Opinion of the Court by

Mr. Justice Marshall,

announced by Mr. Justice Stewart.

The growing public concern about the quality of our natural environment has prompted Congress in recent years to enact legislation1 designed to curb the accelerating destruction of our country’s natural beauty. We are concerned in this case with § 4 (f) of the Department of Transportation Act of 1966, as amended,2 and § 18 (a) of *405the Federal-Aid Highway Act of 1968, 82 Stat. 823, 23 U. S. C. § 138 (1964 ed., Supp. V) (hereafter § 138).3 These statutes prohibit the Secretary of Transportation from authorizing the use of federal funds to finance the construction of highways through public parks if a “feasible and prudent” 4 alternative route exists. If no such route is available, the statutes allow him to approve construction through parks only if there has been “all possible planning to minimize harm” 5 to the park.

*406Petitioners, private citizens as well as local and national conservation organizations, contend that the Secretary has violated these statutes by authorizing the expenditure of federal funds6 for the construction of a six-lane interstate highway through a public park in Memphis, Tennessee. Their claim was rejected by the District Court,7 which granted the Secretary’s motion for summary judgment, and the Court of Appeals for the Sixth Circuit affirmed.8 After oral argument, this Court granted a stay that halted construction and, treating the application for the stay as a petition for certiorari, granted review.9 400 U. S. 939. We now reverse the judgment below and remand for further proceedings in the District Court.

Overton Park is a 342-acre city park located near the center of Memphis. The park contains a zoo, a nine-hole municipal golf course, an outdoor theater, nature trails, a bridle path, an art academy, picnic areas, and 170 acres of forest. The proposed highway, which is to be a six-lane, high-speed, expressway,10 will sever the zoo from the rest of the park. Although the roadway will be depressed below ground level except where it crosses a small creek, 26 acres of the park will be destroyed. The highway is to be a segment of Interstate Highway 1-40, part of the National System of Interstate and *407Defense Highways.11 1-40 will provide Memphis with a major east-west expressway which will allow easier access to downtown Memphis from the residential areas on the eastern edge of the city.12

Although the route through the park was approved by the Bureau of Public Roads in 195613 and by the Federal Highway Administrator in 1966, the enactment of § 4 (f) of the Department of Transportation Act prevented distribution of federal funds for the section of the highway designated to go through Overton Park until the Secretary of Transportation determined whether the requirements of § 4 (f) had been met. Federal funding for the rest of the project was, however, available; and the state acquired a right-of-way on both sides of the park.14 In April 1968, the Secretary announced that he concurred in the judgment of local officials that 1-40 should be built through the park. And in September 1969 the State acquired the right-of-way inside Overton Park from the city.15 Final approval for the project — the route as well as the design— was not announced until November 1969, after Congress had reiterated in § 138 of the Federal-Aid Highway Act *408that highway construction through public parks was to be restricted. Neither announcement approving the route and design of 1-40 was accompanied by a statement of the Secretary’s factual findings. He did not indicate why he believed there were no feasible and prudent alternative routes or why design changes could not be made to reduce the harm to the park.

Petitioners contend that the Secretary’s action is invalid without such formal findings16 and that the Secretary did not make an independent determination but merely relied on the judgment of the Memphis City Council.17 They also contend that it would be “feasible and prudent” to route 1-40 around Overton Park either to the north or to the south. And they argue that if these alternative routes are not “feasible and prudent,” the present plan does not include “all possible” methods for reducing harm to the park. Petitioners claim that 1-40 could be built under the park by using either of two possible tunneling methods,18 and they claim that, at a *409minimum, by using advanced drainage techniques19 the expressway could be depressed below ground level along the entire route through the park including the section that crosses the small creek.

Respondents argue that it was unnecessary for the Secretary to make formal findings, and that he did, in fact, exercise his own independent judgment which was supported by the facts. In the District Court, respondents introduced affidavits, prepared specifically for this litigation, which indicated that the Secretary had made the decision and that the decision was supportable. These affidavits were contradicted by affidavits introduced by petitioners, who also sought to take the deposition of a former Federal Highway Administrator 20 who had participated in the decision to route 1-40 through Overton Park.

The District Court and the Court of Appeals found that formal findings by the Secretary were not necessary and refused to order the deposition of the former Federal Highway Administrator because those courts believed that probing of the mental processes of an administrative decisionmaker was prohibited. And, believing that the Secretary’s authority was wide and reviewing courts’ authority narrow in the approval of highway routes, the lower courts held that the affidavits contained no basis for a determination that the Secretary had exceeded his authority.

We agree that formal findings were not required. But we do not believe that in this case judicial review based solely on litigation affidavits was adequate.

*410A threshold question — whether petitioners are entitled to any judicial review — is easily answered. Section 701 of the Administrative Procedure Act, 5 U. S. C. § 701 (1964 ed., Supp. V), provides that the action of “each authority of the Government of the United States,” which includes the Department of Transportation,21 is subject to judicial review except where there is a statutory prohibition on review or where “agency action is committed to agency discretion by law.” In this case, there is no indication that Congress sought to prohibit judicial review and there is most certainly no “showing of 'clear and convincing evidence’ of a . . . legislative intent” to restrict access to judicial review. Abbott Laboratories v. Gardner, 387 U. S. 136, 141 (1967). Brownell v. We Shung, 352 U. S. 180, 185 (1956).22

Similarly, the Secretary’s decision here does not fall within the exception for action “committed to agency discretion.” This is a very narrow exception.23 Berger, Administrative Arbitrariness and Judicial Review, 65 Col. L. Rev. 55 (1965). The legislative history of the Administrative Procedure Act indicates that it is applicable in those rare instances where “statutes are drawn in such broad terms that in a given case there is no law to apply.” S. Rep. No. 752, 79th Cong., 1st Sess., 26 (1945).

*411Section 4 (f) of the Department of Transportation Act and § 138 of the Federal-Aid Highway Act are clear and specific directives. Both the Department of Transportation Act and the Federal-Aid Highway Act provide that the Secretary “shall not approve any program or project” that requires the use of any public parkland “unless (1) there is no feasible and prudent alternative to the use of such land, and (2) such program includes all possible planning to minimize harm to such park ... 23 U. S. C. § 138 (1964 ed., Supp. V); 49 U. S. C. § 1653 (f) (1964 ed., Supp. V). This language is a plain and explicit bar to the use of federal funds for construction of highways through parks — only the most unusual situations are exempted.

Despite the clarity of the statutory language, respondents argue that the Secretary has wide discretion. They recognize that the requirement that there be no “feasible” alternative route admits of little administrative discretion. For this exemption to apply the Secretary must find that as a matter of sound engineering it would not be feasible to build the highway along any other route.24 Respondents argue, however, that the requirement that there be no other “prudent” route requires the Secretary to engage in a wide-ranging balancing of competing interests. They contend that the Secretary should weigh the detriment resulting from the destruction of parkland against the cost of other routes, safety considerations, and other factors, and determine on the basis of the importance that he attaches to these other factors whether, on balance, alternative feasible routes would be “prudent.”

But no such wide-ranging endeavor was intended. It is obvious that in most cases considerations of cost, directness of route, and community disruption will indicate that parkland should be used for highway construction *412whenever possible. Although it may be necessary to transfer funds from one jurisdiction to another,25 there will always be a smaller outlay required from the public purse26 when parkland is used since the public already owns the land and there will be no need to pay for right-of-way. And since people do not live or work in parks, if a highway is built on parkland no one will have to leave his home or give up his business. Such factors are common to substantially all highway construction. Thus, if Congress intended these factors to be on an equal footing with preservation of parkland there would have been no need for the statutes.

Congress clearly did not intend that cost and disruption of the community were to be ignored27 by the Secretary.28 But the very existence of the statutes29 indicates that protection of parkland was to be given para*413mount importance. The few green havens that are public parks were not to be lost unless there were truly unusual factors present in a particular case or the cost or community disruption resulting from alternative routes reached extraordinary magnitudes. If the statutes are to have any meaning, the Secretary cannot approve the destruction of parkland unless he finds that alternative routes present unique problems.

Plainly, there is “law to apply” and thus the exemption for action “committed to agency discretion” is inapplicable. But the existence of judicial review is only the start: the standard for review must also be determined. For that we must look to § 706 of the Administrative Procedure Act, 5 U. S. C. § 706 (1964 ed., Supp. V), which provides that a “reviewing court shall . . . hold unlawful and set aside agency action, findings, and conclusions found” not to meet six separate standards.30 In all cases *414agency action must be set aside if the action was “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law” or if the action failed to meet statutory, procedural, or constitutional requirements. 5 U. S. C. §§706 (2) (A), (B), (C), (D) (1964 ed., Supp. V). In certain narrow, specifically limited situations, the agency action is to be set aside if the action was not supported by “substantial evidence.” And in other equally narrow circumstances the reviewing court is to engage in a de novo review of the action and set it aside if it was “unwarranted by the facts.” 5 U. S. C. §§ 706 (2)(E), (F) (1964 ed., Supp. V).

Petitioners argue that the Secretary’s approval of the construction of 1-40 through Overton Park is subject to one or the other of these latter two standards of limited applicability. First, they contend that the “substantial evidence” standard of § 706 (2) (E) must be applied. In the alternative, they claim that § 706 (2) (F) applies and that there must be a de novo review to determine if the Secretary’s action was “unwarranted by the facts.” Neither of these standards is, however, applicable.

Review under the substantial-evidence test is authorized only when the agency action is taken pursuant to a rulemaking provision of the Administrative Procedure Act itself, 5 U. S. C. § 553 (1964 ed., Supp. V), or when the agency action is based on a public adjudicatory hearing. See 5 U. S. C. § § 556, 557 (1964 ed., Supp. V). The Secretary’s decision to allow the expenditure of federal funds to build 1-40 through Overton Park was plainly not an exercise of a rulemaking function. See 1 K. Davis, Administrative Law Treatise § 5.01 (1958). And the only hearing that is required by either the Administrative Procedure Act or the statutes regulating the dis*415tribution of federal funds for highway construction is a public hearing conducted by local officials for the purpose of informing the community about the proposed project and eliciting community views on the design and route. 23 U. S. C. § 128 (1964 ed., Supp. V). The hearing is nonadjudicatory, quasi-legislative in nature. It is not designed to produce a record that is to be the basis of agency action — the basic requirement for substantial-evidence review. See H. R. Rep. No. 1980, 79th Cong., 2d Sess.

Petitioners’ alternative argument also fails. De novo review of whether the Secretary’s decision was “unwarranted by the facts” is authorized by § 706 (2) (P) in only two circumstances. First, such de novo review is authorized when the action is adjudicatory in nature and the agency factfinding procedures are inadequate. And, there may be independent judicial factfinding when issues that were not before the agency are raised in a proceeding to enforce nonadjudicatory agency action. H. R. Rep. No. 1980, 79th Cong., 2d Sess. Neither situation exists here.

Even though there is no de novo review in this case and the Secretary’s approval of the route of 1-40 does not have ultimately to meet the substantial-evidence test, the generally applicable standards of § 706 require the reviewing court to engage in a substantial inquiry. Certainly, the Secretary’s decision is entitled to a presumption of regularity. See, e. g., Pacific States Box & Basket Co. v. White, 296 U. S. 176, 185 (1935); United States v. Chemical Foundation, 272 U. S. 1, 14—15 (1926). But that presumption is not to shield his action from a thorough, probing, in-depth review.

The court is first required to decide whether the Secretary acted within the scope of his authority. Schilling v. Rogers, 363 U. S. 666, 676-677 (1960). This determination naturally begins with a delineation of the scope of *416the Secretary’s authority and discretion. L. Jaffe, Judicial Control of Administrative Action 359 (1965). As has been shown, Congress has specified only a small range of choices that the Secretary can make. Also involved in this initial inquiry is a determination of whether on the facts the Secretary’s decision can reasonably be said to be within that range. The reviewing court must consider whether the Secretary properly construed his authority to approve the use of parkland as limited to situations where there are no feasible alternative routes or where feasible alternative routes involve uniquely difficult problems. And the reviewing court must be able to find that the Secretary could have reasonably believed that in this case there are no feasible alternatives or that alternatives do involve unique problems.

Scrutiny of the facts does not end, however, with the determination that the Secretary has acted within the scope of his statutory authority. Section 706 (2) (A) requires a finding that the actual choice made was not “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” 5 U. S. C. § 706 (2)(A) (1964 ed., Supp. V). To make this finding the court must consider whether the decision was based on a consideration of the relevant factors and whether there has been a clear error of judgment. Jaffe, supra, at 182. See McBee v. Bomar, 296 F. 2d 235, 237 (CA6 1961); In re Josephson, 218 F. 2d 174, 182 (CA1 1954); Western Addition Community Organization v. Weaver, 294 F. Supp. 433 (ND Cal. 1968). See also Wong Wing Hang v. Immigration and Naturalization Serv., 360 F. 2d 715, 719 (CA2 1966). Although this inquiry into the facts is to be searching and careful, the ultimate standard of review is a narrow one. The court is not empowered to substitute its judgment for that of the agency.

*417The final inquiry is whether the Secretary’s action followed the necessary procedural requirements. Here the only procedural error alleged is the failure of the Secretary to make formal findings and state his reason for allowing the highway to be built through the park.

Undoubtedly, review of the Secretary’s action is hampered by his failure to make such findings, but the absence of formal findings does not necessarily require that the case be remanded to the Secretary. Neither the Department of Transportation Act nor the Federal-Aid Highway Act requires such formal findings. Moreover, the Administrative Procedure Act requirements that there be formal findings in certain rulemaking and adjudicatory proceedings do not apply to the Secretary’s action here. See 5 U. S. C. §§ 553 (a)(2), 554 (a) (1964 ed., Supp. V). And, although formal findings may be required in some cases in the absence of statutory directives when the nature of the agency action is ambiguous, those situations are rare. See City of Yonkers v. United States, 320 U. S. 685 (1944); American Trucking Assns. v. United States, 344 U. S. 298, 320 (1953). Plainly, there is no ambiguity here; the Secretary has approved the construction of 1-40 through Overton Park and has approved a specific design for the project.

Petitioners contend that although there may not be a statutory requirement that the Secretary make formal findings and even though this may not be a case for the reviewing court to impose a requirement that findings be made, Department of Transportation regulations require them. This argument is based on DOT Order 5610.1,31 which requires the Secretary to make formal *418findings when he approves the use of parkland for highway construction but which was issued after the route for 1-40 was approved.32 Petitioners argue that even though the order was not in effect at the time approval was given to the Overton Park project and even though the order was not intended to have retrospective effect the order represents the law at the time of this Court’s decision and under Thorpe v. Housing Authority, 393 U. S. 268, 281-282 (1969), should be applied to this case.

