number of noncitizens and Hispanics to go uncounted in the upcoming census. That, in turn, would create a risk that some States would wrongfully lose a congressional representative and funding for a host of federal programs. And, the Secretary was told, the adverse consequences would fall most heavily on minority communities. The Secretary decided to ask the question anyway, citing a need for more accurate citizenship data. But the evidence indicated that asking the question would produce citizenship data that is less accurate, not more. And the reason the Secretary gave for needing better citizenship data in the first place-to help enforce the Voting Rights Act of 1965-was not convincing.

 In short, the Secretary's decision to add a citizenship question created a severe risk of harmful consequences, yet he did not adequately consider whether the question was necessary or whether it was an appropriate means of achieving his stated goal. The Secretary thus failed to "articulate a satisfactory explanation" for his decision, "failed to consider ... important aspect[s] of the problem," and "offered an explanation for [his] decision that runs counter to the evidence," all in violation of the APA. Motor Vehicle Mfrs. Assn. of United States, Inc. v. State Farm Mut. Automobile Ins. Co. , 463 U.S. 29 , 43, 103 S.Ct. 2856 , 77 L.Ed.2d 443 (1983). These failures, in my view, risked undermining public confidence in the integrity of our democratic system itself. I would therefore hold that the Secretary's decision-whether pretextual or not-was arbitrary, capricious, and an abuse of discretion.

 I

 A

 Three sets of laws determine the legal outcome of this case. First, the Constitution requires an "actual Enumeration" of the "whole number of persons in each State" every 10 years. Art. I, § 2, cl. 3 ; Amdt. 14, § 2. It does so in order to *2585 "provide a basis for apportioning representatives among the states in the Congress." Baldrige v. Shapiro , 455 U.S. 345 , 353, 102 S.Ct. 1103 , 71 L.Ed.2d 199 (1982) ; see also Art. I, § 2, cl. 3. The inclusion of this provision in the Constitution itself underscores the importance of conducting an accurate census. See Utah v. Evans , 536 U.S. 452 , 478, 122 S.Ct. 2191 , 153 L.Ed.2d 453 (2002) (recognizing "a strong constitutional interest in [the] accuracy" of the enumeration).

 Second, the Census Act contains two directives that constrain the Secretary's ability to add questions to the census. Section 195 says that the Secretary "shall, if he considers it feasible," authorize the use of statistical "sampling" in collecting demographic information. That means the Secretary must, if feasible, obtain demographic information through a survey sent to a sample of households, rather than through the short-form census questionnaire to which every household must respond. The other relevant provision, § 6(c), says that "[ t ] o the maximum extent possible and consistent with the kind, timeliness, quality and scope of the statistics required, the Secretary shall acquire and use information available" from administrative sources "instead of conducting direct inquiries." (Emphasis added.) These provisions, taken together, reflect a congressional preference for keeping the short form short, so that it does not burden recipients and thereby discourage them from responding.

 Third, the APA prohibits administrative agencies from making choices that are "arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law." 5 U. S. C. § 706 (2)(A). We have said that courts, in applying this provision, must decide "whether the decision was based on a consideration of the relevant factors and whether there has been a clear error of judgment." Citizens to Preserve Overton Park, Inc. v. Volpe , 401 U.S. 402 , 416, 91 S.Ct. 814 , 28 L.Ed.2d 136 (1971). The agency must have "examine[d] the relevant data and articulate[d] a satisfactory explanation for its action[,] including a 'rational connection between the facts found and the choice made.' " State Farm , 463 U.S. at 43 , 103 S.Ct. 2856 . An agency ordinarily fails to meet this standard if it has "failed to consider an important aspect of the problem, offered an explanation for its decision that runs counter to the evidence before the agency, or is so implausible that it could not be ascribed to a difference in view or the product of agency expertise." Ibid.

 Courts do not apply these principles of administrative law mechanically. Rather, they take into account, for example, the nature and importance of the particular decision, the relevance and importance of missing information, and the inadequacies of a particular explanation in light of their importance. The Federal Government makes tens of thousands, perhaps millions, of administrative decisions each year. And courts would be wrong to expect or insist upon administrative perfection. But here, the Enumeration Clause, the Census Act, and the nature of the risks created by the agency's decision all make clear that the decision before us is highly important to the proper functioning of our democratic system. It is therefore particularly important that courts here not overlook an agency's (1) failure to consider serious risks of harm, (2) failure to explain its refusal to minimize those risks, or (3) failure to link its conclusion to available evidence. My view, like that of the District Court, is that the agency here failed on all three counts.

 B

 A brief history of how the census has worked over the years will help the reader understand some of the shortcomings of *2586 the Secretary's decisionmaking process. The Framers wrote into the Constitution a mandate to conduct an "actual Enumeration" of the population every 10 years. Art. I, § 2, cl. 3. They did so for good reason. The purpose of the census is to "provide a basis for apportioning representatives among the states in the Congress," Baldrige , 455 U.S. at 353 , 102 S.Ct. 1103 , ensuring that "comparative state political power in the House ... reflect[s] comparative population," Evans , 536 U.S. at 477 , 122 S.Ct. 2191 . The Framers required an actual count of every resident to "limit political chicanery" and to prevent the census count from being "skewed for political ... purposes." Id. , at 500, 122 S.Ct. 2191 (THOMAS, J., concurring in part and dissenting in part).

 Throughout most of the Nation's history, the Federal Government used enumerators, often trained census takers, to conduct the census by going door to door. The enumerators would ask a host of questions, including place of birth, citizenship, and others. But after the 1950 census, the Bureau began to change its approach. Post-census studies revealed that the census had failed to count more than 5 million people and that the undercount disproportionately affected members of minority groups. See M. Anderson, The American Census: A Social History 201-202 (1988); Brief for Historians and Social Scientists as Amici Curiae 15. Studies showed that statistical sampling would produce higher quality data. Anderson, American Census, at 201.

 Beginning with the 1960 census, the Bureau consequently divided its questioning into a short form and a long form. The short form contained a list of questions-a short list-that the census would ask of every household. That list included basic demographic questions like sex, age, race, and marital status. The short form did not include, and has never included, a question about citizenship. See ibid. ; Dept. of Commerce, U. S. Census Bureau, Measuring America: The Decennial Censuses From 1790 to 2000, p. 128 (2002). By way of contrast, the long form set forth a host of questions that would be asked of only a sample of households. In 1960, the long form was sent to one in every four households; in subsequent years, it was sent to approximately one in every six. See 351 F.Supp.3d 502 , 520 (SDNY 2019). And it was more recently replaced by the American Community Survey (ACS), which is sent to approximately 1 in 38 households each year. The long form (and now the ACS) has often included a question about citizenship.

 In 1970, the Census Bureau made another important change to the census. It significantly reduced its reliance upon in-person enumerators. See Anderson, supra , at 206. Instead, it sent nearly all households a questionnaire by mail. Most households received the short form, and a small sample received the long form. Instructions on the form told each household to fill out the questionnaire and return it to the Census Bureau by mail. Enumerators would follow up with households that did not return the questionnaire.

 To maximize accuracy and minimize cost, the Bureau tried to bring about the highest possible "self-response" rate, i.e. , to encourage as many households as possible to respond by mail. For that reason, it tried to keep the short form as short as possible. And it consistently opposed placing a citizenship question on that form. It feared that adding a question about citizenship would "inevitably jeopardize the overall accuracy of the population count," partly because of added response burden but also because, as it explained, noncitizens faced with a citizenship question would be less likely to respond due to *2587 fears of "the information being used against them." Federation for Am. Immigration Reform v. Klutznick , 486 F.Supp. 564 , 568 (DDC 1980).

 Likely for similar reasons, Congress amended the Census Act in 1976, enacting the two statutory provisions to which I previously referred. These two provisions, 13 U. S. C. § 6 (c) and § 195, together encourage the Secretary not to ask demographic questions on the short form if the information can be obtained either through the long form or through administrative records.

 II

 With this statutory and historical background, we can more easily consider the agency decision directly under review. That decision "reinstate[s] [a] citizenship question on the 2020 decennial census." App. to Pet. for Cert. 549a-550a (Memorandum from Wilbur L. Ross, Jr., Secretary of Commerce, to Karen Dunn Kelley, Under Secretary for Economic Affairs (Mar. 26, 2018)). The agency's decision memorandum provided one and only one reason for making that decision-namely, that the question was "necessary to provide complete and accurate data in response to" a request from the Department of Justice (DOJ). Id. , at 562a. The DOJ had requested the citizenship question for "use [in] ... determining violations of Section 2 of the Voting Rights Act." Id. , at 548a.

 The decision memorandum adds that the agency had not been able to "determine definitively how inclusion of a citizenship question on the decennial census will impact responsiveness. However, even if there is some impact on responses, the value of more complete and accurate data derived from surveying the entire population outweighs such concerns." Id. , at 562a. The Secretary's decision thus rests upon a weighing of potentially adverse consequences (diminished responses and a less accurate census count) against potentially offsetting advantages (better citizenship data). In my view, however, the Secretary did not make reasonable decisions about these potential costs and benefits in light of the administrative record.

 A

 Consider first the Secretary's conclusion that he was "not able to determine definitively how inclusion of a citizenship question on the decennial census will impact responsiveness." Ibid. Insofar as this statement implies that adding the citizenship question is unlikely to affect "responsiveness" very much (or perhaps at all), the evidence in the record indicates the contrary.

 1

 The administrative record includes repeated Census Bureau statements that adding the question would produce a less accurate count because noncitizens and Hispanics would be less likely to respond to the questionnaire. See App. 105, 109-112, 158. The Census Bureau's chief scientist said specifically that adding the question would have "an adverse impact on self-response and, as a result, on the accuracy and quality of the 2020 Census." Id. , at 109. And the chief scientist backed this statement up by pointing to "[t]hree distinct analyses." Ibid.

 The first analysis compared nonresponse rates for the short-form census questionnaire (which did not include a citizenship question) to nonresponse rates for the ACS (which did). Obviously, more people fail to respond to the ACS than to the short form. Yet taking into account the fact that the nonresponse rate will be greater for the ACS than for the short form, the Bureau found that the difference *2588 between the two is yet greater for noncitizen households than for citizen households (by 5.1%, according to the Bureau). Id. , at 111. This led the Bureau to say that it was a "reasonable inference" that the presence of the citizenship question accounted for the difference. Ibid.

 The Bureau conducted two additional studies, both analyzing data from the ACS. One study looked at response rates for particular questions on the ACS. It showed that the "no answer" rate for the citizenship question was "much greater than the comparable rates" for other census questions (for example, questions about age, sex, race, and ethnicity). Id. , at 110. And it showed that the "no answer" rate for the citizenship question was significantly higher among Hispanics. Id. , at 109-110. The last study examined "break-off" rates, i.e. , the rate at which respondents stopped answering the questionnaire upon reaching a particular question. It found that Hispanics were significantly more likely than were non-Hispanics to stop answering at the point they reached the citizenship question. Id. , at 112. Together, these two studies provided additional support for the Census Bureau's determination that the citizenship question is likely to mean disproportionately fewer responses from noncitizens and Hispanics than from others. Ibid.

 Putting numbers upon these study results, the Census Bureau estimated that adding the question to the short form would lead to 630,000 additional nonresponding households. Id. , at 114. That is to say, the question would cause households covering more than 1 million additional people to decline to respond to the census. When the Bureau does not receive a response, it follows up with in-person interviews in an effort to obtain the missing information. The Bureau often interviews what it calls "proxies," such as family members and neighbors. But this followup process is subject to error; and the error rate is much greater than the error rate for self-responses. Ibid. The Bureau thus explained that lower self-response rates "degrade data quality" by increasing the risk of error and leading to hundreds of thousands of fewer correct enumerations. Id. , at 113-115. The Bureau added that its estimate was "conservative." Id. , at 115. It expected "differences between citizen and noncitizen response rates and data quality" to be "amplified" in the 2020 census "compared to historical levels." Ibid. Thus, it explained, "the decrease in self-response for citizen households in 2020 could be much greater than the 5.1 percentage points [it] observed during the 2010 Census." Id. , at 115-116. Its conclusion in light of this evidence was clear. Adding the citizenship question to the short form was "very likely to reduce the self-response rate" and thereby "har[m] the quality of the census count." Id. , at 105, 158.

 The Census Bureau's analysis received support from other submissions. Several States pointed out that noncitizens and racial minorities had been undercounted in every prior census. Administrative Record 1091-1092. They also drew attention to recent surveys indicating that noncitizens had significant concerns about the confidentiality of census responses. Ibid. Former directors of the Census Bureau wrote that adding the citizenship question so late in the process "would put the accuracy of the enumeration and success of the census in all communities at grave risk." Id. , at 1057. The American Sociological Association and Census Scientific Advisory Committee echoed these warnings. See id. , at 787, 794-795. On the other hand, the Secretary received submissions by other groups that supported adding the question. See, e.g. , id. , at 1178-1179, 1206, 1276. But as far as I can tell (or as far as the *2589 arguments made here and in the District Court inform the matter), none of these latter submissions significantly added to, or detracted from, the Census Bureau's submissions in respect to the question's likely impact on response rates.

 2

 The Secretary's decision memorandum reached a quite different conclusion from the Census Bureau. The memorandum conceded that "a lower response rate would lead to ... less accurate responses." App. to Pet. for Cert. 556a. But it concluded that neither the Census Bureau nor any stakeholders had provided "definitive, empirical support" for the proposition that the citizenship question would reduce response rates. Id. , at 554a. The memorandum relied for that conclusion upon a number of considerations, but each is contradicted by the record.

 The memorandum first pointed to perceived shortcomings in the Census Bureau's analysis of nonresponse rates. It noted that response rates are generally lower overall for the long form and ACS than they are for the short form. Id. , at 552a-554a. But the Bureau explained that its analysis accounted for this consideration, see App. 111, and no one has given us reason to think the contrary. The Secretary also noted that the Bureau "was not able to isolate what percentage of [the] decline was caused by the inclusion of a citizenship question rather than some other aspect of the long form survey." App. to Pet. for Cert. 554a. But the Bureau said attributing the decline to the citizenship question was a "reasonable inference," App. 111, and again, nothing in the record contradicted the Bureau's judgment. And later analyses have borne out the Bureau's judgment that the citizenship question contributes to the decline in self-response. See, e.g. , id. , at 1002-1006, 1008 (August 2018 Census Bureau study).

 The memorandum next cast doubt on the Census Bureau's analysis of the rate at which people responded to particular questions on the ACS. It noted that the "no answer" rate to the citizenship question was comparable to the "no answer" rate for other questions on the ACS, including educational attainment, income, and property insurance. App. to Pet. for Cert. 553a. But as discussed above, the Bureau found it significant that the "no answer" rate for the citizenship question was "much greater" than the "no answer" rate for the other questions that appear on the short form -that is, the form on which the citizenship question would appear. App. 110, 124. The Secretary offered no reason why the demographic variables to which he pointed provided a better point of comparison.

 Finally, the memorandum relied on information provided by two outside stakeholders. The first was a study conducted by the private survey company Nielsen, in which questions about place of birth and time of arrival had not led to any appreciable decrease in the response rate. App. to Pet. for Cert. 552a. But Nielsen, which in fact urged the Secretary not to add the question, stated that its respondents (unlike census respondents) were paid to respond, and it is consequently not surprising that they did so. Administrative Record 1276. The memorandum also cited statements by former Census Bureau officials suggesting that empirical evidence about the question's potential impact on response rates was "limited." App. to Pet. for Cert. 558a-559a; see also id. , at 552a. But there was no reason to expect the former officials to provide more extensive empirical evidence as to a citizenship question when they were not privy to the internal Bureau analyses on this question. And, like Nielsen, the former *2590 officials strongly urged the Secretary not to ask the question. See Administrative Record 1057.

 The upshot is that the Secretary received evidence of a likely drop in census accuracy by a number somewhere in the hundreds of thousands, and he received nothing significant to the contrary. The Secretary pointed out that the Census Bureau's information was uncertain, i.e. , not "definitive." But that is not a satisfactory answer. Few public-policy-related statistical studies of risks (say, of many health or safety matters) are definitive. As the Court explained in State Farm , "[i]t is not infrequent that the available data do not settle a regulatory issue, and the agency must then exercise its judgment in moving from the facts and probabilities on the record to a policy conclusion." 463 U.S. at 52 , 103 S.Ct. 2856 . But an agency confronted with this situation cannot "merely recite the terms 'substantial uncertainty' as a justification for its actions." Ibid. Instead, it "must explain the evidence which is available" and typically must offer a reasoned explanation for taking action without "engaging in a search for further evidence." Ibid.

 The Secretary did not do so here. He did not explain why he made the decision to add the question without following the Bureau's ordinary practice of extensively testing proposed changes to the census questionnaire. See App. 624-630, 641 (discussing testing process); see also, e.g. , Brief for Former Census Bureau Directors as Amici Curiae 17-21 (discussing prior examples of questions that the Bureau decided not to add after many years of pretesting). Without that testing, the Secretary could not treat the Bureau's expert opinions and its experience with the relevant surveys as worthless merely because its conclusions were not precise. The Bureau's opinions were properly considered as evidence of likelihoods, probabilities, or risks.