The Thorpe litigation resulted from an attempt to evict a tenant from a federally funded housing project under circumstances that suggested that the eviction was prompted by the tenant’s objections to the management of the project. Despite repeated requests, the Housing Authority would not give an explanation for its action. The tenant claimed that the eviction interfered with her exercise of First Amendment rights and that the failure to state the reasons for the eviction and to afford her a hearing denied her due process. After denial of relief in the state courts, this Court granted certiorari “to consider whether [the tenant] was denied due process by the Housing Authority’s refusal to state the reasons for her eviction and to afford her a hearing at which she could contest the sufficiency of those reasons.” 393 U. S., at 272.

While the case was pending in this Court, the Department of Housing and Urban Development issued regulations requiring Housing Authority officials to inform tenants of the reasons for an eviction and to give a tenant the opportunity to reply. The case was then remanded to the state courts to determine if the HUD regulations were applicable to that case. The state court held them not to be applicable and this Court reversed on the *419ground that the general rule is “that an appellate court must apply the law in effect at the time it renders its decision.” 393 U. S., at 281.

While we do not question that DOT Order 5610.1 constitutes the law in effect at the time of our decision, we do not believe that Thorpe compels us to remand for the Secretary to make formal findings.33 Here, unlike the situation in Thorpe, there has been a change in circumstances — additional right-of-way has been cleared and the 26-acre right-of-way inside Overton Park has been purchased by the State. Moreover, there is an administrative record that allows the full, prompt review of the Secretary’s action that is sought without additional delay which would result from having a remand to the Secretary.

That administrative record is not, however, before us. The lower courts based their review on the litigation affidavits that were presented. These affidavits were merely “post hoc” rationalizations, Burlington Truck Lines v. United States, 371 U. S. 156, 168-169 (1962), which have traditionally been found to be an inadequate basis for review. Burlington Truck Lines v. United States, supra; SEC v. Chenery Corp., 318 U. S. 80, 87 (1943). And they clearly do not constitute the “whole record” compiled by the agency: the basis for review required by § 706 of the Administrative Procedure Act. See n. 30, supra.

*420Thus it is necessary to remand this case to the District Court for plenary review of the Secretary’s decision. That review is to be based on the full administrative record that was before the Secretary at the time he made his decision.34 But since the bare record may not disclose the factors that were considered or the Secretary’s construction of the evidence it may be necessary for the District Court to require some explanation in order to determine if the Secretary acted within the scope of his authority and if the Secretary’s action was justifiable under the applicable standard.

The court may require the administrative officials who participated in the decision to give testimony explaining their action. Of course, such inquiry into the mental processes of administrative decisionmakers is usually to be avoided. United States v. Morgan, 313 U. S. 409, 422 (1941). And where there are administrative findings that were made at the same time as the decision, as was the case in Morgan, there must be a strong showing of bad faith or improper behavior before such inquiry may be made. But here there are no such formal findings and it may be that the only way there can be effective judicial review is by examining the decisionmakers themselves. See Shaughnessy v. Accardi, 349 U. S. 280 (1955).

The District Court is not, however, required to make such an inquiry. It may be that the Secretary can prepare formal findings including the information required by DOT Order 5610.1 that will provide an adequate explanation for his action. Such an explanation will, to some extent, be a “post hoc rationalization” and thus must be viewed critically. If the District Court decides *421that additional explanation is necessary, that court should consider which method will prove the most expeditious so that full review may be had as soon as possible.

Reversed and remanded.

Me. Justice Douglas took no part in the consideration or decision of this case.

Separate opinion of

Mr. Justice Black,

with whom Mr. Justice Brennan joins.

I agree with the Court that the judgment of the Court of Appeals is wrong and that its action should be reversed. I do not agree that the whole matter should be remanded to the District Court. I think the case should be sent back to the Secretary of Transportation. It is apparent from the Court’s opinion today that the Secretary of Transportation completely failed to comply with the duty imposed upon him by Congress not to permit a federally financed public highway to run through a public park “unless (1) there is no feasible and prudent alternative to the use of such land, and (2) such program includes all possible planning to minimize harm to such park . . . .” 23 U. S. C. § 138 (1964 ed., Supp. V); 49 U. S. C. § 1653 (f) (1964 ed., Supp. V). That congressional command should not be taken lightly by the Secretary or by this Court. It represents a solemn determination of the highest law-making body of this Nation that the beauty and health-giving facilities of our parks are not to be taken away for public roads without hearings, factfindings, and policy determinations under the supervision of a Cabinet officer — the Secretary of Transportation. The Act of Congress in connection with other federal highway aid legislation,1 it seems to me, *422calls for hearings — hearings that a court can review, hearings that demonstrate more than mere arbitrary defiance by the Secretary. Whether the findings growing out of such hearings are labeled “formal” or “informal” appears to me to be no more than an exercise in semantics. Whatever the hearing requirements might be, the Department of Transportation failed to meet them in this case. I regret that I am compelled to conclude for myself that, except for some too-late formulations, apparently coming from the Solicitor General's office, this record contains not one word to indicate that the Secretary raised even a finger to comply with the command of Congress. It is our duty, I believe, to remand this whole matter back to the Secretary of Transportation for him to give this matter the hearing it deserves in full good-faith obedience to the Act of Congress. That Act was obviously passed to protect our public parks from forays by road builders except in the most extraordinary and imperative circumstances.2 This record does not demonstrate the existence of such circumstances. I dissent from the Court’s failure to send the case back to the Secretary, whose duty has not yet been performed.

Me. Justice Blackmun.

I fully join the Court in its opinion and in its judgment. I merely wish to state the obvious: (1) The case comes to this Court as the end product of more than a decade of endeavor to solve the interstate highway problem at Memphis. (2) The administrative decisions under attack here are not those of a single Secretary; some were made by the present Secretary’s predecessor and, before him, by the Department of Commerce’s Bureau of Public *423Roads. (3) The 1966 Act and the 1968 Act have cut across former methods and here have imposed new standards and conditions upon a situation that already was largely developed.

This undoubtedly is why the record is sketchy and less than one would expect if the project were one which had been instituted after the passage of the 1966 Act.

Justice Souter

delivered the opinion of the Court.

The question is whether a tariff classification ruling by the United States Customs Service deserves judicial deference. The Federal Circuit rejected Customs’s invocation of Chevron U S. A. Inc. v. Natural Resources Defense Council, Inc., 467 U. S. 837 (1984), in support of such a ruling, to which it gave no deference. We agree that a tariff classification has no claim to judicial deference under Chevron, there being no indication that Congress intended such a ruling to carry the force of law, but we hold that under Skidmore v. Swift & Co., 323 U. S. 134 (1944), the ruling is eligible to claim respect according to its persuasiveness.

I

A

Imports are taxed under the Harmonized Tariff Schedule of the United States (HTSUS), 19 U. S. C. § 1202. Title 19 U. S. C. § 1500(b) provides that Customs “shall, under rules *222and regulations prescribed by the Secretary [of the Treasury,] ... fix the final classification and rate of duty applicable to... merchandise” under the HTSUS. Section 1502(a) provides that

“[t]he Secretary of the Treasury shall establish and promulgate such rules and regulations not inconsistent with the law (including regulations establishing procedures for the issuance of binding rulings prior to the entry of the merchandise concerned), and may disseminate such information as may be necessary to secure a just, impartial, and uniform appraisement of imported merchandise and the classification and assessment of duties thereon at the various ports of entry.”1

See also §1624 (general delegation to Secretary to issue rules and regulations for the admission of goods).

The Secretary provides for tariff rulings before the entry of goods by regulations authorizing “ruling letters” setting tariff classifications for particular imports. 19 CFR § 177.8 (2000). A ruling letter

“represents the official position of the Customs Service with respect to the particular transaction or issue described therein and is binding on all Customs Service personnel in accordance with the provisions of this section until modified or revoked. In the absence of a change of practice or other modification or revocation which affects the principle of the ruling set forth in the ruling letter, that principle may be cited as authority in the disposition of transactions involving the same circumstances.” § 177.9(a).

*223After the transaction that gives it birth, a ruling letter is to “be applied only with respect to transactions involving articles identical to the sample submitted with the ruling request or to articles whose description is identical to the description set forth in the ruling letter.” § 177.9(b)(2). As a general matter, such a letter is “subject to modification or revocation without notice to any person, except the person to whom the letter wás addressed,” § 177.9(c), and the regulations consequently provide that “no other person should rely on the ruling letter or assume that the principles of that ruling will be applied in connection with any transaction other than the one described in the letter,” ibid. Since ruling letters respond t'o transactions of the moment, they are not subject to notice and comment before being issued, may be published but need only be made “available for public inspection,” 19 U. S. C. § 1625(a), and, at the time this action arose, could be modified without notice and comment under most circumstances, 19 CFR § 177.10(c) (2000).2 A broader notice-and-comment requirement for modification of prior rulings was added by statute in 1993, Pub. L. 103-182, § 623, 107 Stat. 2186, codified at 19 U. S. C. § 1625(c), and took effect after this case arose.3

*224Any of the 464 port-of-entry5 Customs offices may issue ruling letters, and so may the Customs Headquarters Office, in providing “[a]dvice or guidance as to the interpretation or proper application of the Customs and related laws with respect to a specific Customs transaction [which] may be requested by Customs Service field offices ... at any time, whether the transaction is prospective, current, or completed,” 19 CFR § 177.11(a) (2000). Most ruling letters contain little or no reasoning, but simply describe goods and state the appropriate category and tariff. A few letters, like the Headquarters ruling at issue here, set out a rationale in some detail.

B

Respondent, the Mead Corporation, imports “day planners,” three-ring binders with pages having room for notes of daily schedules and phone numbers and addresses, together with a calendar and suchlike. The tariff schedule on point falls under the HTSUS heading for “[registers, account books, notebooks, order books, receipt books, letter pads, memorandum pads, diaries and similar articles,” HTSUS subheading 4820.10, which comprises two subcategories. Items in the first, “[d]iaries, notebooks and address books, bound; memorandum pads, letter pads and similar articles,” were subject to a tariff of 4.0% at the time in controversy. 185 F. 3d 1304, 1305 (CA Fed. 1999) (citing subheading 4820.10.20); see also App. to Pet. for Cert. 46a. Objects in the second, covering “[o]ther” items, were free *225of duty. HTSUS subheading 4820.10.40; see also App. to Pet. for Cert. 46a.

Between 1989 and 1993, Customs repeatedly treated day planners under the “other” HTSUS subheading. In January 1993, however, Customs changed its position, and issued a Headquarters ruling letter classifying Mead’s day planners as “Diaries . . . , bound” subject to tariff under subheading 4820.10.20. That letter was short on explanation, App. to Brief in Opposition 4a-6a, but after Mead’s protest, Customs Headquarters issued a new letter, carefully reasoned but never published, reaching the same conclusion, App. to Pet. for Cert. 28a-47a. This letter considered two definitions of “diary” from the Oxford English Dictionary, the first covering a daily journal of the past day’s events, the second a book including “ ‘printed dates for daily memoranda and jottings; also . . . calendars . . . .’” Id., at 33a-34a (quoting Oxford English Dictionary 321 (Compact ed. 1982)). Customs concluded that “diary” was not confined to the first, in part because the broader definition reflects commercial usage and hence the “commercial identity of these items in the marketplace.” App. to Pet. for Cert. 34a. As for the definition of “bound,” Customs concluded that HTSUS was not referring to “bookbinding,” but to a less exact sort of fastening described in the Harmonized Commodity Description and Coding System Explanatory Notes to Heading 4820, which spoke of binding by “‘reinforcements or fittings of metal, plastics, etc.’ ” Id., at 45a.

Customs rejected Mead’s further protest of the second Headquarters ruling letter, and Mead filed suit in the Court of International Trade (CIT). The CIT granted the Government’s motion for summary judgment, adopting Customs’s reasoning without saying anything about deference. 17 F. Supp. 2d 1004 (1998).

Mead then went to the United States Court of Appeals for the Federal Circuit. While the case was pending there this Court decided United States v. Haggar Apparel Co., 526 *226U. S. 380 (1999), holding that Customs regulations receive the deference described in Chevron U S. A. Inc. v. Natural Resources Defense Council, Inc., 467 U. S. 837 (1984). The appeals court requested briefing on the impact of Haggar, and the Government argued that classification rulings, like Customs regulations, deserve Chevron deference.

The Federal Circuit, however, reversed the CIT and held that Customs classification rulings should not get Chevron deference, owing to differences from the regulations at issue in Haggar. Rulings are not preceded by notice and comment as under the Administrative Procedure Act (APA), 5 U. S. C. § 553, they “do not carry the force of law and are not, like regulations, intended to clarify the rights and obligations of importers beyond the specific case under review.” 185 F. 3d, at 1307. The appeals court thought classification rulings had a weaker Chevron claim even than Internal Revenue Service interpretive rulings, to which that court gives no deference; unlike rulings by the IRS, Customs rulings issue from many locations and need not be published. 185 F. 3d, at 1307-1308.

The Court of Appeals accordingly gave no deference at all to the ruling classifying the Mead day planners and rejected the agency’s reasoning as to both “diary” and “bound.” It thought that planners were not diaries because they had no space for “relatively extensive notations about events, observations, feelings, or thoughts” in the past. Id., at 1310. And it concluded that diaries “bound” in subheading 4810.10.20 presupposed “unbound” diaries, such that treating ring-fastened diaries as “bound” would leave the “unbound diary” an empty category. Id., at 1311.

We granted certiorari, 530 U. S. 1202 (2000), in order to consider the limits of Chevron deference owed to administrative practice in applying a statute. We hold that administrative implementation of a particular statutory provision qualifies for Chevron deference when it appears that Congress delegated authority to the agency generally to make *227rules carrying the force of law, and that the agency interpretation claiming deference was promulgated in the exercise of that authority. Delegation of such authority may be shown in a variety of ways, as by an agency’s power to engage in adjudication or notice-and-comment rulemaking, or by some other indication of a comparable congressional intent. The Customs ruling at issue here fails to qualify, although the possibility that it deserves some deference under Skidmore leads us to vacate and remand.