 As noted above, the consequences of mistakes in the census count, of even a few hundred thousand, are grave. Differences of a few thousand people, as between one State and another, can mean a loss or gain of a congressional seat-a matter of great consequence to a State. See 351 F.Supp.3d at 594 . And similar small differences can make a large difference to the allocation of federal funds among competing state programs. Id. , at 596-597 ; see also Baldrige , 455 U.S. at 353-354 , n. 9 , 102 S.Ct. 1103 . If near-absolute certainty is what the Secretary meant by "definitive," that insistence would itself be arbitrary in light of the constitutional and statutory consequences at stake. And if the Secretary instead meant that the evidence does not indicate a serious risk of a less accurate count, that conclusion does not find support in the record.

 B

 Now consider the Secretary's conclusion that, even if adding a citizenship question diminishes the accuracy of the enumeration, "the value of more complete and accurate data derived from surveying the entire population outweighs ... concerns" about diminished accuracy. App. to Pet. for Cert. 562a (emphasis added). That conclusion was also arbitrary. The administrative record indicates that adding a citizenship question to the short form would produce less "complete and accurate data," not more.

 1

 The Census Bureau informed the Secretary that, for about 90% of the population, accurate citizenship data is available from administrative records maintained by the Social Security Administration and Internal Revenue Service. App. 146. The Bureau *2591 further informed the Secretary that it had "high confidence" that it could develop a statistical model that would accurately impute citizenship status for the remaining 10% of the population. Ibid. The Bureau stated that these methods alone-using existing administrative records for 90% of the population and statistical modeling for the remaining 10%-would yield more accurate citizenship data than also asking a citizenship question. Id. , at 159. How could that be so? The answer is somewhat technical but readily understandable.

 First , consider the 90% of the population (about 295 million people) as to whom administrative records are available. The Government agrees that using these administrative records would provide highly reliable information about citizenship, because the records "require proof of citizenship." Id. , at 117. By contrast, if responses to a citizenship question were used for this group, the Census Bureau predicted without contradiction that about one-third of the noncitizens in this group who respond would answer the question untruthfully, claiming to be citizens when they are not. Id. , at 147. Those incorrect answers-about 9.5 million in total-would conflict with the administrative records on file for those noncitizens. And what would the Census Bureau do with the conflicting data? If it accepts the answer to the citizenship question as determinative, it will have less accurate data. If it accepts the citizenship data from administrative records as determinative, asking the question will have served no purpose.

 Thus, as to 295 million people-the overwhelming majority of the population-asking the citizenship question would at best add nothing at all. I say "at best" because, for one thing, the Census Bureau informed the Secretary that asking the question would produce 1 million more people who could not be linked to administrative records, which in turn would require the Census Bureau to resort to a less accurate source of citizenship data for these people. See id. , at 147-149; see also 351 F.Supp.3d at 538 -539 . For another, the policy of the Census Bureau has always been to use census responses rather than administrative records in cases where the two conflict. App. 147. In this case, that practice would mean accepting 9.5 million inaccurate responses even though accurate administrative records are available. See ibid. The Census Bureau could perhaps change that practice, but the Secretary's decision memorandum said nothing about the matter. It did not address the problem.

 Second , consider the remaining 10% of the population (about 35 million people) for whom the Government lacks administrative records. The question here is which approach would yield the most "complete and accurate" citizenship data for this group-adding a citizenship question or using statistical modeling alone? To answer this question, we must further divide this group into two categories-those who would respond to the citizenship question if it were asked and those who would not.

 Start with the category of about 22 million people who would answer a citizenship question if it were asked. Would their answers regarding citizenship be more accurate than citizenship data produced by statistical modeling? The Census Bureau said no. That is because many of the noncitizens in this group would answer the question falsely, resulting in an estimated 500,000 inaccurate answers. See id. , at 148. And those who answer the question falsely would be commingled, perhaps randomly, with those who answer it correctly, thereby casting doubt on the answers of all 22 million, with no way of knowing which answers are correct and which are false. By contrast, the Bureau believed that it could develop a statistical model that *2592 would produce more accurate citizenship data than these census responses. The Bureau therefore informed the Secretary that it could do better. As the Bureau's chief scientist explained, although "[o]ne might think" that asking the question "could help fill the ... gaps" in the administrative records, the data did not support that assumption. Id. , at 157. Instead, he explained, responses to the citizenship question "may not be reliable," which "calls into question their ability to improve upon" the Bureau's statistical modeling process. Ibid.

 Next, turn to the more than 13 million remaining people who would not answer the citizenship question even if it were asked. As to this category, the Census Bureau would still need to use statistical modeling to obtain citizenship data, because there would be no census response to use instead. Hence, asking the citizenship question would add nothing at all as to this group. To the contrary, as the Government concedes, asking the question would reduce the accuracy of the citizenship data for this group, because the relatively inaccurate answers to the citizenship question would diminish the overall accuracy of the Census Bureau's statistical model. See Brief for Petitioners 34 (conceding that the Census Bureau model will be "highe[r] quality" without the question than with it); 351 F.Supp.3d at 640 (explaining that asking the question would "corrup[t] ... the data generated by extrapolating from self-responses through imputation").

 In sum, in respect to the 295 million persons for whom administrative records exist, asking the question on the short form would, at best, be no improvement over using administrative records alone. And in respect to the remaining 35 million people for whom no administrative records exist, asking the question would be no better, and in some respects would be worse, than using statistical modeling. The Census Bureau therefore told the Secretary that asking the citizenship question, even in addition to using administrative records, "would result in poorer quality citizenship data" than using administrative records alone, and would "still have all the negative cost and quality implications" of asking the citizenship question. App. 159. I could find no evidence contradicting that prediction.

 2

 If my description of the record is correct, it raises a serious legal problem. How can an agency support the decision to add a question to the short form, thereby risking a significant undercount of the population, on the ground that it will improve the accuracy of citizenship data, when in fact the evidence indicates that adding the question will harm the accuracy of citizenship data? Of course it cannot. But, as I have just said, I have not been able to find evidence to suggest that adding the question would result in more accurate citizenship data. Neither could the District Court. After reviewing the record in detail, the District Court found that "all of the relevant evidence before Secretary Ross- all of it-demonstrated that using administrative records ... would actually produce more accurate [citizenship] data than adding a citizenship question to the census." 351 F.Supp.3d at 650 .

 What consideration did the Secretary give to this problem? He stated simply that "[a]sking the citizenship question of 100 percent of the population gives each respondent the opportunity to provide an answer," which "may eliminate the need for the Census Bureau to have to impute an answer for millions of people." App. to Pet. for Cert. 556a. He therefore must have assumed, sub silentio , exactly what *2593 the Census Bureau experts urged him not to assume-that answers to the citizenship question would be more accurate than statistical modeling. And he ignored the undisputed respects in which asking the question would make the existing data less accurate. Other than his assumption, the Secretary said nothing, absolutely nothing, to suggest a reasoned basis for disagreeing with the Bureau's expert statistical judgment.

 The Government now maintains that the Secretary reasonably discounted the Census Bureau's recommendation because it was based on an untested prediction about the accuracy of its model. But this is not a case in which the Secretary was presented with a policy choice between two reasonable but uncertain options. For one thing, the record is much less uncertain than the Government acknowledges. Although it is true that the Census Bureau at one point told the Secretary that it could not "quantify the relative magnitude of the errors across the alternatives at this time," App. 148, it unequivocally stated that asking the question " would result in poorer quality citizenship data " than omitting it, id. , at 159 (emphasis added). Thus, even if the Bureau could not "quantify" the relative accuracy of the options, it could and did conclude that one option was likely more accurate than the other. Even in the face of some uncertainty, where all available evidence indicates that one option is better than the other, it is unreasonable to choose the worse option without explanation.

 For another thing, to the extent the record reflects some uncertainty regarding the accuracy of the Census Bureau's statistical model, that is because the model needed to be "developed and tested" before it could be employed. Id. , at 146. But the Secretary made his decision before any such development or testing could be completed. Having decided to make an immediate decision rather than wait for testing, the Secretary could not dismiss the Bureau's prediction about the inadvisability of that decision on the ground that the prediction reflected likelihoods, probabilities, and risks rather than certainties.

 Finally, recall that the Census Act requires the Secretary to use administrative records rather than direct inquiries to "the maximum extent possible." 13 U. S. C. § 6 (c). That statutory requirement highlights what should be obvious: Whether adding a citizenship question to the short form would produce more accurate citizenship data is a relevant factor-indeed, a critically important factor-that the Secretary was required to consider. Here, the Secretary did not adequately explain why he rejected the evidence that adding the question would yield less accurate data. He did not even acknowledge that the Census Act obliged him to use administrative records rather than asking a question to the extent possible. And he did not explain how obtaining citizenship data that is no better or worse than the data otherwise available could justify jeopardizing the accuracy of the census count.

 In these respects, the Secretary failed to consider "important aspect[s] of the problem" and "offered an explanation for [his] decision that runs counter to the evidence before the agency." State Farm , 463 U.S. at 43 , 103 S.Ct. 2856 .

 C

 The Secretary's failure to consider this evidence-that adding the question would harm the census count in the interest of obtaining less accurate citizenship data-provides a sufficient basis for setting the decision aside. But there is more. The reason that the Secretary provided for needing more accurate citizenship information in the first place-to help the DOJ

 *2594 enforce the Voting Rights Act-is unconvincing.

 The Secretary stated that adding the citizenship question was "necessary to provide complete and accurate data in response to the DOJ request." App. to Pet. for Cert. 562a. The DOJ's request in turn asserted that the citizenship data currently available from the ACS was not "ideal" for enforcing the Voting Rights Act. Id. , at 567a. One of the DOJ's principal complaints was that ACS data is reported for groups of census blocks rather than for each census block itself. The DOJ letter stated that adding a citizenship question could provide it with individual block-by-block data which, the DOJ maintained, would allow it to better enforce the Voting Rights Act's protections for minority voters. Id. , at 568a.

 This rationale is difficult to accept. One obvious problem is that the DOJ provided no basis to believe that more precise data would in fact help with Voting Rights Act enforcement. Congress enacted the Voting Rights Act in 1965-15 years after the census last asked every household about citizenship. Actions to enforce the Act have therefore always used citizenship data derived from sampling. Yet I am aware of no one-not in the Department of Commerce proceeding, in the District Court, or in this Court-who has provided a single example in which enforcement of the Act has suffered due to lack of more precise citizenship data. Organizations with expertise in this area tell us that asking the citizenship question will not help enforce the Act. See, e.g. , Brief for NAACP Legal Defense & Educational Fund, Inc., as Amicus Curiae 30-36. Rather, the question will, by depressing the count of minority groups, hurt those whom the Act seeks to help. See, e.g. , Brief for Leadership Conference on Civil and Human Rights et al. as Amici Curiae 21-29.

 Another problem with the Secretary's rationale is that, even assuming the DOJ needed more detailed citizenship data, there were better ways of obtaining the needed data. The Census Bureau offered to provide the DOJ with data using administrative records, which, as I have pointed out, are likely just as accurate, if not more accurate, than responses to a citizenship question. The Census Bureau offered to provide this data at the census block level, which would resolve each of the DOJ's complaints about the existing ACS data. See Administrative Record 3289. But the Secretary rejected this alternative without explaining why it would not fully respond to the DOJ's request. That failure was particularly problematic given that the Census Act requires the Secretary to use other methods of obtaining demographic information if at all possible. See §§ 6(c), 195.

 Normally, the Secretary would be entitled to place considerable weight upon the DOJ's expertise in matters involving the Voting Rights Act, but there are strong reasons for discounting that expertise here. The administrative record shows that DOJ's request to add a citizenship question originated not with the DOJ, but with the Secretary himself. See Administrative Record 3710. The Voting Rights Act rationale was in fact first proposed by Commerce Department officials. See ibid. DOJ officials, for their part, were initially uninterested in obtaining more detailed citizenship data, App. 414, and they agreed to request the data only after the Secretary personally spoke to the Attorney General about the matter, see Administrative Record 2651. And when the acting director of the Census Bureau proposed alternative means of obtaining better citizenship data, DOJ officials declined to meet to discuss the proposal. See id. , at 3460.

 *2595 Taken as a whole, the evidence in the administrative record indicates that the Voting Rights Act rationale offered by the Secretary was not just unconvincing, but pretextual. And, as the Court concludes, further evidence outside the administrative record but present in the trial record supports the finding of pretext. See Part V, ante. Among other things, that evidence reveals that the DOJ official who wrote the letter agreed that adding the question "is not necessary for DOJ's VRA enforcement efforts." App. 1113. And that official further acknowledged that he did not "know whether or not [citizenship] data produced from responses to the citizenship question ... will, in fact, be more precise than the [citizenship] data on which the DOJ is currently relying for purposes of VRA enforcement." Id. , at 1102.

 The Court explains, and I agree, that a court normally should not "reject an agency's stated reasons for acting simply because the agency might also have had other unstated reasons." Ante , at 2573. But in this case, "the evidence tells a story that does not match the explanation the Secretary gave for his decision." Ante , at 2575. This evidence strongly suggests that the Secretary's stated rationale was pretextual. I consequently join Part V of the Court's opinion (except insofar as it concludes that the Secretary's decision was reasonable apart from the question of pretext). And I agree that the pretextual nature of the Secretary's decision provides a sufficient basis to affirm the District Court's decision to send the matter back to the agency.

 * * *

 I agree with the Court that the APA gives agencies broad leeway to carry out their legislatively delegated duties. And I recognize that Congress has specifically delegated to the Secretary of Commerce the authority to conduct a census of the population "in such form and content as he may determine." § 141(a). But although this delegation is broad, it is not without limits. The APA supplies one such limit. In an effort to ensure rational decisionmaking, the APA prohibits an agency from making decisions that are "arbitrary, capricious, [or] an abuse of discretion." 5 U. S. C. § 706 (2)(A).

 This provision, of course, does not insist that decisionmakers think through every minor aspect of every problem that they face. But here, the Secretary's decision was a major one, potentially affecting the proper workings of our democratic government and the proper allocation of hundreds of billions of dollars in federal funds. Cf. ante , at 2565 - 2566. Yet the decision was ill considered in a number of critically important respects. The Secretary did not give adequate consideration to issues that should have been central to his judgment, such as the high likelihood of an undercount, the low likelihood that a question would yield more accurate citizenship data, and the apparent lack of any need for more accurate citizenship data to begin with. The Secretary's failures in considering those critical issues make his decision unreasonable. They are the kinds of failures for which, in my view, the APA's arbitrary and capricious provision was written.

 As I have said, I agree with the Court's conclusion as to pretext and with the decision to send the matter back to the agency. I do not agree, however, with several of the Court's conclusions concerning application of the arbitrary and capricious standard. In my view, the Secretary's decision-whether pretextual or not-was arbitrary, capricious, and an abuse of his lawfully delegated discretion. I consequently concur in the Court's judgment to the extent that it affirms the judgment of the District Court.

 The annual alien registration requirement was repealed in 1981. See § 11, 95 Stat. 1617 (1981).

 Several months after the Secretary made his decision, the Bureau updated its prediction to 5.8%, the figure the District Court later relied on in its standing analysis. See 351 F.Supp.3d 502 , 579 (SDNY 2019).

 The full text of subsections (a) and (b) provides:

 "(a) The Secretary, whenever he considers it advisable, may call upon any other department, agency, or establishment of the Federal Government, or of the government of the District of Columbia, for information pertinent to the work provided for in this title.
 "(b) The Secretary may acquire, by purchase or otherwise, from States, counties, cities, or other units of government, or their instrumentalities, or from private persons and agencies, such copies of records, reports, and other material as may be required for the efficient and economical conduct of the censuses and surveys provided for in this title." 13 U. S. C. § 6 .

 As a 2016 Census Bureau guidance document explained, obtaining citizenship statistics is "essential for agencies and policy makers setting and evaluating immigration policies and laws, understanding how different immigrant groups are assimilated, and monitoring against discrimination." Dept. of Commerce, Census Bureau, American Community Survey, Why We Ask: Place of Birth, Citizenship and Year of Entry, www2.census.gov/programs-surveys/acs/about/qbyqfact/2016/Citizenship.pdf (all Internet materials as last visited June 25, 2019).

 United Nations, Dept. of Economic and Social Affairs Statistics Div., Principles and Recommendations for Population and Housing Censuses 163, 191 (rev. 3, 2017).

 See, e.g. , Brief for Petitioners 29 (" '[O]ther major democracies inquire about citizenship on their census, including Australia, Canada, France, Germany, Indonesia, Ireland, Mexico, Spain, and the United Kingdom, to name a few' " (quoting App. to Pet. for Cert. 561a)).