II

A

When Congresshas “explicitly left a gap for an agency to fill, there is an express delegation of authority to the agency to elucidate a specific provision of the statute by regulation,” Chevron, 467 U. S., at 843-844, and any ensuing regulation is binding in the courts unless procedurally defective, arbitrary or capricious in substance, or manifestly contrary to the statute.6 See id., at 844; United States v. Morton, 467 U. S. 822, 834 (1984); APA, 5 U. S. C. §§706(2)(A), (D). But whether or not they enjoy any express delegation of authority on a particular question, agencies charged with applying a statute necessarily make all sorts of interpretive choices, and while not all of those choices bind judges to follow them, they certainly may influence courts facing questions the agencies have already answered. “[T]he well-reasoned views of the agencies implementing a statute ‘constitute a body of experience and informed judgment to which courts and litigants may properly resort for guidance,’ ” Bragdon v. Abbott, 524 U. S. 624,642 (1998) (quoting Skidmore, 323 U. S., at 139-140), and “[w]e have long recognized that considerable weight should be accorded to an executive department’s *228construction of a statutory scheme it is entrusted to administer . . . Chevron, supra, at 844 (footnote omitted); see also Ford Motor Credit Co. v. Milhollin, 444 U. S. 555, 565 (1980); Zenith Radio Corp. v. United States, 437 U. S. 443, 450 (1978). The fair measure of deference to an agency administering its own statute has been understood to vary with circumstances, and courts have looked to the degree of the agency’s care,7 its consistency,8 formality,9 and relative expertness,10 and to the persuasiveness of the agency’s position, see Skidmore, supra, at 139-140. The approach has produced a spectrum of judicial responses, from great respect at one end, see, e. g., Aluminum Co. of America v. Central Lincoln Peoples’ Util. Dish, 467 U. S. 380, 389-390 (1984) (“ 'substantial deference’ ” to administrative construction), to near indifference at the other, see, e. g., Bowen v. Georgetown Univ. Hospital, 488 U. S. 204, 212-213 (1988) (interpretation advanced for the first time in a litigation brief). Justice Jackson summed things up in Skidmore v. Swift & Co.:

“The weight [accorded to an administrative] judgment in a particular case will depend upon the thoroughness evident in its consideration, the validity of its reasoning, its consistency with earlier and later pronouncements, and all those factors which give it power to persuade, if lacking power to control.” 323 U. S., at 140.

*229Since 1984, we have identified a category of interpretive choices distinguished by an additional reason for judicial deference. This Court in Chevron recognized that Congress not only engages in express delegation of specific interpretive authority, but that “[s]ometimes the legislative delegation to an agency on a particular question is implicit.” 467 U. S., at 844. Congress, that is, may not have expressly delegated authority or responsibility to implement a particular provision or fill a particular gap. Yet it can still be apparent from the agency’s generally conferred authority and other statutory circumstances that Congress would expect the agency to be able to speak with the force of law when it addresses ambiguity in the statute or fills a space in the enacted law, even one about which “Congress did not actually have an intent” as to a particular result. Id,., at 845. When circumstances implying such an expectation exist, a reviewing court has no business rejecting an agency’s exercise of its generally conferred authority to resolve a particular statutory ambiguity simply because the agency’s chosen resolution seems unwise, see id., at 845-846, but is obliged to accept the agency’s position if Congress has not previously spoken to the point at issue and the agency’s interpretation is reasonable, see id., at 842-845; cf. 5 U. S. C. §706(2) (a reviewing court shall set aside agency action, findings, and conclusions found to be “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law”).

We have recognized a very good indicator of delegation meriting Chevron treatment in express congressional authorizations to engage in the process of rulemaking or adjudication that produces regulations or rulings for which deference is claimed. See, e. g., EEOC v. Arabian American Oil Co., 499 U. S. 244, 257 (1991) (no Chevron deference to agency guideline where congressional delegation did not include the power to “ ‘promulgate rules or regulations’ ” (quoting General Elec. Co. v. Gilbert, 429 U. S. 125, 141 *230(1976))); see also Christensen v. Harris County, 529 U. S. 576, 596-597 (2000) (Breyer, J., dissenting) (where it is in doubt that Congress actually intended to delegate particular interpretive authority to an agency, Chevron is “inapplicable”). It is fair to assume generally that Congress contemplates administrative action with the effect of law when it provides for a relatively formal administrative procedure tending to foster the fairness and deliberation that should underlie a pronouncement of such force.11 Cf. Smiley v. Citibank (South Dakota), N. A., 517 U. S. 735, 741 (1996) (APA notice and comment “designed to assure due deliberation”). Thus, the overwhelming number of our cases applying Chevron deference have reviewed the fruits of notice-and-comment rulemaking or formal adjudication.12 That said, and as sig*231nificant as notice-and-comment is in pointing to Chevron authority, the want of that procedure here does not decide the case, for we have sometimes found reasons for Chevron deference even when no such administrative formality was required and none was afforded, see, e. g., NationsBank of N. C., N. A. v. Variable Annuity Life Ins. Co., 513 U. S. 251, 256-257, 263 (1995).13 The fact that the tariff classification here was not a product of such formal process does not alone, therefore, bar the application of Chevron.

There are, nonetheless, ample reasons to deny Chevron deference here. The authorization for classification rulings, and Customs’s practice in making them, present a case far removed not only from notice-and-comment process, but from any other circumstances reasonably suggesting that Congress ever thought of classification rulings as deserving the deference claimed for them here.

B

No matter which angle we choose for viewing the Customs ruling letter in this case, it fails to qualify under Chevron. On the face of the statute, to begin with, the terms of the congressional delegation give no indication that Congress meant to delegate authority to Customs to issue classifica*232tion rulings with the force of law. We are not, of course, here making any global statement about Customs’s authority, for it is true that the general rulemaking power conferred on Customs, see 19 U. S. C. § 1624, authorizes some regulation with the force of law, or “legal norms,” as we put it in Haggar, 526 U. S., at 391.14 It is true as well that Congress had classification rulings in mind when it explicitly authorized, in a parenthetical, the issuance of “regulations establishing procedures for the issuance of binding rulings prior to the entry of the merchandise concerned,” 19 U. S. C. § 1502(a).15 The reference to binding classifications does not, however, bespeak the legislative type of activity that would naturally bind more than the parties to the ruling, once the goods classified are admitted into this country. And though the statute’s direction to disseminate “information” necessary to “secure” uniformity, ibid., seems to assume that a ruling may be precedent in later transactions, precedential value alone does not add up to Chevron entitlement; interpretive rules may sometimes function as precedents, see Strauss, The Rulemaking Continuum, 41 Duke L. J. 1463, 1472-1473 (1992), and they enjoy no Chevron status as a class. In any event, any precedential claim of a classification ruling is counterbalanced by the provision for independent review of Customs classifications by the CIT, see 28 U. S. C. §§2638-2640; the scheme for CIT review includes a provision that treats classification rulings on par with the Secretary’s rulings on “valuation, rate of duty, marking, restricted mer*233chandise, entry requirements, drawbacks, vessel repairs, or similar matters,” § 1581(h); see § 2639(b). It is hard to imagine a congressional understanding more at odds with the Chevron regime.16

It is difficult, in fact, to see in the agency practice itself any indication that Customs ever set out with a lawmaking pretense in mind when it undertook to make classifications like these. Customs does not generally engage in notice- and-comment practice when issuing them, and their treatment by the agency makes it clear that a letter’s binding character as a ruling stops short of third parties; Customs has regarded a classification as conclusive only as between itself and the importer to whom it was issued, 19 CFR § 177.9(c) (2000), and even then only until Customs has given advance notice of intended change, §§ 177.9(a), (c). Other importers are in fact warned against assuming any right of detrimental reliance. § 177.9(c).

Indeed, to claim that classifications have legal force is to ignore the reality that 46 different Customs offices issue 10,000 to 15,000 of them each year, see Brief for Respondent 5; CITBA Brief 6 (citing Treasury Advisory Committee on the Commercial Operations of the United States Customs Service, Report of the COAC Subcommittee on OR&R, Exhs. 1, 3 (Jan. 26, 2000) (reprinted in App. to CITBA Brief 20a-21a)). Any suggestion that rulings intended to have the force of law are being churned out at a rate of 10,000 a year at an agency’s 46 scattered offices is simply self-refuting. Although the circumstances are less startling here, with a Headquarters letter in issue, none of the relevant statutes recognizes this category of rulings as separate or different from others; there is thus no indication that a *234more potent delegation might have been understood as going to Headquarters even when Headquarters provides developed reasoning, as it did in this instance.

Nor do the amendments to the statute made effective after this case arose disturb our conclusion. The new law requires Customs to provide notice-and-comment procedures only when modifying or revoking a prior classification ruling or modifying the treatment accorded to substantially identical transactions, 19 U. S. C. § 1625(c); and under its regulations, Customs sees itself obliged to provide notice-and-comment procedures only when “changing a practice” so as to produce a tariff increase, or in the imposition of a restriction or prohibition, or when Customs Headquarters determines that “the matter is of sufficient importance to involve the interests of domestic industry,” 19 CFR §§ 177.10(c)(1), (2) (2000). The statutory changes reveal no new congressional objective of treating classification decisions generally as rulemaking with force of law, nor do they suggest any intent to create a Chevron patchwork of classification rulings, some with force of law, some without.

In sum, classification rulings are best treated like “interpretations contained in policy statements, agency manuals, and enforcement guidelines.” Christensen, 529 U. S., at 587. They are beyond the Chevron pale.

C

To agree with the Court of Appeals that Customs ruling letters do not fall within Chevron is not, however, to place them outside the pale of any deference whatever. Chevron did nothing to eliminate Skidmore’s holding that an agency’s interpretation may merit some deference whatever its form, given the “specialized experience and broader investigations and information” available to the agency, 328 U. S., at 139, and given the value of uniformity in its administrative and judicial understandings of what a national law requires, id., at 140. See generally Metropolitan Stevedore Co. v. *235Rambo, 521 U. S. 121, 136 (1997) (reasonable agency interpretations carry “at least some added persuasive force” where Chevron is inapplicable); Reno v. Koray, 515 U. S. 50, 61 (1995) (according “some deference” to an interpretive rule that “do[es] not require notice and comment”); Martin v. Occupational Safety and Health Review Comm’n, 499 U. S. 144, 157 (1991) (“some weight” is due to informal interpretations though not “the same deference as norms that derive from the exercise of... delegated lawmaking powers”).

There is room at least to raise a Skidmore claim here, where the regulatory scheme is highly detailed, and Customs can bring the benefit of specialized experience to bear on the subtle questions in this case: whether the daily planner with room for brief daily entries falls under “diaries,” when diaries are grouped with “notebooks and address books, bound; memorandum pads, letter pads and similar articles,” HTSUS subheading 4820.10.20; and whether a planner with a ring binding should qualify as “bound,” when a binding may be typified by a book, but also may have “reinforcements or fittings of metal, plastics, etc.,” Harmonized Commodity Description and Coding System Explanatory Notes to Heading 4820, p. 687 (cited in Customs Headquarters letter, App. to Pet. for Cert. 45a. A classification ruling in this situation may therefore at least seek a respect proportional to its “power to persuade,” Skidmore, supra, at 140; see also Christensen, 529 U. S., at 587; id., at 595 (Stevens, J., dissenting); id., at 596-597 (Breyer, J., dissenting). Such a ruling may surely claim the merit of its writer’s thoroughness, logic, and expertness, its fit with prior interpretations, and any other sources of weight.

D

Underlying the position we take here, like the position expressed by Justice Scalia in dissent, is a choice about the best way to deal with an inescapable feature of the *236body of congressional legislation authorizing administrative action. That feature is the great variety of ways in which the laws invest the Government’s administrative arms with discretion, and with procedures for exercising it, in giving meaning to Acts of Congress. Implementation of a statute may occur in formal adjudication or the choice to defend against judicial challenge; it may occur in a central board or office or in dozens of enforcement agencies dotted across the country; its institutional lawmaking may be confined to the resolution of minute detail or extend to legislative rulemaking on matters intentionally left by Congress to be worked out at the agency level.

Although we all accept the position that the Judiciary should defer to at least some of this multifarious administrative action, we have to decide how to take account of the great range of its variety. If the primary objective is to simplify the judicial process of giving or withholding deference, then the diversity of statutes authorizing discretionary administrative action must be declared irrelevant or minimized. If, on the other hand, it is simply implausible that Congress intended such a broad range of statutory authority to produce only two varieties of administrative action, demanding either Chevron deference or none at all, then the breadth of the spectrum of possible agency action must be taken into account. Justice Scalia’^ first priority over the years has been to limit and simplify. The Court’s choice has been to tailor deference to variety.17 This accept-*237anee of the range of statutory variation has led the Court to recognize more than one variety of judicial deference, just as the Court has recognized a variety of indicators that Congress would expect Chevron deference.18

Our respective choices are repeated today. Justice Scalia would pose the question of deference as an either- or choice. On his view that Chevron rendered Skidmore anachronistic, when courts owe any deference it is Chevron deference that they owe, post, at 250. Whether courts do owe deference in a given case turns, for him, on whether the agency action (if reasonable) is “authoritative,” post, at 257. The character of the authoritative derives, in turn, not from breadth of delegation or the agency’s procedure in implementing it, but is defined as the “official” position of an agency, ibid., and may ultimately be a function of administrative persistence alone, ante, at 258.

The Court, on the other hand, said nothing in Chevron to eliminate Skidmore’s recognition of various justifications for deference depending on statutory circumstances and agency action; Chevron was simply a case recognizing that even without express authority to fill a specific statutory gap, circumstances pointing to implicit congressional delegation present a particularly insistent call for deference. Indeed, in holding here that Chevron left Skidmore intact and applicable where statutory circumstances indicate no intent to delegate general authority to make rules with force of law, or where such authority was not invoked, we hold nothing more than we said last Term in response to the particular *238statutory circumstances in Christensen, to which Justice Scalia then took exception, see 529 U. S., at 589, just as he does again today.

We think, in sum, that Justice Scalia’s efforts to simplify ultimately run afoul of Congress’s indications that different statutes present different reasons for considering respect for the exercise of administrative authority or deference to it. Without being at odds with congressional intent much of the time, we believe that judicial responses to administrative action must continue to differentiate between Chevron and Skidmore, and that continued recognition of Skidmore is necessary for just the reasons Justice Jackson gave when that ease was decided.19

* * *

Since the Skidmore assessment called for here ought to be made in the first instance by the Court of Appeals for the *239Federal Circuit or the CIT, we go no further than to vacate the judgment and remand the case for further proceedings consistent with this opinion.

It is so ordered.

Justice Scalia,

dissenting.

Today’s opinion makes an avulsive change in judicial review of federal administrative action. Whereas previously a reasonable agency application of an ambiguous statutory provision had to be sustained so long as it represented the agency’s authoritative interpretation, henceforth such an application can be set aside unless “it appears that Congress delegated authority to the agency generally to make rules carrying the force of law,” as by giving an agency “power to engage in adjudication or notiee-and-comment rulemaking, or . . . some other [procedure] indicating] comparable congressional intent,” and “the agency interpretation claiming deference was promulgated in the exercise of that authority.” Ante, at 226-227.1 What was previously a general presumption of authority in agencies to resolve ambiguity in the statutes they have been authorized to enforce has been changed to a presumption of no such authority, which must be overcome by affirmative legislative intent to the contrary. And whereas previously, when agency authority to resolve ambiguity did not exist the court was free to give the statute what it considered the best interpretation, henceforth the court must supposedly give the agency view some indeterminate amount of so-called Skidmore deference. Skidmore v. Swift & Co., 323 U. S. 134 (1944). We will be sorting out the consequences of the Mead doctrine, which has today replaced the Chevron doctrine, Chevron U. S. A. Inc. v. Natural Resources Defense Council, Inc., 467 U. S. 837 (1984), for years to come. I would adhere to our established jurisprudence, *240defer to the reasonable interpretation the Customs Service has given to the statute it is charged with enforcing, and reverse the judgment of the Court of Appeals.

I

Only five years ago, the Court described the Chevron doctrine as follows: “We accord deference to agencies under Chevron . .. because of a presumption that Congress, when it left ambiguity in a statute meant for implementation by an agency, understood that the ambiguity would be resolved, first and foremost, by the agency, and desired the agency (rather than the courts) to possess whatever degree of discretion the ambiguity allows,” Smiley v. Citibank (South Dakota), N. A., 517 U. S. 735, 740-741 (1996) (citing Chevron, supra, at 843-844). Today the Court collapses this doctrine, announcing instead a presumption that agency discretion does not exist unless the statute, expressly or impliedly, says so. While the Court disclaims any hard- and-fast rule for determining the existence of discretion-conferring intent, it asserts that “a very good indicator [is] express congressional authorizations to engage in the process of rulemaking or adjudication that produces regulations or rulings for which deference is claimed,” ante, at 229. Only when agencies act through “adjudieation[,] notice-and-comment rulemaking, or . . . some other [procedure] indicating] comparable congressional intent [whatever that means]” is Chevron deference applicable — because these “relatively formal administrative proeedure[s] [designed] to foster . . . fairness and deliberation” bespeak (according to the Court) congressional willingness to have the agency, rather than the courts, resolve statutory ambiguities. Ante, at 227, 230. Once it is determined that Chevron deference is not in order, the uncertainty is not at an end — and indeed is just beginning. Litigants cannot then assume that the statutory question is one for the courts to determine, accord*241ing to traditional interpretive principles and by their own judicial lights. No, the Court now resurrects, in full force, the pre-Chevron doctrine of Skidmore deference, see Skid-more, supra, whereby “[t]he fair measure of deference to an agency administering its own statute . . . varfies] with circumstances,” including “the degree of the agency’s care, its consistency, formality, and relative expertness, and . . . the persuasiveness of the agency’s position,” ante, at 228 (footnotes omitted). The Court has largely replaced Chevron, in other words, with that test most beloved by a court unwilling to be held to rules (and most feared by litigants who want to know what to expect): th’ol’ “totality of the circumstances” test.

The Court’s new doctrine is neither sound in principle nor sustainable in practice.

A

As to principle: The doctrine of Chevron — that all authoritative agency interpretations of statutes they are charged with administering deserve deference — was rooted in a legal presumption of congressional intent, important to the division of powers between the Second and Third Branches. When, Chevron said, Congress leaves an ambiguity in a statute that is to be administered by an executive agency, it is presumed that Congress meant to give the agency discretion, within the limits of reasonable interpretation, as to how the ambiguity is to be resolved. By committing enforcement of the statute to an agency rather than the courts, Congress committed its initial and primary interpretation to that branch as well.

There is some question whether Chevron was faithful to the text of the Administrative Procedure Act (APA), which it did not even bother to cite.2 But it was in accord with the *242origins of federal-court judicial review. Judicial control of federal executive officers was principally exercised through the prerogative writ of mandamus. See L. Jaffe, Judicial Control of Administrative Action 166, 176-177 (1965). That writ, generally would not issue unless the executive officer was acting plainly beyond the scope of his authority.

“The questions mooted before the Secretary and decided by him were whether the fund is a tribal fund, whether the tribe is still existing and whether the distribution of the annuities is to be confined to members of the tribe .... These are all questions of law the solution of which requires a construction of the act of 1889 and other related acts. A reading of these acts shows that they fall short of plainly requiring that any of the questions be answered in the negative and that in some aspects they give color to the affirmative answers of the Secretary. That the construction of the acts insofar as they have a bearing on the first and third questions is sufficiently uncertain to involve the exercise of judgment and discretion is rather plain....

“From what has been said it follows that the case is not one in which mandamus will lie.” t Wilbur v. United States ex rel. Kadrie, 281 U. S. 206, 221-222 (1930).

*243Statutory ambiguities, in other words, were left to reasonable resolution by the Executive.

The basis in principle for today’s new doctrine can be described as follows: The background rule is that ambiguity in legislative instructions to agencies is to be resolved not by the agencies but by the judges. Specific congressional intent to depart from this rule must be found — and while there is no single touchstone for such intent it can generally be found when Congress has authorized the agency to act through (what the Court says is) relatively formal procedures such as informal rulemaking and formal (and informal?) adjudication, and when the agency in fact employs such procedures. The Court’s background rule is contradicted by the origins of judicial review of administrative action. But in addition, the Court’s principal criterion of congressional intent to supplant its background rule seems to me quite implausible. There is no necessary connection between the formality of procedure and the power of the entity administering the procedure to resolve authoritatively questions of law. The most formal of the procedures the Court refers to — -formal adjudication — is modeled after the process used in trial courts, which of course are not generally accorded deference on questions of law. The purpose of such a procedure is to produce a closed record for determination and review of the facts — which implies nothing about the power of the agency subjected to the procedure to resolve authoritatively questions of law.

As for informal rulemaking: While formal adjudication procedures are prescribed (either by statute or by the Constitution), see 5 U. S. C..§§554, 556; Wong Yang Sung v. McGrath, 339 U. S. 33, 50 (1950), informal rulemaking is more typically authorized but not required. Agencies with such authority are free to give guidance through rulemaking, but they may proceed to administer their statute case-by-case, “making law” as they implement their program (not necessarily through formal adjudication). See NLRB v. Bell *244Aerospace Co., 416 U. S. 267, 290-295 (1974); SEC v. Chenery Corp., 332 U. S. 194, 202-203 (1947). Is it likely — or indeed even plausible — that Congress meant, when such an agency chooses rulemaking, to accord the administrators of that agency, and their successors, the flexibility of interpreting the ambiguous statute now one way, and later another; but, when such an agency chooses case-by-case administration, to eliminate all future agency discretion by having that same ambiguity resolved authoritatively (and forever) by the courts?3 Surely that makes no sense. It is also the case that certain significant categories of rules — those involving grant and benefit programs, for example, are exempt from the requirements of informal rulemaking. See 5 U. S. C. § 553(a)(2). Under the Court’s novel theory, when an agency takes advantage of that exemption its rules will be deprived of Chevron deference, i. e., authoritative effect. Was this either the plausible intent of the APA rulemaking exemption, or the plausible intent of the Congress that established the grant or benefit program?

Some decisions that are neither informal rulemaking nor formal adjudication are required to be made personally by a Cabinet Secretary, without any prescribed procedures. See, e. g., United States v. Giordano, 416 U. S. 505, 508 (1974) (involving application of 18 U. S. C. §2516 (1970 ed.), requiring wiretap applications to be authorized by “[t]he Attorney General, or any Assistant Attorney General specially designated by the Attorney General”); D. C. Federation of Civic Assns. v. Volpe, 459 F. 2d 1231, 1248-1249 (CADC 1971) (involving application of 23 U. S. C. § 138 (1970 ed.) requiring the Secretary of Transportation to determine that there is “no feasible and prudent alternative to the use of” publicly owned parkland for a federally funded highway), cert, denied, 405 U. S. 1030 (1972). Is it conceivable that decisions *245specifically committed to these high-level officers are meant to be accorded no deference, while decisions by an administrative law judge left in place without further discretionary agency review, see 5 U. S. C. § 557(b), are authoritative? This seems to me quite absurd, and not at all in accord with any plausible actual intent of Congress.

B

As for the practical effects of the new rule:

1

The principal effect will be protracted confusion. As noted above, the one test for Chevron deference that the Court enunciates is wonderfully imprecise: whether “Congress delegated authority to the agency generally to make rules carrying the force of law,.. . as by .. . adjudication[,] notice-and-comment rulemaking, or . . . some other [procedure] indicating] comparable congressional intent.” But even this description does not do justice to the utter flabbiness of the Court’s criterion, since, in order to maintain the fiction that the new test is really just the old one, applied consistently throughout our case law, the Court must make a virtually open-ended exception to its already imprecise guidance: In the present case, it tells us, the absence of notice-and-comment rulemaking (and “[who knows?] [of] some other [procedure] indieati[ng] comparable congressional intent”) is not enough to decide the question of Chevron deference, “for we have sometimes found reasons for Chevron deference even when no such administrative formality was required and none was afforded.” Ante, at 226-227, 231. The opinion then goes on to consider a grab bag of other factors — including the factor that used to be the sole criterion for Chevron deference: whether the interpretation represented the authoritative position of the agency, see ante, *246at 231-234. It is hard to know what the lower courts are to make of today’s guidance.

2

Another practical effect of today’s opinion will be an artificially induced increase in informal rulemaking. Buy stock in the GPO. Since informal rulemaking and formal adjudication are the only more-or-less safe harbors from the storm that the Court has unleashed; and since formal adjudication is not an option but must be mandated by statute or constitutional command; informal rulemaking — which the Court was once careful to make voluntary unless required by statute, see Bell Aerospace, supra, and Chenery, supra— will now become a virtual necessity. As I have described, the Court’s safe harbor requires not merely that the agency have been given rulemaking authority, but also that the agency have employed rulemaking as the means of resolving the statutory ambiguity. (It is hard to understand why that should be so. Surely the mere conferral of rulemaking authority demonstrates — if one accepts the Court’s logic — a congressional intent to allow the agency to resolve ambiguities. And given that intent, what difference does it make that the agency chooses instead to use another perfectly permissible means for that purpose?) Moreover, the majority’s approach will have a perverse effect on the rules that do emerge, given the principle (which the Court leaves untouched today) that judges must defer to reasonable agency interpretations of their own regulations. See, e. g., United States, v. Cleveland Indians Baseball Co., 532 U. S. 200, 220 (2001) (“We need not decide whether the [informal] Revenue Rulings themselves are entitled to deference[,. . . because] the Rulings simply reflect the agency’s longstanding interpretation of its own regulations”). Agencies will now have high incentive to rush out barebones, ambiguous, rules construing statutory ambiguities, which they can then in turn further clarify through informal rulings entitled to judicial respect.

*2473

Worst of all, the majority’s approach will lead to the ossification of large portions of our statutory law. Where Chevron applies, statutory ambiguities remain ambiguities subject to the agency’s ongoing clarification. They create a space, so to speak, for the exercise of continuing agency discretion. As Chevron itself held, the Environmental Protection Agency can interpret “stationary source” to mean a single smokestack, can later replace that interpretation with the “bubble concept” embracing an entire plant, and if that proves undesirable can return again to the original interpretation. 467 U. S., at 853-859, 865-866. For the indeterminately large number of statutes taken out of Chevron by today’s decision, however, ambiguity (and hence flexibility) will cease with the first judicial resolution. Skidmore deference gives the agency’s current position some vague and uncertain amount of respect, but it does not, like Chevron, leave the matter within the control of the Executive Branch for the future. Once the court has spoken, it becomes unlawful for the agency to take a contradictory position; the statute now says what the court has prescribed. See Neal v. United States, 516 U. S. 284, 295 (1996); Lechmere, Inc. v. NLRB, 502 U. S. 527, 536-537 (1992); Maislin Industries, U. S., Inc. v. Primary Steel, Inc., 497 U. S. 116, 131 (1990). It will be bad enough when this ossification occurs as a result of judicial determination (under today’s new principles) that there is no affirmative indication of congressional intent to “delegate”; but it will be positively bizarre when it occurs simply because of an agency’s failure to act by rulemaking (rather than informal adjudication) before the issue is presented to the.courts.

One might respond that such ossification would not result if the agency were simply to readopt its interpretation, after a court reviewing it under Skidmore had rejected it, by re-promulgating it through one of the Chevron-eligible procedural formats approved by the Court today. Approving this *248procedure would be a landmark abdication of judicial power. It is worlds apart from Chevron proper, where the court does not purport to give the statute a judicial interpretation— except in identifying the scope of the statutory ambiguity, as to which the court’s judgment is final and irreversible. (Under Chevron proper, when the agency’s authoritative interpretation comes within the scope of that ambiguity — and the court therefore approves it — the agency will not be “overruling” the court’s decision when it later decides that a different interpretation (still within the scope of the ambiguity) is preferable.) By contrast, under this view, the reviewing court will not be holding the agency’s authoritative interpretation within the scope of the ambiguity; but will be holding that the agency has not used the “delegation-conferring” procedures, and that the court must therefore interpret the statute on its own — but subject to reversal if and when the agency uses the proper procedures.

One is reminded of Justice Jackson’s words in Chicago & Southern Air Lines, Inc. v. Waterman S. S. Corp., 333 U. S. 103, 113 (1948):

“The court below considered that after it reviewed the Board’s order its judgment would be submitted to the President, that his power to disapprove would apply after as well as before the court acts, and hence that there would be no chance of a deadlock and no conflict of function. But if the President may completely disregard the judgment of the court, it would be only because it is one the courts were not authorized to render. Judgments within the powers vested in courts by the Judiciary Article of the Constitution may not lawfully be revised, overturned or refused faith and credit by another Department of Government.”

I know of no case, in the entire history of the federal courts, in which we have allowed a judicial interpretation of a statute to be set aside by an agency — or have allowed a *249lower court to render an interpretation of a statute subject to correction by an agency. As recently as 1996, we rejected an attempt to do 'precisely that. In Chapman v. United States, 500 U. S. 453 (1991), we had held that the weight of the blotter paper bearing the lysergic acid diethyl-amide (LSD) must be counted for purposes of determining whether the quantity crossed the 10-gram threshold of 21 U. S. C. § 841(b)(1)(A)(v) imposing a minimum sentence of 10 years. At that time the United States Sentencing Commission applied a similar approach under the Sentencing Guidelines, but had taken no position regarding the meaning of the statutory provision. The Commission later changed its Guidelines approach, and, according to the petitioner in Neal v. United States, 516 U. S. 284 (1996), made clear its view that the statute bore that meaning as well. The petitioner argued that we should defer to that new approach. We would have none of it.