 We do not have before us a claim that information outside the administrative record calls into question the legality of an agency action based on an unstated, unlawful bias or motivation ( e.g. , a claim of religious discrimination under the Free Exercise Clause). But to the extent such a claim is viable, the analysis would have nothing to do with the arbitrary-and-capricious review pressed by respondents. See §§ 706(2)(A)-(C) (addressing agency actions that violate "constitutional" or "statutory" requirements, or that "otherwise [are] not in accordance with law").

 Insofar as Overton Park authorizes an exception to review on the administrative record, it has been criticized as having "no textual grounding in the APA" and as "created by the Court, without citation or explanation, to facilitate Article III review." Gavoor & Platt, Administrative Records and the Courts, 67 U. Kan. L. Rev. 1 , 44 (2018) ; see id., at 22 (further arguing that the exception was "neither presented by the facts of the case nor briefed by the parties"). The legitimacy and scope of the exception-which by its terms contemplates only "administrative officials who participated in the decision ... giv[ing] testimony explaining their action," Overton Park , 401 U.S. at 420 , 91 S.Ct. 814 -is an important question that may warrant future consideration. But because the Court's holding is incorrect regardless of the validity of the Overton Park exception, I will apply it here.

Justice ALITO, concurring in part and dissenting in part.
 *2596 It is a sign of our time that the inclusion of a question about citizenship on the census has become a subject of bitter public controversy and has led to today's regrettable decision. While the decision to place such a question on the 2020 census questionnaire is attacked as racist, there is a broad international consensus that inquiring about citizenship on a census is not just appropriate but advisable. No one disputes that it is important to know how many inhabitants of this country are citizens. 1 And the most direct way to gather this information is to ask for it in a census. The United Nations recommends that a census inquire about citizenship, 2 and many countries do so. 3

 Asking about citizenship on the census also has a rich history in our country. Every census, from the very first one in 1790 to the most recent in 2010, has sought not just a count of the number of inhabitants but also varying amounts of additional demographic information. In 1800, Thomas Jefferson, as president of the American Philosophical Society, signed a letter to Congress asking for the inclusion on the census of questions regarding " 'the respective numbers of native citizens, citizens of foreign birth, and of aliens' " " 'for the purpose ... of more exactly distinguishing the increase of population by birth and immigration.' " C. Wright, History and Growth of the United States Census (prepared for the Senate Committee on the Census), S. Doc. No. 194, 56th Cong., 1st Sess., 19 (1900). In 1820, John Quincy Adams, as Secretary of State, was responsible for conducting the census, and consistent with the 1820 Census Act, he instructed the marshals who were charged with gathering the information to ask about citizenship. 4 In 1830, when Martin Van Buren was Secretary of State, a question about citizenship was again included. 5 With the exception of the census of 1840, at least some portion of the population was asked a question about citizenship as part of the census through 2000, after which the question was moved to the American Community Survey, which is sent to only a small fraction of the population. All these census inquiries were made by the Executive pursuant to congressional authorization. None were reviewed by the courts.

 Now, for the first time, this Court has seen fit to claim a role with respect to the inclusion of a citizenship question on the census, and in doing so, the Court has set a dangerous precedent, both with regard *2597 to the census itself and with regard to judicial review of all other executive agency actions. For the reasons ably stated by Justice THOMAS, see ante , p. ---- (opinion concurring in part and dissenting in part), today's decision is either an aberration or a license for widespread judicial inquiry into the motivations of Executive Branch officials. If this case is taken as a model, then any one of the approximately 1,000 district court judges in this country, upon receiving information that a controversial agency decision might have been motivated by some unstated consideration, may order the questioning of Cabinet officers and other high-ranking Executive Branch officials, and the judge may then pass judgment on whether the decision was pretextual. What Bismarck is reputed to have said about laws and sausages comes to mind. And that goes for decisionmaking by all three branches.

 To put the point bluntly, the Federal Judiciary has no authority to stick its nose into the question whether it is good policy to include a citizenship question on the census or whether the reasons given by Secretary Ross for that decision were his only reasons or his real reasons. Of course, we may determine whether the decision is constitutional. But under the considerations that typically guide this Court in the exercise of its power of judicial review of agency action, we have no authority to decide whether the Secretary's decision was rendered in compliance with the Administrative Procedure Act (APA).

 I

 The APA authorizes judicial review of "agency action" taken in violation of law, 5 U. S. C. §§ 706 (2)(A)-(D), but § 701(a)(2) of the APA bars judicial review of agency actions that are "committed to agency discretion by law." Although we have characterized the scope of § 701(a)(2) as " 'narrow,' " Heckler v. Chaney , 470 U.S. 821 , 830, 105 S.Ct. 1649 , 84 L.Ed.2d 714 (1985), there are circumstances in which it applies. And while our cases recognize a strong presumption in favor of judicial review of agency action, see, e.g. , Weyerhaeuser Co. v. United States Fish and Wildlife Serv. , 586 U. S. ----, ----, 139 S.Ct. 361 , 370, 202 L.Ed.2d 269 (2018), this "is 'just' a presumption," and like all real presumptions, it may be (and has been) rebutted, Lincoln v. Vigil , 508 U.S. 182 , 190, 113 S.Ct. 2024 , 124 L.Ed.2d 101 (1993). 6

 In considering whether the general presumption in favor of judicial review has been rebutted in specific cases, we have identified factors that are relevant to the inquiry: whether the text and structure of the relevant statutes leave a court with any " 'meaningful standard against which to judge the agency's exercise of discretion,' " Webster v. Doe , 486 U.S. 592 , 600, 108 S.Ct. 2047 , 100 L.Ed.2d 632 (1988) (quoting Heckler , supra , at 830, 105 S.Ct. 1649 ); whether the matter at hand has traditionally been viewed as committed to agency discretion, see ICC v. Locomotive Engineers , 482 U.S. 270 , 282, 107 S.Ct. 2360 , 96 L.Ed.2d 222 (1987) ; whether the challenged action manifests a "general unsuitability" for judicial review because it involves a "complicated balancing of a number of factors," including judgments regarding the allocation of agency resources or matters otherwise committed to *2598 another branch, Heckler , supra , at 831-832, 105 S.Ct. 1649 ; and whether judicial review would produce "disruptive practical consequences," Southern R. Co. v. Seaboard Allied Milling Corp. , 442 U.S. 444 , 457, 99 S.Ct. 2388 , 60 L.Ed.2d 1017 (1979) (applying this factor to the reviewability inquiry under § 701(a)(1) ).

 Applying those factors, I conclude that the decision of the Secretary of Commerce to add core demographic questions to the decennial census questionnaire is committed to agency discretion by law and therefore may not be challenged under the APA. 7

 II

 A

 I start with the question whether the relevant statutory provisions provide any standard that courts can apply in reviewing the Secretary's decision to restore a citizenship question to the census. The provision that directly addresses this question is 13 U. S. C. § 141 (a), the statute that vests the Secretary with authority to administer the decennial census. This provision gives the Secretary unfettered discretion to include on the census questions about basic demographic characteristics like citizenship. It begins by providing that the Secretary

 "shall, in the year 1980 and every 10 years thereafter, take a decennial census of population ... in such form and content as he may determine , including the use of sampling procedures and special surveys." Ibid. (emphasis added).
 The two phrases I have highlighted-"census of population" and "in such form and content as he may determine"-are of immediate importance. A "census of population" is broader than a mere head count. The term is defined as "a census of population ... and matters relating to population ." § 141(g) (emphasis added). Because this definition refers to both "a census of population" and "matters relating to population," the latter concept must include more than a "census of population" in the strict sense of a head count. And it seems obvious that what this additional information must include is the sort of basic demographic information that has long been sought in the census. So the statute clearly authorizes the Secretary to gather such information.

 The second phrase, "in such form and content as he may determine," specifies how this information is to be gathered, namely, by a method having the "form and content" that the Secretary "may determine." In other words, this is left purely to the Secretary's discretion. A clearer and less restricted conferral of discretion is hard to imagine.

 It is instructive to compare this delegation of authority to the statutory language at issue in one of our most well-known § 701(a)(2) cases, Webster v. Doe , 486 U.S. 592 , 108 S.Ct. 2047 , 100 L.Ed.2d 632 . There, the relevant statute allowed termination of a Central Intelligence Agency employee whenever the Director "shall deem such termination necessary or advisable in the interests of the United States." Id. , at 600, 108 S.Ct. 2047 (internal quotation marks omitted and emphasis deleted). Reasoning that the statute's "shall deem " standard "fairly exudes *2599 deference to the Director," the Court concluded that the text of the statute "appear[ed] ... to foreclose the application of any meaningful judicial standard of review." Ibid.