“Were we, for argument’s sake, to adopt petitioner’s view that the Commission intended the commentary as an interpretation of § 841(b)(1), and that the last sentence of the commentary states the Commission’s view that the dose-based method is consistent with the term 'mixture or substance’ in the statute, he still would not prevail. The Commission’s dose-based method cannot be squared with Chapman. ... In these circumstances, we need not decide what, if any, deference is owed the Commission in order to reject its alleged contrary interpretation. Once we have determined a statute’s meaning, we adhere to our ruling under the doctrine of stare decisis, and we assess an agency’s later interpretation of the statute against that settled law.” Id., at 294-295 (citations omitted).

There is, in short, no way to avoid the ossification of federal law that today’s opinion sets in motion. What a court says is the law after according Skidmore deference will be the *250law forever, beyond the power of the agency to change even through rulemaking.

4

And finally, the majority’s approach compounds the confusion it creates by breathing new life into the anachronism of Skidmore, which sets forth a sliding scale of deference owed an agency’s interpretation of a statute that is dependent “upon the thoroughness evident in [the agency’s] consideration, the validity of its reasoning, its consistency with earlier and later pronouncements, and all those factors which give it power to persuade, if lacking power to control”; in this way, the appropriate measure of deference will be accorded the “body of experience and informed judgment” that such interpretations often embody, 323 U. S., at 140. Justice Jackson’s eloquence notwithstanding, the rule of. Skidmore deference is an empty truism and a trifling statement of the obvious: A judge should take into account the well-considered views of expert observers.

It was possible to live with the indeterminacy of Skid-more deference in earlier times. But in an era when federal statutory law administered by federal agencies is pervasive, and when the ambiguities (intended or unintended) that those statutes contain are innumerable, totality-of-the-circumstances Skidmore deference is a recipe for uncertainty, unpredictability, and endless litigation. To condemn a vast body of agency action to that regime (all except rule-making, formal (and informal?) adjudication, and whatever else might now and then be included within today’s intentionally vague formulation of affirmative congressional intent to “delegate”) is irresponsible.

II

The Court s pretense that today’s opinion is nothing more than application of our prior case law does not withstand analysis. It is, to be sure, impossible to demonstrate that any of our cases contradicts the rule of decision that the *251Court prescribes, because the Court prescribes none. More precisely, it at one and the same time (1) renders meaningless its newly announced requirement that there be an affirmative congressional intent to have ambiguities resolved by the administering agency, and (2) ensures that no prior decision can possibly be cited which contradicts that requirement, by simply announcing that all prior decisions according Chevron deference exemplify the multifarious ways in which that congressional intent can be manifested: “[A]s significant as notice-and-eomment is in pointing to Chevron authority, the want of that procedure here does not decide the case, for we have sometimes found reasons for Chevron deference even when no such administrative formality was required and none was afforded,” ante, at 230-231.4

*252The principles central to today’s opinion have no antecedent in our jurisprudence. Chevron, the case that the opinion purportedly explicates, made no mention of the “relatively formal administrative procedure[s]," ante, at 230, that the Court today finds the best indication of an affirmative intent by Congress to have ambiguities resolved by the administering agency. Which is not so remarkable, since Chevron made no mention of any need to find such an affirmative intent; it said that in the event of statutory ambiguity agency authority to clarify was to be presumed. And our cases have followed that prescription.

Six years ago, we unanimously accorded Chevron deference to an interpretation of the National Bank Act, 12 U. S. C. §24 Seventh (1988 ed. and Supp. V), contained in a letter to a private party from a Senior Deputy Comptroller of the Currency. See NationsBank of N. C, N. A. v. Variable Annuity Life Ins. Co., 513 U. S. 251, 255, 257 (1995). We did so because the letter represented (and no one contested) that it set forth the official position of the Comptroller of the Currency, see id., at 263.

Several cases decided virtually in the wake of Chevron, which the Court conveniently ignores, demonstrate that Congress could not (if it was reading our opinions) have acted in reliance on a background assumption that Chevron deference would generally be accorded only to agency interpretations arrived at through formal adjudication, notice-and-comment rulemaking, or other procedures assuring “fairness and deliberation,” ante, at 230. In FDIC v. Philadelphia Gear Corp., 476 U. S. 426, 438-439 (1986), we accorded Chevron deference to the Federal Deposit Insurance Corporation’s interpretation of the statutory term “deposit” reflected in a course of unstructured administrative actions, and gave particular weight to the agency’s “contemporaneous understanding” reflected in the response given by an FDIC official to a question asked at a meeting of FDIC and bank officials. It was clear that the position reflected *253the official position of the agency, and that was enough to command Chevron deference. In Young v. Community Nutrition Institute, 476 U. S. 974 (1986), the statutory ambiguity at issue pertained to a provision that “the Secretary [of Health and Human Services] shall promulgate regulations limiting the quantity [of any poisonous or deleterious substance added to any food] to such extent as he finds necessary for the protection of public health.” The Secretary had regularly interpreted the phrase “to such extent as he finds necessary” as conferring discretion not to issue a rule, rather than merely discretion regarding the quantity that the rule would permit. This interpretation was not, of course, reflected in any formal adjudication, and had not been the subject of any informal rulemaking — it was the Secretary’s understanding consistently applied in the course of the Department’s practice. We accorded it Chevron deference, as unquestionably we should have. And in Mead Cory. v. Tilley, 490 U. S. 714 (1989), a private suit by retirees against their former employer under the Employee Retirement Income Security Act of 1974 (ERISA), we accorded Chevron deference to the Pension Benefit Guaranty Corporation’s interpretation of § 4044(a) of ERISA, 29 U. S. C. § 1344(a) (1982 ed. and Supp. V), that was reflected only in an amicus brief to this Court and in several opinion letters issued without benefit of any prescribed procedures. See 490 U. S., at 722.

I could continue to enumerate cases according Chevron deference to agency interpretations not arrived at through formal proceedings — for example, Pension Benefit Guaranty Corporation v. LTV Corp., 496 U. S. 633, 642-643, 647-648 (1990) (according Chevron deference to the PBGC’s interpretation of the requirements for its restoring a terminated plan under §4047 of ERISA, 29 U. S. C. §1347 (1988 ed.), which interpretation was reflected in nothing more than the agency’s act of issuing a notice of restoration). Suffice it to say that many cases flatly contradict the theory of Chevron set forth in today’s opinion, and with one exception *254not a single case can be found with language that supports the theory. That exception, a very recent one, deserves extended discussion.

In Christensen v. Harris County, 529 U. S. 576 (2000), the Court said the following:

“[W]e confront an interpretation contained in an opinion letter, not one arrived at after, for example, a formal adjudication or notice-and-comment rulemaking. Interpretations such as those in opinion letters — like interpretations contained in policy statements, agency manuals, and enforcement guidelines, all of which lack the force of law — do not warrant Chevron-style deference.” Id., at 587.

This statement was dictum, unnecessary to the Court’s holding. Since the Court went on to find that the Secretary of Labor’s position “ma[de] little sense” given the text and structure of the statute, id., at 585-586, Chevron deference could not have been accorded no matter what the conditions for its application. See 529 U. S., at 591 (Scalia, J., concurring in part and concurring in judgment). It was, moreover, dictum unsupported by the precedent that the Court cited.

The Christensen majority followed its above-quoted dictum with a string citation of three cases, none of which sustains its point. In Reno v. Koray, 515 U. S. 50 (1995), we had no occasion to consider what level of deference was owed the Bureau of Prisons’ interpretation of 18 U. S. C. § 3585(b) set forth in an internal agency guideline, because our opinion made clear that we would have independently arrived at the same interpretation on our own, see 515 U. S., at 57-60. And although part of one sentence in Koray might be read to suggest that the Bureau’s “Program Statement]” should be accorded a measure of deference less than that mandated by Chevron, this aside is ultimately inconclusive, *255since the sentence ends by observing that the statement was “a ‘permissible construction of the statute’ ” under Chevron. 515 U. S., at 61 (quoting Chevron, 467 U. S., at 843). In the second case cited, EEOC v. Arabian American Oil Co., 499 U. S. 244 (1991), it was again unnecessary to our holding whether the agency’s interpretation of the statute warranted Chevron deference, since the “longstanding . . . ‘canon of [statutory] construction’ ” disfavoring extraterritoriality, 499 U. S., at 248, would have required the same result even if Chevron applied. See 499 U. S., at 260 (Scalia, J., concurring in part and concurring in judgment). While the opinion did purport to accord the Equal Employment Opportunity Commission’s informally promulgated interpretation only Skidmore deference, it did so because the Court thought itself bound by its pre-Chevron, EEOC-specific decision in General Elec. Co. v. Gilbert, 429 U. S. 125 (1976), which noted that “‘Congress, in enacting Title VII, did not’” intend to give the EEOC substantive authority to resolve statutory ambiguities, Arabian American Oil, supra, at 257 (quoting Gilbert, supra, at 141). Lastly, in Martin v. Occupational Safety and Health Review Comm’n, 499 U. S. 144 (1991), the question of the level of deference owed the Secretary of Labor’s interpretation of the Occupational Safety and Health Act of 1970, 84 Stat. 1590, as amended, 29 U. S. C. § 651 et seq., was neither presented by the case nor considered in our opinion. The only question before the Court was which of two competing interpretations of 29 CFR §1910.1029 (1990) — the Secretary’s or the Occupational Safety and Health Review Commission’s — should have been deferred to by the court below. See 499 U. S., at 150. The dicta the Christensen Court cited, 529 U. S., at 587 (citing 499 U. S., at 157), opined on the measure of deference owed the Secretary’s interpretation, not of the statute, but of his own regulations, see generally Manning, Constitutional Structure *256and Judicial Deference to Agency Interpretations of Agency Rules, 96 Colum. L. Rev. 612 (1996).

To make matters worse, the arguments marshaled by Christensen in support of its dictum — its observation that “interpretations contained in policy statements, agency manuals, and enforcement guidelines, all . . . lack the force of law,” and its citation of 1 K. Davis & R. Pierce, Administrative Law Treatise § 3.5 (3d ed. 1994), 529 U. S., at 587 — are not only unpersuasive but bear scant resemblance to the reasoning of today’s opinion. Davis and Pierce, and Professor Robert Anthony upon whom they rely, see Anthony, Which Agency Interpretations Should Bind Citizens and the Courts?, 7 Yale J. on Reg. 1 (1990), do indeed set forth the argument I have criticized above, that congressional authorization of informal rulemaking or formal (and perhaps even informal) adjudication somehow bespeaks a congressional intent to “delegate” power to resolve statutory ambiguities. But their analysis does not permit the broad add-ons that the Court’s opinion contains — “some other [procedure] indicating] comparable congressional intent,” ante, at 227, and “we have sometimes found reasons for Chevron deference even when no such administrative formality was required and none was afforded,” ante, at 231.

III

To decide the present case, I would adhere to the original formulation of Chevron. “ ‘The power of an administrative agency to administer a congressionally created ... program necessarily requires the formulation of policy and the making of rules to fill any gap left, implicitly or explicitly, by Congress,’ ” 467 U. S., at 843 (quoting Morton v. Ruiz, 415 U. S. 199, 231 (1974)). We accordingly presume — and our precedents have made clear to Congress that we presume— that, absent some clear textual indication to the contrary, “Congress, when it left ambiguity in a statute meant for im*257plementation by an agency, understood that the ambiguity would be resolved, first and foremost, by the agency, and desired the agency (rather than the courts) to possess whatever degree of discretion the ambiguity allows,” Smiley, 517 U. S., at 740-741 (citing Chevron, supra, at 843-844). Chevron sets forth an across-the-board presumption, which operates as a background rule of law against which Congress legislates: Ambiguity means Congress intended agency discretion. Any resolution of the ambiguity by the administering agency that is authoritative — that represents the official position of the agency — must be accepted by the courts if it is reasonable.

Nothing in the statute at issue here displays an intent to modify the background presumption on which Chevron deference is based. The Court points, ante, at 233, n. 16, to 28 U. S. C. § 2640(a), which provides that, in reviewing the ruling by the Customs Service, the Court of International Trade (CIT) “shall make its determinations upon the basis of the record made before the court.” But records are made to determine the facts, not the law. All this provision means is that new evidence may be introduced at the CIT stage; it says nothing about whether the CIT must respect the Customs Service’s authoritative interpretation of the law. More significant than § 2640(a), insofar as the CIT’s obligation to defer to the Customs Service’s legal interpretations is concerned, is § 2639(a)(1), which requires the CIT to accord a “presumfption of] correctness]” to the Customs Service’s decision. Another provision cited by the Court, ante, at 233, n. 16, is §2638, which provides that the CIT, “by rule, may consider any new ground in support” of the challenge to the Customs Service’s ruling. Once again, it is impossible to see how this has any connection to the degree of deference the CIT must accord the Customs Service’s interpretation of its statute. Such “new ground[s]” may be intervening or newly discovered facts, or some intervening *258law or regulation that might render the Customs Service’s ruling unsound.5

There is no doubt that the Customs Service’s interpretation represents the authoritative view of the agency. Although the actual ruling letter was signed by only the Director of the Commercial Rulings Branch of Customs Headquarters’ Office of Regulations and Rulings, see Pet. for Cert. 47a, the Solicitor General of the United States has filed a brief, cosigned by the General Counsel of the Department of the Treasury, that represents the position set forth in the ruling letter to be the official position of the Customs Service. Cf. Christensen, 529 U. S., at 591 (Scalia, J., concurring in part and concurring in judgment). No one contends that it is merely a “post hoc rationalization]” or an “agency litigating positio[n] wholly unsupported by regulations, rulings, or administrative practice,” Bowen v. Georgetown Univ. Hospital, 488 U. S. 204, 212 (1988).6

*259There is also no doubt that the Customs Service’s interpretation is a reasonable one, whether or not judges would consider it the best. I will not belabor this point, since the Court evidently agrees: An interpretation that was unreasonable would not merit the remand that the Court decrees for consideration of Skidmore deference.

IV

Finally, and least importantly, even were I to accept the Court’s revised version of Chevron as a correct statement *260of the law, I would still accord deference to the tariff classification ruling at issue in this case. For the case is indistinguishable, in that regard, from NationsBank of N. C., N. A. v. Variable Annuity Life Ins. Co., 513 U. S. 251 (1995), which the Court acknowledges as an instance in which Chevron deference is warranted notwithstanding the absence of formal adjudication, notice-and-comment rulemaking, or comparable “administrative formality,” ante, at 231. Here, as in NationsBank, there is a tradition of great deference to the opinions of the agency head, ante, at 231, n. 13. Just two Terms ago, we observed:

“As early as 1809, Chief Justice Marshall noted in a customs case that ‘[i]f the question had been doubtful, the court would have respected the uniform construction which it is understood has been given by the treasury department of the United States upon similar questions.’ United States v. Vowell, 5 Cranch 368, 372. See also P. Reed, The Role of Federal Courts in U. S. Customs & International Trade Law 289 (1997) (‘Consistent with the Chevron methodology, and as has long been the rule in customs cases, customs regulations are sustained if they represent reasonable interpretations of the statute’); cf. Zenith Radio Corp. v. United States, 437 U. S. 443, 450 (1978) (deferring to the Treasury Department’s ‘longstanding and consistent administrative interpretation’ of the countervailing duty provision of the Tariff Act.” United States v. Haggar Apparel Co., 526 U. S. 380, 393 (1999).

And here, as in NationsBank, the agency interpretation in question is officially that of the agency head. Consequently, even on the Court’s own terms, the Customs ruling at issue in this case should be given Chevron deference.

*261For the reasons stated, I respectfully dissent from the Court’s judgment. I would uphold the Customs Service’s construction of Subheading 4820.10.20 of thé Harmonized Tariff Schedule of the United States, 19 U. S. C. § 1202, and would reverse the contrary decision of the Court of Appeals. I dissent even more vigorously from the reasoning that produces the Court’s judgment, and that makes today’s decision one of the most significant opinions ever rendered by the Court dealing with the judicial review of administrative action. Its consequences will be enormous, and almost uniformly bad.

Justice SCALIA delivered the opinion of the Court.

*293We consider whether an agency's interpretation of a statutory ambiguity that concerns the scope of its regulatory authority (that is, its jurisdiction) is entitled to deference under Chevron U.S.A. Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984).

I

Wireless telecommunications networks require towers and antennas; proposed sites for those towers and antennas must be approved by local zoning authorities. In the Telecommunications Act of 1996, Congress "impose[d] specific limitations on the traditional authority of state and local governments to regulate the location, construction, and modification of such facilities," Rancho Palos Verdes v. Abrams, 544 U.S. 113, 115, 125 S.Ct. 1453, 161 L.Ed.2d 316 (2005), and incorporated those limitations into the Communications Act of 1934, see 110 Stat. 56, 151. Section 201(b) of that Act empowers the Federal Communications Commission to "prescribe such rules and regulations as may be necessary in the public interest to carry out [its] provisions." Ch. 296, 52 Stat. 588, codified at 47 U.S.C. § 201(b). Of course, that rulemaking authority extends to the subsequently added portions of the Act. See AT & T Corp. v. Iowa Utilities Bd., 525 U.S. 366, 377-378, 119 S.Ct. 721, 142 L.Ed.2d 835 (1999).

*294The Act imposes five substantive limitations, which are codified in 47 U.S.C. § 332(c)(7)(B) ; only one of them, § 332(c)(7)(B)(ii), is at issue here. That provision requires state or local governments to act on wireless siting applications "within a reasonable period of time after the request is duly filed." Two other features of § 332(c)(7) are relevant. First, subparagraph (A), known as the "saving clause," provides that nothing in the Act, except those limitations provided in § 332(c)(7)(B), "shall limit or affect the authority of a State or local government" over siting decisions. Second, *1867§ 332(c)(7)(B)(v) authorizes a person who believes a state or local government's wireless-siting decision to be inconsistent with any of the limitations in § 332(c)(7)(B) to "commence an action in any court of competent jurisdiction."

In theory, § 332(c)(7)(B)(ii) requires state and local zoning authorities to take prompt action on siting applications for wireless facilities. But in practice, wireless providers often faced long delays. In July 2008, CTIA-The Wireless Association,1 which represents wireless service providers, petitioned the FCC to clarify the meaning of § 332(c)(7)(B)(ii)'s requirement that zoning authorities act on siting requests "within a reasonable period of time." In November 2009, the Commission, relying on its broad statutory authority to implement the provisions of the Communications Act, issued a declaratory ruling responding to CTIA's petition. In re Petition for Declaratory Ruling, 24 FCC Rcd. 13994, 14001. The Commission found that the "record evidence demonstrates that unreasonable delays in the personal wireless service facility siting process have obstructed the provision of wireless services" and that such delays "impede the promotion of advanced services and competition *295that Congress deemed critical in the Telecommunications Act of 1996." Id., at 14006, 14008. A " reasonable period of time" under § 332(c)(7)(B)(ii), the Commission determined, is presumptively (but rebuttably) 90 days to process a collocation application (that is, an application to place a new antenna on an existing tower) and 150 days to process all other applications. Id., at 14005.

Some state and local governments opposed adoption of the Declaratory Ruling on the ground that the Commission lacked "authority to interpret ambiguous provisions of Section 332(c)(7)." Id., at 14000. Specifically, they argued that the saving clause, § 332(c)(7)(A), and the judicial review provision, § 337(c)(7)(B)(v), together display a congressional intent to withhold from the Commission authority to interpret the limitations in § 332(c)(7)(B). Asserting that ground of objection, the cities of Arlington and San Antonio, Texas, petitioned for review of the Declaratory Ruling in the Court of Appeals for the Fifth Circuit.

Relying on Circuit precedent, the Court of Appeals held that the Chevron framework applied to the threshold question whether the FCC possessed statutory authority to adopt the 90- and 150-day timeframes. 668 F.3d 229, 248 (C.A.5 2012) (citing Texas v. United States, 497 F.3d 491, 501 (C.A.5 2007) ). Applying Chevron, the Court of Appeals found " § 332(c)(7)(A)'s effect on the FCC's authority to administer § 332(c)(7)(B)'s limitations ambiguous," 668 F.3d, at 250, and held that "the FCC's interpretation of its statutory authority" was a permissible construction of the statute. Id., at 254. On the merits, the court upheld the presumptive 90- and 150-day deadlines as a "permissible construction of § 332(c)(7)(B)(ii) and (v)... entitled to Chevron deference." Id., at 256.

We granted certiorari, 568 U.S. ----, 133 S.Ct. 524, 184 L.Ed.2d 252 (2012), limited to the first question presented: "Whether ... a court should apply Chevron to ... an agency's determination of its *1868own jurisdiction." Pet. for Cert. in No. 11-1545, p. i.

II

A

*296As this case turns on the scope of the doctrine enshrined in Chevron, we begin with a description of that case's now-canonical formulation. "When a court reviews an agency's construction of the statute which it administers, it is confronted with two questions." 467 U.S., at 842, 104 S.Ct. 2778. First, applying the ordinary tools of statutory construction, the court must determine "whether Congress has directly spoken to the precise question at issue. If the intent of Congress is clear, that is the end of the matter; for the court, as well as the agency, must give effect to the unambiguously expressed intent of Congress." Id., at 842-843, 104 S.Ct. 2778. But "if the statute is silent or ambiguous with respect to the specific issue, the question for the court is whether the agency's answer is based on a permissible construction of the statute." Id., at 843, 104 S.Ct. 2778.

Chevron is rooted in a background presumption of congressional intent: namely, "that Congress, when it left ambiguity in a statute" administered by an agency, "understood that the ambiguity would be resolved, first and foremost, by the agency, and desired the agency (rather than the courts) to possess whatever degree of discretion the ambiguity allows." Smiley v. Citibank (South Dakota), N. A., 517 U.S. 735, 740-741, 116 S.Ct. 1730, 135 L.Ed.2d 25 (1996). Chevron thus provides a stable background rule against which Congress can legislate: Statutory ambiguities will be resolved, within the bounds of reasonable interpretation, not by the courts but by the administering agency. See Iowa Utilities Bd., 525 U.S., at 397, 119 S.Ct. 721. Congress knows to speak in plain terms when it wishes to circumscribe, and in capacious terms when it wishes to enlarge, agency discretion.

B

The question here is whether a court must defer under Chevron to an agency's interpretation of a statutory ambiguity that concerns the scope of the agency's statutory authority *297(that is, its jurisdiction). The argument against deference rests on the premise that there exist two distinct classes of agency interpretations: Some interpretations-the big, important ones, presumably-define the agency's "jurisdiction." Others-humdrum, run-of-the-mill stuff-are simply applications of jurisdiction the agency plainly has. That premise is false, because the distinction between "jurisdictional" and "nonjurisdictional" interpretations is a mirage. No matter how it is framed, the question a court faces when confronted with an agency's interpretation of a statute it administers is always, simply, whether the agency has stayed within the bounds of its statutory authority .

The misconception that there are, for Chevron purposes, separate "jurisdictional" questions on which no deference is due derives, perhaps, from a reflexive extension to agencies of the very real division between the jurisdictional and nonjurisdictional that is applicable to courts. In the judicial context, there is a meaningful line: Whether the court decided correctly is a question that has different consequences from the question whether it had the power to decide at all . Congress has the power (within limits) to tell the courts what classes of cases they may decide, see Trainmen v. Toledo, P. & W.R. Co., 321 U.S. 50, 63-64, 64 S.Ct. 413, 88 L.Ed. 534 (1944) ;

*1869Lauf v. E.G. Shinner & Co., 303 U.S. 323, 330, 58 S.Ct. 578, 82 L.Ed. 872 (1938), but not to prescribe or superintend how they decide those cases, see Plaut v. Spendthrift Farm, Inc., 514 U.S. 211, 218-219, 115 S.Ct. 1447, 131 L.Ed.2d 328 (1995). A court's power to decide a case is independent of whether its decision is correct, which is why even an erroneous judgment is entitled to res judicata effect. Put differently, a jurisdictionally proper but substantively incorrect judicial decision is not ultra vires.

That is not so for agencies charged with administering congressional statutes. Both their power to act and how they are to act is authoritatively prescribed by Congress, so that when they act improperly, no less than when they act beyond their jurisdiction, what they do is ultra vires. Because the *298question-whether framed as an incorrect application of agency authority or an assertion of authority not conferred-is always whether the agency has gone beyond what Congress has permitted it to do, there is no principled basis for carving out some arbitrary subset of such claims as " jurisdictional."

An example will illustrate just how illusory the proposed line between "jurisdictional" and "nonjurisdictional" agency interpretations is. Imagine the following validly-enacted statute:

COMMON CARRIER ACT

SECTION 1. The Agency shall have jurisdiction to prohibit any common carrier from imposing an unreasonable condition upon access to its facilities.

There is no question that this provision-including the terms "common carrier" and "unreasonable condition"-defines the Agency's jurisdiction. Surely, the argument goes, a court must determine de novo the scope of that jurisdiction.

Consider, however, this alternative formulation of the statute:

COMMON CARRIER ACT

SECTION 1. No common carrier shall impose an unreasonable condition upon access to its facilities.

SECTION 2. The Agency may prescribe rules and regulations necessary in the public interest to effectuate Section 1 of this Act.

Now imagine that the Agency, invoking its Section 2 authority, promulgates this Rule: "(1) The term 'common carrier' in Section 1 includes Internet Service Providers. (2) The term 'unreasonable condition' in Section 1 includes unreasonably high prices. (3) A monthly fee greater than $25 is an unreasonable condition on access to Internet service." By this Rule, the Agency has claimed for itself jurisdiction that is doubly questionable: Does its authority extend to Internet Service *299Providers? And does it extend to setting prices? Yet Section 2 makes clear that Congress, in petitioners' words, "conferred interpretive power on the agency" with respect to Section 1. Brief for Petitioners in No. 1545, p. 14. Even under petitioners' theory, then, a court should defer to the Agency's interpretation of the terms "common carrier" and "unreasonable condition"-that is to say, its assertion that its "jurisdiction" extends to regulating Internet Service Providers and setting prices.

In the first case, by contrast, petitioners' theory would accord the agency no deference. The trouble with this is that in both cases, the underlying question is exactly the same : Does the statute give the agency authority to regulate Internet Service Providers and cap prices, or not?2

*1870The reality, laid bare, is that there is no difference, insofar as the validity of agency action is concerned, between an agency's exceeding the scope of its authority (its "jurisdiction") and its exceeding authorized application of authority that it unquestionably has. "To exceed authorized application is to exceed authority. Virtually any administrative action can be characterized as either the one or the other, depending on how generally one wishes to describe the 'authority.' " Mississippi Power & Light Co. v. Mississippi ex rel. Moore, 487 U.S. 354, 381, 108 S.Ct. 2428, 101 L.Ed.2d 322 (1988) (SCALIA, J., concurring in judgment); see also Monaghan, Marbury and the Administrative State, 83 Colum. L. Rev. 1, 29 (1983) ("Administrative application of law is administrative formulation of law whenever it involves elaboration of the statutory norm.").

*300This point is nicely illustrated by our decision in National Cable & Telecommunications Assn., Inc. v. Gulf Power Co., 534 U.S. 327, 122 S.Ct. 782, 151 L.Ed.2d 794 (2002). That case considered whether the FCC's "jurisdiction" to regulate the rents utility-pole owners charge for "pole attachments" (defined as attachments by a cable television system or provider of telecommunications service) extended to attachments that provided both cable television and high-speed Internet access (attachments for so-called "commingled services"). Id. , at 331-336, 122 S.Ct. 782. We held, sensibly, that Chevron applied. 534 U.S., at 333, 339, 122 S.Ct. 782. Whether framed as going to the scope of the FCC's delegated authority or the FCC's application of its delegated authority, the underlying question was the same: Did the FCC exceed the bounds of its statutory authority to regulate rents for "pole attachments" when it sought to regulate rents for pole attachments providing commingled services?

The label is an empty distraction because every new application of a broad statutory term can be reframed as a questionable extension of the agency's jurisdiction. One of the briefs in support of petitioners explains, helpfully, that "[j]urisdictional questions concern the who, what, where, and when of regulatory power: which subject matters may an agency regulate and under what conditions." Brief for IMLA Respondents 18-19. But an agency's application of its authority pursuant to statutory text answers the same questions. Who is an "outside salesman"? What is a "pole attachment"? Where do the "waters of the United States" end? When must a Medicare provider challenge a reimbursement determination in order to be entitled to an administrative appeal? These can all be reframed as questions about the scope of agencies' regulatory jurisdiction-and they are all questions to which the Chevron framework applies. See Christopher v. SmithKline Beecham Corp., 567 U.S. ----, ----, ----, 132 S.Ct. 2156, 2162, 2165, 183 L.Ed.2d 153 (2012) ; National Cable & Telecommunications Assn., supra, at 331, 333, 122 S.Ct. 782; United States v. Riverside Bayview Homes, Inc., 474 U.S. 121, 123, 131, 106 S.Ct. 455, 88 L.Ed.2d 419 (1985) ; Sebelius v. Auburn *301Regional Medical Center, 568 U.S. ----, ----, ----, 133 S.Ct. 817, 821, 826-827, 184 L.Ed.2d 627 (2013).