 The § 141(a) language discussed above is even more sweeping than that of the statute in Webster . Unlike the Census Act, the statute in Webster placed a condition on the Director's action-in particular, the requirement that he terminate an employee only after concluding that doing so would further the "interests of the United States." No such condition applies to the Secretary's determination about the form and content of the decennial census, a fact that distinguishes the statute at issue here from others this Court has found to fall outside § 701(a)(2) and thus within courts' power to review. See, e.g., Weyerhaeuser Co. , 586 U. S., at ----, 139 S.Ct., at 370 (statute conditioning agency power to exclude land from critical habitat designation on agency's consideration of " 'economic impact' " of designation and " 'determin[ation] that the benefits of such exclusion outweigh the benefits of specifying such area as part of the critical habitat' ").

 B

 Those arguing in favor of judicial review contend that the § 141(a) language that I have discussed so far is limited by language that follows immediately after. That part of § 141(a) states:

 "In connection with any such census [ i.e. , the decennial "census of population"], the Secretary is authorized to obtain such other census information as necessary ." (Emphasis added.)
 This means, it is argued, that information about citizenship may be obtained by means of the census only if that is "necessary." But this argument is clearly wrong. The information that must be "necessary" (whatever that means in this context) is " other census information." That refers to information other than that obtained in the "census of population," and as explained, the term "census of population" includes not just a head count but other "matters relating to population," a category that encompasses basic demographic information such as citizenship. Accordingly, this argument is definitively refuted by the text of § 141. And although it is not necessary to look beyond that text, it is worth noting that this argument, if accepted, would require that the term "necessary" be given a less than strictly literal meaning; otherwise, it would run contrary to the broad delegation effected by the first portion of § 141(a) by making it all but impossible for the Secretary to include on the census anything other than questions relating to the number of persons living at a particular address. That would be so because it will often not be "necessary" to obtain this information via the census rather than by some other means.

 C

 Another argument in favor of review relies on 13 U. S. C. § 195 , which states:

 "Except for the determination of population for purposes of apportionment of Representatives in Congress among the several States, the Secretary shall, if he considers it feasible, authorize the use of the statistical method known as 'sampling' in carrying out the provisions of this title."
 Justice BREYER, for example, interprets this provision to mean that "the Secretary must, if feasible, obtain demographic information through a survey sent to a sample of households, rather than through the short-form census questionnaire to which every household must respond." Ante , at 2585 (opinion concurring in part and dissenting in part). Under that reading of § 195, it is asserted, the provision sets *2600 forth a judicially reviewable limit on the Secretary's authority to obtain information through direct inquiries.

 This argument fails to take into account that the current version of § 195 was enacted as part of the same Act of Congress that included the present version of § 141 8 and that the two provisions are both parts of a unified scheme regarding the use of sampling. Section 141, a provision concerned exclusively with the census, addresses the use of sampling in that particular context. I previously quoted the relevant language, but I repeat it now so that it is clearly in mind. Section 141(a) provides that the Secretary

 "shall, in the year 1980 and every 10 years thereafter, take a decennial census of population ... in such form and content as he may determine, including the use of sampling procedures and special surveys ." (Emphasis added.)
 What this means is that the Secretary, in conducting the "census of population," has discretion to choose the form and content of the vehicles used in that project, and among the methods that he may employ, if he sees fit, are sampling and special surveys.

 Section 195 is not a census-specific provision, but it does have one (important) thing to say specifically about the census: It prohibits the use of sampling "for the determination of population for purposes of apportionment of Representatives in Congress." In this one way, it qualifies the Secretary's discretion regarding the "form and content" of the vehicles used in conducting the "census of population." And that is what we meant in Department of Commerce v. United States House of Representatives , 525 U.S. 316 , 338, 119 S.Ct. 765 , 142 L.Ed.2d 797 (1999), when we said that § 141(a) 's "broad grant of authority ... is informed ... by the narrower and more specific § 195." Otherwise, the text of § 195 does not deal specifically with the census. It addresses all the many information-gathering activities conducted by the Commerce Department, and as to these, it says that the Secretary shall use sampling if he deems it "feasible."

 If § 195 were read to mean that no information other than a head count can be sought by means of a census questionnaire unless it is not "feasible" to get that information by sampling, then there would be little if anything left of the broad discretion "to use sampling techniques" conferred on the Secretary by § 141(a). "Feasible" means "capable of being done, executed, or effected," Webster's Third New International Dictionary 831 (1961), and it is not clear that the gathering of any core demographic information is not "capable of being done" by sampling. So if that were what § 195 means, then Congress, in the same Act, would have given the Secretary discretion to use sampling in the census "as he may determine" but also compelled him to use sampling in almost all instances. That is no way to read the provisions of a single Act. A law's provisions should be read to work together. See A. Scalia & B. Garner, Reading Law 180 (2012) ("The provisions of a text should be interpreted in a way that renders them compatible, not contradictory"). See also, e.g. , Parker Drilling Management Services, Ltd. v. Newton , 587 U. S. ----, ---- - ----, 139 S.Ct. 1881 , 1887-1889, L.Ed.2d ---- (2019) (slip op., at 5-6) ; Star Athletica, L. L. C. v. Varsity Brands, Inc. , 580 U. S. ----, ---- - ----, 137 S.Ct. 1002 , 1009-1010, 197 L.Ed.2d 354 (2017) ; Kawasaki Kisen Kaisha Ltd. v. Regal-Beloit Corp. , 561 U.S. 89 , 108, 130 S.Ct. 2433 , 177 L.Ed.2d 424 (2010). And if there is tension between a specific provision, like *2601 § 141 's instruction regarding the use of sampling in the decennial census, and a general one, like § 195 's directive regarding the use of sampling in all data-collection activities, the specific provision must take precedence. Cf. NLRB v. SW General, Inc. , 580 U. S. ----, ----, 137 S.Ct. 929 , 941-942, 197 L.Ed.2d 263 (2017).

 When §§ 141 and 195 are read in this way, it is easy to see how they fit together. In using the census to gather information "relating to population" for any use other than the actual enumeration, the Secretary may use sampling "as he may determine." In conducting all the Department's efforts to collect data by other means, he may authorize the use of sampling if he thinks that is "feasible." The upshot for present purposes is that § 195 does not require the "counterintuitive resul[t]" of barring the Secretary from including on the census questionnaire the kinds of basic demographic questions that have been asked as part of every census in U. S. history. RJR Nabisco, Inc. v. European Community , 579 U. S. ----, ----, 136 S.Ct. 2090 , 2104, 195 L.Ed.2d 476 (2016).

 D

 One additional provision, 13 U. S. C. § 6 (c), 9 requires close consideration. This provision, which was enacted in 1976 in the same Act as §§ 141(a) and 195, has three subsections. Subsection (a) provides that the Secretary may call on other components of the Federal Government to obtain information that is "pertinent to" the Department's work. Subsection (b) authorizes the Secretary to "acquire, by purchase or otherwise" from state and local governments and private sources "such copies of records, reports, and other material as may be required for the efficient and economical conduct of the censuses and surveys provided for in this title." Finally, subsection (c) provides:

 "To the maximum extent possible and consistent with the kind, timeliness, quality and scope of the statistics required, the Secretary shall acquire and use information available from any source referred to in subsection (a) or (b) of this section instead of conducting direct inquiries."
 The District Court interpreted subsection (c) to mean that the Secretary must turn to another federal agency or outside source for demographic information (rather than seeking the information on the census) unless doing so would not be "possible" or "consistent with the kind, timeliness, quality and scope of the statistics required." This argument fails for reasons similar to those that sank the § 195 argument just discussed. Section 6(c) is not a census-specific provision but instead applies generally to all the Commerce Department's information-gathering activities. If it is read to apply to the "census of population," it cannot be reconciled with § 141(a), which, as noted, broadly authorizes the Secretary to use that vehicle for obtaining information "relating to population,"

 *2602 i.e. , core demographic information. If § 6(c) applied to the gathering of such information, it would make it hard to justify the inclusion of any demographic questions on the census, even though this has been done since 1790. (Is it not possible to get information about age and sex, for example, from any outside source (or combination of sources), even if the Department offers to acquire it from a private source by purchase?) Reading § 6(c) to mean what the District Court thought would turn it into the proverbial elephant stuffed into a mouse hole. Section 6(c), however, is a decidedly mouse-like provision. It was enacted with no fanfare and no real explanation, 10 and remained in the shadows, virtually unused and unnoticed, for more than 40 years.