In sum, judges should not waste their time in the mental acrobatics needed to decide whether an agency's interpretation of a statutory provision is "jurisdictional" or "nonjurisdictional." Once those labels *1871are sheared away, it becomes clear that the question in every case is, simply, whether the statutory text forecloses the agency's assertion of authority, or not. See H. Edwards & L. Elliott, Federal Standards of Review 146 (2007) ("In practice, it does not appear to matter whether delegated authority is viewed as a threshold inquiry."). The federal judge as haruspex, sifting the entrails of vast statutory schemes to divine whether a particular agency interpretation qualifies as "jurisdictional," is not engaged in reasoned decisionmaking.

C

Fortunately, then, we have consistently held "that Chevron applies to cases in which an agency adopts a construction of a jurisdictional provision of a statute it administers." 1 R. Pierce, Administrative Law Treatise § 3.5, p. 187 (2010). One of our opinions explicitly says that no "exception exists to the normal [deferential] standard of review" for " 'jurisdictional or legal question[s] concerning the coverage' " of an Act. NLRB v. City Disposal Systems, Inc., 465 U.S. 822, 830, n. 7, 104 S.Ct. 1505, 79 L.Ed.2d 839 (1984). A prime example of deferential review for questions of jurisdiction is Commodity Futures Trading Comm'n v. Schor, 478 U.S. 833, 106 S.Ct. 3245, 92 L.Ed.2d 675 (1986). That case involved a CFTC interpretation of 7 U.S.C. § 18(c), which provides that before the Commission takes action on a complaint, the complainant must file a bond to cover "any reparation award that may be issued by the Commission against the complainant on any counterclaim by respondent." (Emphasis added.) The CFTC, pursuant to its broad rulemaking authority, see § 12a(5), interpreted that oblique reference to counterclaims as granting it "the power to take jurisdiction *302over" not just federal-law counterclaims, but state-law counterclaims as well. Schor, supra, at 844, 106 S.Ct. 3245. We not only deferred under Chevron to the Commission's "eminently reasonable ... interpretation of the statute it is entrusted to administer," but also chided the Court of Appeals for declining to afford deference because of the putatively " 'statutory interpretation-jurisdictional' nature of the question at issue." 478 U.S., at 844-845, 106 S.Ct. 3245.

Similar examples abound. We have afforded Chevron deference to the Commerce Department's determination that its authority to seek antidumping duties extended to uranium imported under contracts for enrichment services, United States v. Eurodif S. A., 555 U.S. 305, 316, 129 S.Ct. 878, 172 L.Ed.2d 679 (2009) ; to the Interstate Commerce Commission's view that courts, not the Commission, possessed "initial jurisdiction with respect to the award of reparations" for unreasonable shipping charges, Reiter v. Cooper, 507 U.S. 258, 269, 113 S.Ct. 1213, 122 L.Ed.2d 604 (1993) (internal quotation marks and ellipsis omitted); and to the Army Corps of Engineers' assertion that its permitting authority over discharges into "waters of the United States" extended to "freshwater wetlands" adjacent to covered waters, Riverside Bayview Homes, supra, at 123-124, 131, 106 S.Ct. 455. We have even deferred to the FCC's assertion that its broad regulatory authority extends to pre-empting conflicting state rules. City of New York v. FCC, 486 U.S. 57, 64, 108 S.Ct. 1637, 100 L.Ed.2d 48 (1988) ; Capital Cities Cable, Inc. v. Crisp, 467 U.S. 691, 700, 104 S.Ct. 2694, 81 L.Ed.2d 580 (1984).3

*1872*303Our cases hold that Chevron applies equally to statutes designed to curtail the scope of agency discretion. For instance, in Chemical Mfrs. Assn. v. Natural Resources Defense Council, Inc., 470 U.S. 116, 123, 105 S.Ct. 1102, 84 L.Ed.2d 90 (1985), we considered a statute prohibiting the Environmental Protection Agency from "modify[ing] any requirement of this section as it applies to any specific pollutant which is on the toxic pollutant list." The EPA construed the statute as not precluding it from granting variances with respect to certain toxic pollutants. Finding no " clear congressional intent to forbid EPA's sensible variance mechanism," id., at 134, 105 S.Ct. 1102, we deferred to the EPA's construction of this express limitation on its own regulatory authority, id., at 125, 105 S.Ct. 1102 (citing Chevron, 467 U.S. 837, 104 S.Ct. 2778, 81 L.Ed.2d 694); see also, e.g., Japan Whaling Assn. v. American Cetacean Soc., 478 U.S. 221, 226, 232-234, 106 S.Ct. 2860, 92 L.Ed.2d 166 (1986).

The U.S. Reports are shot through with applications of Chevron to agencies' constructions of the scope of their own jurisdiction. And we have applied Chevron where concerns about agency self-aggrandizement are at their apogee: in cases where an agency's expansive construction of the extent of its own power would have wrought a fundamental change in the regulatory scheme. In FDA v. Brown & Williamson Tobacco Corp., 529 U.S. 120, 120 S.Ct. 1291, 146 L.Ed.2d 121 (2000), the threshold question was the "appropriate framework for analyzing" the FDA's *304assertion of "jurisdiction to regulate tobacco products," id., at 126, 132, 120 S.Ct. 1291 -a question of vast "economic and political magnitude," id., at 133, 120 S.Ct. 1291."Because this case involves an administrative agency's construction of a statute that it administers," we held, Chevron applied. 529 U.S., at 132, 120 S.Ct. 1291. Similarly, in MCI Telecommunications Corp. v. American Telephone & Telegraph Co., 512 U.S. 218, 224, 229, 231, 114 S.Ct. 2223, 129 L.Ed.2d 182 (1994), we applied the Chevron framework to the FCC's assertion that the statutory phrase "modify any requirement" gave it authority to eliminate rate-filing requirements, "the essential characteristic of a rate-regulated industry," for long-distance telephone carriers.

The false dichotomy between "jurisdictional" and "nonjurisdictional" agency interpretations may be no more than a bogeyman, but it is dangerous all the same. Like the Hound of the Baskervilles, it is *1873conjured by those with greater quarry in sight: Make no mistake-the ultimate target here is Chevron itself. Savvy challengers of agency action would play the "jurisdictional" card in every case. See, e.g., Cellco Partnership v. FCC, 700 F.3d 534, 541 (C.A.D.C.2012). Some judges would be deceived by the specious, but scary-sounding, "jurisdictional"-"nonjurisdictional" line; others tempted by the prospect of making public policy by prescribing the meaning of ambiguous statutory commands. The effect would be to transfer any number of interpretive decisions-archetypal Chevron questions, about how best to construe an ambiguous term in light of competing policy interests-from the agencies that administer the statutes to federal courts.4 We have cautioned that "judges ought to *305refrain from substituting their own interstitial lawmaking" for that of an agency. Ford Motor Credit Co. v. Milhollin, 444 U.S. 555, 568, 100 S.Ct. 790, 63 L.Ed.2d 22 (1980). That is precisely what Chevron prevents.

III

A

One group of respondents contends that Chevron deference is inappropriate here because the FCC has "assert[ed] jurisdiction over matters of traditional state and local concern." Brief for IMLA Respondents 35. But this case has nothing to do with federalism. Section 332(c)(7)(B)(ii) explicitly supplants state authority by requiring zoning authorities to render a decision "within a reasonable period of time," and the meaning of that phrase is indisputably a question of federal law. We rejected a similar faux-federalism argument in the Iowa Utilities Board case, in terms that apply equally here: "This is, at bottom, a debate not about whether the States will be allowed to do their own thing, but about whether it will be the FCC or the federal courts that draw the lines to which they must hew." 525 U.S., at 379, n. 6, 119 S.Ct. 721. These lines will be drawn either by unelected federal bureaucrats, or by unelected (and even less politically accountable) federal judges. "[I]t is hard to spark a passionate 'States' rights' debate over that detail." Ibid.

B

A few words in response to the dissent. The question on which we granted certiorari was whether "a court should apply Chevron to review an agency's determination of its own jurisdiction." Pet. for Cert. i.5 Perhaps sensing the *306incoherence of the "jurisdictional-nonjurisdictional" line, the dissent does not even attempt to defend it, see post, at 1864, but proposes a much *1874broader scope for de novo judicial review: Jurisdictional or not, and even where a rule is at issue and the statute contains a broad grant of rulemaking authority, the dissent would have a court search provision-by-provision to determine "whether [that] delegation covers the 'specific provision' and 'particular question' before the court." Post, at 1882 - 1883.

The dissent is correct that United States v. Mead Corp., 533 U.S. 218, 121 S.Ct. 2164, 150 L.Ed.2d 292 (2001), requires that, for Chevron deference to apply, the agency must have received congressional authority to determine the particular matter at issue in the particular manner adopted. No one disputes that. But Mead denied Chevron deference to action, by an agency with rulemaking authority, that was not rulemaking. What the dissent needs, and fails to produce, is a single case in which a general conferral of rulemaking or adjudicative authority has been held insufficient to support Chevron deference for an exercise of that authority within the agency's substantive field. There is no such case, and what the dissent proposes is a massive revision of our Chevron jurisprudence.

Where we differ from the dissent is in its apparent rejection of the theorem that the whole includes all of its parts-its view that a general conferral of rulemaking authority does not validate rules for all the matters the agency is charged with administering. Rather, the dissent proposes that even when general rulemaking authority is clear, every agency rule must be subjected to a de novo judicial determination of whether the particular issue was committed to agency discretion. It offers no standards at all to guide this open-ended hunt for congressional intent (that is to say, for evidence of congressional intent more specific than the conferral of general rulemaking authority). It would simply punt that question back to the Court of Appeals, presumably *307for application of some sort of totality-of-the-circumstances test-which is really, of course, not a test at all but an invitation to make an ad hoc judgment regarding congressional intent. Thirteen Courts of Appeals applying a totality-of-the-circumstances test would render the binding effect of agency rules unpredictable and destroy the whole stabilizing purpose of Chevron. The excessive agency power that the dissent fears would be replaced by chaos. There is no need to wade into these murky waters. It suffices to decide this case that the preconditions to deference under Chevron are satisfied because Congress has unambiguously vested the FCC with general authority to administer the Communications Act through rulemaking and adjudication, and the agency interpretation at issue was promulgated in the exercise of that authority.

* * *

Those who assert that applying Chevron to "jurisdictional" interpretations "leaves the fox in charge of the henhouse" overlook the reality that a separate category of "jurisdictional" interpretations does not exist. The fox-in-the-henhouse syndrome is to be avoided not by establishing an arbitrary and undefinable category of agency decisionmaking that is accorded no deference, but by taking seriously, and applying rigorously, in all cases, statutory limits on agencies' authority. Where Congress has established a clear line, the agency cannot go beyond it; and where Congress has established an ambiguous line, the agency can go no further than the ambiguity will fairly allow. But in rigorously applying the latter rule, a court need not pause to puzzle over whether the interpretive question presented is "jurisdictional." If "the agency's answer is based on a permissible construction of the statute,"

*1875that is the end of the matter. Chevron, 467 U.S., at 842, 104 S.Ct. 2778.

The judgment of the Court of Appeals is affirmed.

It is so ordered.

Justice BREYER, concurring in part and concurring in the judgment.

*308I agree with the Court that normally "the question a court faces when confronted with an agency's interpretation of a statute it administers" is, "simply, whether the agency has stayed within the bounds of its statutory authority." Ante, at 1879 - 1880. In this context, "the distinction between 'jurisdictional' and 'non-jurisdictional' interpretations is a mirage." Ante, at 1879 - 1880.

Deciding just what those statutory bounds are, however, is not always an easy matter, and the Court's case law abounds with discussion of the subject. A reviewing judge, for example, will have to decide independently whether Congress delegated authority to the agency to provide interpretations of, or to enact rules pursuant to, the statute at issue-interpretations or rules that carry with them "the force of law." United States v. Mead Corp., 533 U.S. 218, 229, 121 S.Ct. 2164, 150 L.Ed.2d 292 (2001). If so, the reviewing court must give special leeway or "deference" to the agency's interpretation. See id., at 227-228, 121 S.Ct. 2164.

We have added that, if "[e]mploying traditional tools of statutory construction," INS v. Cardoza-Fonseca, 480 U.S. 421, 446, 107 S.Ct. 1207, 94 L.Ed.2d 434 (1987), the court determines that Congress has spoken clearly on the disputed question, then "that is the end of the matter," Chevron U.S.A. Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 842, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984). The agency is due no deference, for Congress has left no gap for the agency to fill. Id., at 842-844, 104 S.Ct. 2778. If, on the other hand, Congress has not spoken clearly, if, for example it has written ambiguously, then that ambiguity is a sign-but not always a conclusive sign-that Congress intends a reviewing court to pay particular attention to (i.e., to give a degree of deference to) the agency's interpretation. See Gonzales v. Oregon, 546 U.S. 243, 258-269, 126 S.Ct. 904, 163 L.Ed.2d 748 (2006) ; Mead, supra, at 229, 121 S.Ct. 2164

I say that the existence of statutory ambiguity is sometimes not enough to warrant the conclusion that Congress *309has left a deference-warranting gap for the agency to fill because our cases make clear that other, sometimes context-specific, factors will on occasion prove relevant. (And, given the vast number of government statutes, regulatory programs, and underlying circumstances, that variety is hardly surprising.) In Mead , for example, we looked to several factors other than simple ambiguity to help determine whether Congress left a statutory gap, thus delegating to the agency the authority to fill that gap with an interpretation that would carry "the force of law." 533 U.S., at 229-231, 121 S.Ct. 2164. Elsewhere, we have assessed

"the interstitial nature of the legal question, the related expertise of the Agency, the importance of the question to administration of the statute, the complexity of that administration, and the careful consideration the Agency has given the question over a long period of time." Barnhart v. Walton, 535 U.S. 212, 222, 122 S.Ct. 1265, 152 L.Ed.2d 330 (2002).

The subject matter of the relevant provision-for instance, its distance from the agency's ordinary statutory duties or its falling within the scope of another agency's authority-has also proved relevant. See Gonzales, supra, at 265-266, 126 S.Ct. 904.

*1876See also Gellhorn & Verkuil, Controlling Chevron -Based Delegations, 20 Cardozo L.Rev. 989, 1007-1010 (1999).