 E

 Respondents and the Court cite two other provisions in support of reviewability, but neither has anything to do with the issue of putting a citizenship question on the census. In determining whether statutory provisions include standards that could provide a basis for judicial review, it is necessary to focus on the precise claims at issue, see, e.g., Webster , 486 U.S. at 601 -602 , 108 S.Ct. 2047 (distinguishing between statutory and constitutional claims); Locomotive Engineers , 482 U.S. at 277 -279 , 107 S.Ct. 2360 (parsing claims under different prongs of reopener statute); Heckler , 470 U.S. at 836 , 105 S.Ct. 1649 (rejecting as "irrelevant" to the agency decision at issue two statutory provisions that were argued to provide " 'law to apply' "). And when viewed in this way, the remaining statutory provisions cited in support of reviewability are of no value.

 Respondents point to § 141(b), which requires the Secretary to complete the tabulation of total population by States "within 9 months after the census date" and then to report the results to the President. That provision sets out an easily administered deadline, and it has nothing to do with the content of the census questionnaire.

 Respondents also claim that § 141(f) is relevant to the question of judicial review, but that provision concerns congressional review. It directs the Secretary to report to Congress, at specified times, the subjects and questions that he intends to include on the census. According to respondents, the Secretary's compliance with those requirements is judicially reviewable, and that, they contend, takes the Secretary's decision to include a citizenship question out from under § 701(a)(2).

 Respondents fundamentally misunderstand the significance of congressional reporting requirements in evaluating whether a particular agency action is subject to judicial review. Congressional reporting requirements are "legion in federal law," Natural Resources Defense Council, Inc. v. Hodel , 865 F.2d 288 , 317 (CADC 1988), and their purpose is to permit Congress to monitor and, if it sees fit, to correct Executive Branch actions to which it objects. When a congressional reporting requirement "[l]ack[s] a provision for judicial review," compliance "by its nature seems singularly committed to congressional discretion in measuring the fidelity of the *2603 Executive Branch actor to legislatively mandated requirements." Id. , at 318. In other words, it is Congress, not the Judiciary, that is best situated to determine whether an agency's responses to Congress are sufficient and, if not, to "take what it deems to be the appropriate action." Id. , at 319.

 In that respect, § 141(f) actually cuts against judicial review. The Constitution gives Congress the authority to "direct" the "Manner" in which the census is conducted, and by imposing the § 141(f) reporting requirements, Congress retained some of that supervisory authority. It did not transfer it to the courts. 11

 Respondents protest that congressional review may not be enough to guard against a Secretary's abuses, especially when the party in control of Congress stands to benefit. But that complaint simply expresses disagreement with the Framers' choice to vest power over the census in a political body, cf. Baldrige v. Shapiro , 455 U.S. 345 , 347-348, 102 S.Ct. 1103 , 71 L.Ed.2d 199 (1982) ("Under [the] Constitution, responsibility for conducting the decennial census rests with Congress"), and the manner in which Congress has chosen to exercise that power, see Wisconsin v. City of New York , 517 U.S. 1 , 19, 116 S.Ct. 1091 , 134 L.Ed.2d 167 (1996) (Congress has delegated its "virtually unlimited discretion" in conducting the census to the Secretary). In any event, the ability to press constitutional challenges to the Secretary's decisions, see n. 7, supra, answers many of the examples in respondents' parade of horribles.

 In short, the relevant text of § 141(a) "fairly exudes deference" to the Secretary. Webster , 486 U.S. at 600 , 108 S.Ct. 2047 . And no other provision of law cited by respondents or my colleagues provides any "meaningful judicial standard" for reviewing the Secretary's selection of demographic questions for inclusion on the census. Ibid.

 III

 In addition to requiring an examination of the text and structure of the relevant statutes, our APA § 701(a)(2) cases look to whether the agency action in question is a type that has traditionally been viewed as committed to agency discretion or whether it is instead one that "federal courts regularly review." Weyerhaeuser Co. , 586 U. S., at ----, 139 S.Ct., at 370. In cases where the Court has found that agency action is committed to agency discretion by law, an important factor has been the absence of an established record of judicial review prior to the adoption of the APA. See Heckler , 470 U.S. at 832-833 , 105 S.Ct. 1649 (agency nonenforcement); Locomotive Engineers , 482 U.S. at 282 , 107 S.Ct. 2360 (agency decision not to reopen final decision based on material error); Lincoln , 508 U.S. at 192 , 113 S.Ct. 2024 (agency use of lump-sum appropriations).

 *2604 Here, there is no relevant record of judicial review. We are confronted with a practice that reaches back two centuries. The very first census went beyond a mere head count and gathered additional demographic information, and during virtually the entire period prior to the enactment of the APA, a citizenship question was asked of everyone. Notably absent from that long record is any practice of judicial review of the content of the census. Indeed, this Court has never before encountered a direct challenge to a census question. App. to Pet. for Cert. 416a. And litigation in the lower courts about the census is sparse and generally of relatively recent vintage.

 Not only is this sort of history significant in all § 701(a)(2) cases, see Locomotive Engineers , supra , at 282, 107 S.Ct. 2360 , but we have previously stressed the particular "importance of historical practice" when it comes to evaluating the Secretary's authority over the census. Wisconsin , supra , at 21, 116 S.Ct. 1091 ; see also ante, at 2567 (opinion of the Court). Moreover, where the relevant question is not whether review may be had at all, but rather the branch with the authority to exercise review, the absence of any substantial record of judicial review is especially revealing. See, e.g., NLRB v. Noel Canning , 573 U.S. 513 , 525, 134 S.Ct. 2550 , 189 L.Ed.2d 538 (2014) (it is "neither new nor controversial" that "longstanding practice of the government can inform our determination of what the law is" (internal quotation marks and citation omitted)); United States v. Midwest Oil Co. , 236 U.S. 459 , 473, 35 S.Ct. 309 , 59 L.Ed. 673 (1915) ("in determining ... the existence of a power, weight [is] given to ... usage"). Thus, the absence of any real tradition of judicial review of decisions regarding the content of the census counsels against review in this case.

 In an attempt to show that there is no relevant "tradition of nonreviewability," Locomotive Engineers , supra , at 282, 107 S.Ct. 2360 , respondents contend that this Court has recently engaged in review of the "conduct of the census," Brief for Government Respondents 26-27. But in none of the cases they cite did the Court address an APA challenge to the content of census questions. 12 Some involved constitutional claims about enumeration and apportionment. See Franklin v. Massachusetts , 505 U.S. 788 , 790, 801, 112 S.Ct. 2767 , 120 L.Ed.2d 636 (1992) (constitutional challenge to "method used for counting federal employees serving overseas" as part of "reapportionment determination"); Wisconsin , 517 U.S. at 20 , 116 S.Ct. 1091 (constitutional challenge to Secretary's decision not to adjust count). Others concerned enforcement of statutes with specific directives. See Department of Commerce , 525 U.S. at 343 , 119 S.Ct. 765 (holding that § 195 bars use of "sampling" to reach actual enumeration for apportionment); Utah v. Evans , 536 U.S. 452 , 464-465, 122 S.Ct. 2191 , 153 L.Ed.2d 453 (2002) (considering whether statistical method violated § 195 's bar on use of "sampling" in apportionment enumeration). According to respondents, these cases mean that all the Secretary's census-related decisions are suitable for judicial review and thus fall outside of § 701(a)(2), and the Court apparently agrees, rejecting the Government's § 701(a)(2) argument in part because "[w]e and other courts have entertained both constitutional and statutory challenges to census-related decisionmaking." Ante , at 2568.

 *2605 This argument misses the point of § 701(a)(2). The question under that provision is whether the challenged action "is committed to agency discretion by law," not whether a different action by the same agency is reviewable under the APA, much less whether an action taken by the same agency can be challenged under the Constitution. Take the example of Heckler v. Chaney , supra , where the Court considered whether a particular Food and Drug Administration (FDA) decision was reviewable under the APA. Many FDA actions are subject to APA review, see, e.g., Weinberger v. Hynson, Westcott & Dunning, Inc. , 412 U.S. 609 , 627, 93 S.Ct. 2469 , 37 L.Ed.2d 207 (1973), but that did not prevent the Heckler Court from holding that the particular FDA decision at issue there fell within § 701(a)(2). See also, e.g. , Heckler , supra, at 836-837, 105 S.Ct. 1649 .

 Respondents and some of their amici contend that the Secretary's decision is at least amenable to judicial review for consistency with the APA's reasoned-explanation requirement. See Motor Vehicle Mfrs. Assn. of United States, Inc. v. State Farm Mut. Automobile Ins. Co. , 463 U.S. 29 , 43, 103 S.Ct. 2856 , 77 L.Ed.2d 443 (1983) (describing requirement). Thus, the argument goes, even if no statute sets out a standard that can be used in reviewing the particular agency action in question, a court may review an agency's explanation of the reasons for its action and set it aside if the court finds those reasons to be arbitrary or irrational.

 This argument would obliterate § 701(a)(2). Even if a statute expressly gave an agency absolute, unrestricted, unfettered, unlimited, and unqualified discretion with respect to a particular decision, a court could still review the agency's explanation of the reasons for its decision. That is not what § 701(a)(2) means. As we put it previously in answering a similar argument against application of § 701(a)(2), it is "fals[e]" to suggest "that if the agency gives a 'reviewable' reason for otherwise unreviewable action, the action becomes reviewable." Locomotive Engineers , 482 U.S. at 283 , 107 S.Ct. 2360 . That is because when an action "is committed to agency discretion by law," the Judiciary has no role to play, even when an agency sets forth "an eminently 'reviewable' proposition." Id., at 282-283 , 107 S.Ct. 2360 .

 IV

 In sum, neither respondents nor my colleagues have been able to identify any relevant, judicially manageable limits on the Secretary's decision to put a core demographic question back on the census. And without an "adequate standard of review for such agency action," id. , at 282, 107 S.Ct. 2360 , courts reviewing decisions about the "form and content" of the census would inevitably be drawn into second-guessing the Secretary's assessment of complicated policy tradeoffs, 13 another indicator of "general unsuitability" for judicial review. Heckler , supra , at 831, 105 S.Ct. 1649 .

 Indeed, if this litigation is any indication, widespread judicial review of the Secretary's conduct of the census will usher in an era of "disruptive practical consequences,"

 *2606 and this too weighs against review. Seaboard Allied Milling Corp. , 442 U.S. at 457 , 99 S.Ct. 2388 . Cf. Tucker v. United States Dept. of Commerce , 958 F.2d 1411 , 1418 (CA7 1992) (expressing doubt about "both the provenance and the practicability" of allowing judicial review of census-related decisions).

 Respondents protest that the importance of the census provides a compelling reason to allow APA review. See also ante, at 2595 (opinion of BREYER, J.). But this argument overlooks the fact that the Secretary is accountable in other ways for census-related decisionmaking. 14 If the Secretary violates the Constitution or any applicable statutory provision related to the census, his action is reviewable. The Secretary is also accountable to Congress with respect to the administration of the census since he has that power only because Congress has found it appropriate to entrust it to him. And the Secretary is always answerable to the President, who is, in turn, accountable to the people.

 * * *

 Throughout our Nation's history, the Executive Branch has decided without judicial supervision or interference whether and, if so, in what form the decennial census should inquire about the citizenship of the inhabitants of this country. Whether to put a citizenship question on the 2020 census questionnaire is a question that is committed by law to the discretion of the Secretary of Commerce and is therefore exempt from APA review. The District Court had the authority to decide respondents' constitutional claims, but the remainder of their complaint should have been dismissed.

 I join Parts I, II, III, IV-B, and IV-C 15 of the opinion of the Court. I do not join the remainder, and insofar as the Court holds that the Secretary's decision is reviewable under the APA, I respectfully dissent.

 Justice KAVANAUGH and I join Parts I, II, III, and IV of the opinion of the Court. Justice GORSUCH joins Parts I, II, III, IV-B, and IV-C.

 Justice ALITO has made a strong argument that the specific decision at issue here-whether to include a citizenship question on the census-is a matter "committed to agency discretion by law." 5 U. S. C. § 701 (a)(2) ; see post , at 2596 - 2597 (opinion concurring in part and dissenting in part). As he explains, the Secretary's decision plainly falls within the scope of the Secretary's constitutional authority, does not implicate any statutory prohibition, and is among the "inquiries" and "content[s]" of the census that the Secretary is expressly directed to "determine" for himself. §§ 5, 141(a) ; see post , at 2598 - 2603. Nevertheless, I assume, for the purpose of this opinion, that the Secretary's decision is subject to judicial review.

 Deferential review of the agency's discretionary choices and reasoning under the arbitrary-and-capricious standard stands in marked contrast to a court's plenary review of the agency's interpretation and application of the law. See §§ 706(A)-(D) (court must review agency action to ensure that it complies with all "constitutional," "statutory," and "procedur[al]" requirements, and is otherwise "in accordance with law").

 See Act of Mar. 14, 1820, ch. 24, 3 Stat. 550 ; Wright, History and Growth of the United States Census, S. Doc. No. 194, 56th Cong., 1st Sess., 133-137.

 See Dept. of Commerce, Census Bureau, History: 1830 Census Questionnaire, https://www.census.gov/history/www/through_the_decades/questionnaires/1830_2.html.

 Because the § 701(a)(2) analysis dictates whether APA review may be had, Justice BREYER's assertion that the APA "supplies [a] limit" on the Secretary's otherwise "broad" delegation, ante , at 2595 (opinion concurring in part and dissenting in part), mistakenly assumes the answer to the reviewability question. Cf. Heckler v. Chaney , 470 U.S. 821 , 828, 105 S.Ct. 1649 , 84 L.Ed.2d 714 (1985) ("[B]efore any review at all may be had, a party must first clear the hurdle of § 701(a)").

 The Government concedes that courts may review constitutional challenges to the Secretary's actions. Cf. Webster v. Doe , 486 U.S. 592 , 603, 108 S.Ct. 2047 , 100 L.Ed.2d 632 (1988). For the reasons given in the Court's opinion, see ante , at 2566 - 2567, I agree that the only remaining constitutional claim at issue-respondents' Enumeration Clause claim-lacks merit and thus does not constitute a basis for enjoining the addition of the citizenship question.

 See 90 Stat. 2459 .

 Section 6 states:

 "(a) The Secretary, whenever he considers it advisable, may call upon any other department, agency, or establishment of the Federal Government, or of the government of the District of Columbia, for information pertinent to the work provided for in this title.

 "(b) The Secretary may acquire, by purchase or otherwise, from States, counties, cities, or other units of government, or their instrumentalities, or from private persons and agencies, such copies of records, reports, and other material as may be required for the efficient and economical conduct of the censuses and surveys provided for in this title.

 "(c) To the maximum extent possible and consistent with the kind, timeliness, quality and scope of the statistics required, the Secretary shall acquire and use information available from any source referred to in subsection (a) or (b) of this section instead of conducting direct inquiries."

 The most respondents can muster are snippets from the legislative history of the 1976 Census Act indicating that § 6(c) was enacted to decrease the Secretary's use of "direct inquiries" in the interest of "reducing respondent burden." H. R. Rep. No. 94-1719, p. 10 (1976). Even accepting that premise, it simply raises the same question just discussed-namely, whether Congress's desire to reduce respondent burden, as reflected by § 6(c), yields to the Secretary's broad authorization in § 141(a) to "determine" the "form and content" of any direct inquiries on the census. Cf. id., at 11 (characterizing § 141 as a "provisio[n] directly related to decennial ... census").

 It is notable that Congress, pursuant to its supervisory authority, has in some cases limited the particular demographic characteristics about which the Secretary may require information through census questionnaires. In § 221(c), for example, Congress has dictated that "no person shall be compelled to disclose information relative to his religious beliefs or to membership in a religious body." Similarly, in a series of appropriation Acts, Congress has specified that "none of the funds provided in this or any other Act for any fiscal year may be used for the collection of census data on race identification that does not include 'some other race' as a category." 123 Stat. 3115 , note following 13 U. S. C. § 5 . Those examples highlight that when Congress wishes to limit the Secretary's authority to require responses to particular demographic questions, it "knows precisely how to do so." Limelight Networks, Inc. v. Akamai Technologies, Inc. , 572 U.S. 915 , 923, 134 S.Ct. 2111 , 189 L.Ed.2d 52 (2014).

 The same can be said for the lower court cases on which respondents rely. See, e.g., Brief for Government Respondents 26, and n. 6 (collecting cases, none of which "involved the census questionnaire" or the Secretary's selection of questions).

 In determining how the census is to be conducted, the Secretary must make decisions about a bevy of matters, such as the best way to count particular persons or categories of persons with an adequate degree of accuracy ( e.g. , by face-to-face interviews, telephone calls, questionnaires to be mailed back, contacts with neighbors, or use of existing records); the use of followup procedures and other quality control measures; which persons should be included in which households; and issues concerning where a person should be enumerated. These and countless other factors may affect whether an individual receives or responds to the census questionnaire.

 Since the time Secretary Ross publicly announced his intent to add the citizenship question, "Congress has questioned the Secretary about his decision in public hearings on several occasions." Brief for Petitioners 50 (collecting examples).

 Although I would hold that the Secretary's decision is not reviewable under the APA, in the alternative I would conclude that the decision survives review under the applicable standards. I join Parts IV-B and IV-C on that understanding.