Moreover, the statute's text, its context, the structure of the statutory scheme, and canons of textual construction are relevant in determining whether the statute is ambiguous and can be equally helpful in determining whether such ambiguity comes accompanied with agency authority to fill a gap with an interpretation that carries the force of law. See Household Credit Services, Inc. v. Pfennig, 541 U.S. 232, 239-242, 124 S.Ct. 1741, 158 L.Ed.2d 450 (2004) ; Zuni Public School Dist. No. 89 v. Department of Education, 550 U.S. 81, 98-99, 127 S.Ct. 1534, 167 L.Ed.2d 449 (2007) ; FDA v. Brown & Williamson Tobacco Corp., 529 U.S. 120, 133, 120 S.Ct. 1291, 146 L.Ed.2d 121 (2000) ; Dole v. Steelworkers, 494 U.S. 26, 36, 110 S.Ct. 929, 108 L.Ed.2d 23 (1990). Statutory purposes, including those revealed in part by legislative *310and regulatory history, can be similarly relevant. See Brown & Williamson Tobacco Corp.,supra, at 143-147, 120 S.Ct. 1291; Pension Benefit Guaranty Corporation v. LTV Corp., 496 U.S. 633, 649, 110 S.Ct. 2668, 110 L.Ed.2d 579 (1990) ; Global Crossing Telecommunications, Inc. v. Metrophones Telecommunications, Inc., 550 U.S. 45, 48-49, 127 S.Ct. 1513, 167 L.Ed.2d 422 (2007). See also AT & T Corp. v. Iowa Utilities Bd., 525 U.S. 366, 412-413, 119 S.Ct. 721, 142 L.Ed.2d 835 (1999) (BREYER, J., concurring in part and dissenting in part).

Although seemingly complex in abstract description, in practice this framework has proved a workable way to approximate how Congress would likely have meant to allocate interpretive law-determining authority between reviewing court and agency. The question whether Congress has delegated to an agency the authority to provide an interpretation that carries the force of law is for the judge to answer independently. The judge, considering "traditional tools of statutory construction," Cardoza-Fonseca, supra, at 446, 107 S.Ct. 1207, will ask whether Congress has spoken unambiguously. If so, the text controls. If not, the judge will ask whether Congress would have intended the agency to resolve the resulting ambiguity. If so, deference is warranted. See Mead, supra, at 229, 121 S.Ct. 2164 Even if not, however, sometimes an agency interpretation, in light of the agency's special expertise, will still have the "power to persuade, if lacking power to control," Skidmore v. Swift & Co., 323 U.S. 134, 140, 65 S.Ct. 161, 89 L.Ed. 124 (1944).

The case before us offers an example. The relevant statutory provision requires state or local governments to act on wireless siting applications "within a reasonable period of time after" a wireless service provider files such a request. 47 U.S.C. § 332(c)(7)(B)(ii). The Federal Communications Commission (FCC) argued that this provision granted it a degree of leeway in determining the amount of time that is reasonable. Many factors favor the agency's view: (1) the language of the Telecommunications Act grants the FCC broad authority (including rulemaking authority) to administer the Act; (2) the words are open-ended-i.e. "ambiguous";

*311(3) the provision concerns an interstitial administrative matter, in respect to which the agency's expertise could have an important role to play; and (4) the matter, in context, is complex, likely making the agency's expertise useful in helping to answer the "reasonableness" question that the statute poses. See § 151 (creating the FCC); § 201(b) (providing rulemaking authority); National Cable & Telecommunications Assn. v. Brand X Internet Services, 545 U.S. 967, 980-981, 125 S.Ct. 2688, 162 L.Ed.2d 820 (2005) (acknowledging *1877the FCC's authority to administer the Act).

On the other side of the coin, petitioners point to two statutory provisions which, they believe, require a different conclusion-namely, that the FCC lacked authority altogether to interpret § 332(c)(7)(B)(ii). First, a nearby saving clause says: "Except as provided in this paragraph, nothing in this chapter shall limit or affect the authority of a State or local government or instrumentality thereof over decisions regarding the placement, construction, and modification of personal wireless service facilities." § 332(c)(7)(A). Second, a judicial review provision, says: "Any person adversely affected by any final action or failure to act by a State or local government or any instrumentality thereof that is inconsistent with this subparagraph may, within 30 days after such action or failure to act, commence an action in any court of competent jurisdiction." § 332(c)(7)(B)(v).

In my view, however, these two provisions cannot provide good reason for reaching the conclusion advocated by petitioners. The first provision begins with an exception, stating that it does not apply to (among other things) the "reasonableness" provision here at issue. The second simply sets forth a procedure for judicial review, a review that applies to most government actions. Both are consistent with a statutory scheme that gives States, localities, the FCC, and reviewing courts each some role to play in the location of wireless service facilities. And neither "expressly describ[es] an exception" to the FCC's plenary authority to interpret *312the Act. American Hospital Assn. v. NLRB, 499 U.S. 606, 613, 111 S.Ct. 1539, 113 L.Ed.2d 675 (1991).

For these reasons, I would reject petitioners' argument and conclude that § 332(c)(7)(B)(ii) -the "reasonableness" statute-leaves a gap for the FCC to fill. I would hold that the FCC's lawful efforts to do so carry "the force of law." Mead, 533 U.S., at 229, 121 S.Ct. 2164. The Court of Appeals ultimately reached the same conclusion (though for somewhat different reasons), and the majority affirms the lower court. I consequently join the majority's judgment and such portions of its opinion as are consistent with what I have written here.

Chief Justice ROBERTS, with whom Justice KENNEDY and Justice ALITO join, dissenting.

My disagreement with the Court is fundamental. It is also easily expressed: A court should not defer to an agency until the court decides, on its own, that the agency is entitled to deference. Courts defer to an agency's interpretation of law when and because Congress has conferred on the agency interpretive authority over the question at issue. An agency cannot exercise interpretive authority until it has it; the question whether an agency enjoys that authority must be decided by a court, without deference to the agency.

I

One of the principal authors of the Constitution famously wrote that the "accumulation of all powers, legislative, executive, and judiciary, in the same hands, ... may justly be pronounced the very definition of tyranny." The Federalist No. 47, p. 324 (J. Cooke ed. 1961) (J. Madison). Although modern administrative agencies fit most comfortably within the Executive Branch, as a practical matter they exercise legislative power, by promulgating regulations with the force of law; executive power, by policing compliance with those regulations; and judicial power, by adjudicating enforcement actions and imposing sanctions on those found to *313have *1878violated their rules. The accumulation of these powers in the same hands is not an occasional or isolated exception to the constitutional plan; it is a central feature of modern American government.

The administrative state "wields vast power and touches almost every aspect of daily life." Free Enterprise Fund v. Public Company Accounting Oversight Bd., 561 U.S. ----, ----, 130 S.Ct. 3138, 3156, 177 L.Ed.2d 706 (2010). The Framers could hardly have envisioned today's "vast and varied federal bureaucracy" and the authority administrative agencies now hold over our economic, social, and political activities. Ibid. "[T]he administrative state with its reams of regulations would leave them rubbing their eyes." Alden v. Maine, 527 U.S. 706, 807, 119 S.Ct. 2240, 144 L.Ed.2d 636 (1999) (Souter, J., dissenting), quoted in Federal Maritime Comm'n v. South Carolina Ports Authority, 535 U.S. 743, 755, 122 S.Ct. 1864, 152 L.Ed.2d 962 (2002). And the federal bureaucracy continues to grow; in the last 15 years, Congress has launched more than 50 new agencies. Compare Office of the Federal Register, United States Government Manual 1997/1998, with Office of the Federal Register, United States Government Manual 2012. And more are on the way. See, e.g., Congressional Research Service, C. Copeland, New Entities Created Pursuant to the Patient Protection and Affordable Care Act 1 (2010) (The PPACA "creates, requires others to create, or authorizes dozens of new entities to implement the legislation").

Although the Constitution empowers the President to keep federal officers accountable, administrative agencies enjoy in practice a significant degree of independence. As scholars have noted, "no President (or his executive office staff) could, and presumably none would wish to, supervise so broad a swath of regulatory activity." Kagan, Presidential Administration, 114 Harv. L.Rev. 2245, 2250 (2001) ; see also S. Breyer, Making Our Democracy Work 110 (2010) ( "the president may not have the time or willingness to review [agency] decisions"). President Truman colorfully described his power over the administrative state by complaining, "I

*314thought I was the president, but when it comes to these bureaucrats, I can't do a damn thing." See R. Nathan, The Administrative Presidency 2 (1986). President Kennedy once told a constituent, "I agree with you, but I don't know if the government will." See id., at 1. The collection of agencies housed outside the traditional executive departments, including the Federal Communications Commission, is routinely described as the "headless fourth branch of government," reflecting not only the scope of their authority but their practical independence. See, e.g., Administrative Conference of United States, D. Lewis & J. Selin, Sourcebook of United States Executive Agencies 11 (2012).

As for judicial oversight, agencies enjoy broad power to construe statutory provisions over which they have been given interpretive authority. In Chevron U.S.A. Inc. v. Natural Resources Defense Council, Inc., we established a test for reviewing "an agency's construction of the statute which it administers." 467 U.S. 837, 842, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984). If Congress has "directly spoken to the precise question at issue," we said, "that is the end of the matter." Ibid. A contrary agency interpretation must give way. But if Congress has not expressed a specific intent, a court is bound to defer to any "permissible construction of the statute," even if that is not "the reading the court would have reached if the question initially had arisen in a judicial proceeding." Id., at 843, and n. 11, 104 S.Ct. 2778.

*1879When it applies, Chevron is a powerful weapon in an agency's regulatory arsenal. Congressional delegations to agencies are often ambiguous-expressing "a mood rather than a message." Friendly, The Federal Administrative Agencies: The Need for Better Definition of Standards, 75 Harv. L.Rev. 1263, 1311 (1962). By design or default, Congress often fails to speak to "the precise question" before an agency. In the absence of such an answer, an agency's interpretation has the full force and effect of law, unless it "exceeds the bounds of the permissible." Barnhart v. Walton, 535 U.S. 212, 218, 122 S.Ct. 1265, 152 L.Ed.2d 330 (2002).

*315It would be a bit much to describe the result as "the very definition of tyranny," but the danger posed by the growing power of the administrative state cannot be dismissed. See, e.g., Talk America, Inc. v. Michigan Bell Telephone Co., 564 U.S. ----, ----, 131 S.Ct. 2254, 2266, 180 L.Ed.2d 96 (2011) (SCALIA, J., concurring) (noting that the FCC "has repeatedly been rebuked in its attempts to expand the statute beyond its text, and has repeatedly sought new means to the same ends"); Sackett v. EPA, 566 U.S. ----, ---- - ----, 132 S.Ct. 1367, 1374, 182 L.Ed.2d 367 (2012) (rejecting agency argument that would "enable the strong-arming of regulated parties into 'voluntary compliance' without the opportunity for judicial review").

What the Court says in footnote 4 of its opinion is good, and true (except of course for the "dissent overstates" part). Ante, at 1873, n. 4. The Framers did divide governmental power in the manner the Court describes, for the purpose of safeguarding liberty. And yet ... the citizen confronting thousands of pages of regulations-promulgated by an agency directed by Congress to regulate, say, "in the public interest"-can perhaps be excused for thinking that it is the agency really doing the legislating. And with hundreds of federal agencies poking into every nook and cranny of daily life, that citizen might also understandably question whether Presidential oversight-a critical part of the Constitutional plan-is always an effective safeguard against agency overreaching.

It is against this background that we consider whether the authority of administrative agencies should be augmented even further, to include not only broad power to give definitive answers to questions left to them by Congress, but also the same power to decide when Congress has given them that power.

Before proceeding to answer that question, however, it is necessary to sort through some confusion over what this litigation is about. The source of the confusion is a familiar culprit: the concept of "jurisdiction," which we have repeatedly described as a word with " 'many, too many, meanings.' "

*316Union Pacific R. Co. v. Locomotive Engineers, 558 U.S. 67, 81, 130 S.Ct. 584, 175 L.Ed.2d 428 (2009).

The Court states that the question "is whether a court must defer under Chevron to an agency's interpretation of a statutory ambiguity that concerns the scope of the agency's statutory authority (that is, its jurisdiction)." Ante, at 1868. That is fine-until the parenthetical. The parties, amici, and court below too often use the term "jurisdiction" imprecisely, which leads the Court to misunderstand the argument it must confront. That argument is not that "there exist two distinct classes of agency interpretations," some "big, important ones" that "define the agency's 'jurisdiction,' " and other "humdrum, run-of-the-mill" ones that "are simply applications of jurisdiction the agency plainly has." Ibid. The argument is instead that a court should not defer to an agency on *1880whether Congress has granted the agency interpretive authority over the statutory ambiguity at issue.

You can call that "jurisdiction" if you'd like, as petitioners do in the question presented. But given that the term is ambiguous, more is required to understand its use in that question than simply "having read it." Ante, at 1873, n. 5. It is important to keep in mind that the term, in the present context, has the more precise meaning noted above, encompassing congressionally delegated authority to issue interpretations with the force and effect of law. See 668 F.3d 229, 248 (C.A.5 2012) (case below) ("The issue in the instant case is whether the FCC possessed statutory authority to administer § 332(c)(7)(B)(ii) and (v) by adopting the 90- and 150-day time frames"). And that has nothing do with whether the statutory provisions at issue are "big" or "small."

II

"It is emphatically the province and duty of the judicial department to say what the law is." Marbury v. Madison, 1 Cranch 137, 177, 2 L.Ed. 60 (1803). The rise of the modern administrative state has not changed that duty. Indeed, the Administrative *317Procedure Act, governing judicial review of most agency action, instructs reviewing courts to decide "all relevant questions of law." 5 U.S.C. § 706.

We do not ignore that command when we afford an agency's statutory interpretation Chevron deference; we respect it. We give binding deference to permissible agency interpretations of statutory ambiguities because Congress has delegated to the agency the authority to interpret those ambiguities "with the force of law." United States v. Mead Corp., 533 U.S. 218, 229, 121 S.Ct. 2164, 150 L.Ed.2d 292 (2001) ; see also Monaghan, Marbury and the Administrative State, 83 Colum. L.Rev. 1, 27-28 (1983) ("the court is not abdicating its constitutional duty to 'say what the law is' by deferring to agency interpretations of law: it is simply applying the law as 'made' by the authorized law-making entity").

But before a court may grant such deference, it must on its own decide whether Congress-the branch vested with lawmaking authority under the Constitution-has in fact delegated to the agency lawmaking power over the ambiguity at issue. See ante, at 1876 (BREYER, J., concurring in part and concurring in judgment) ("The question whether Congress has delegated to an agency the authority to provide an interpretation that carries the force of law is for the judge to answer independently."). Agencies are creatures of Congress; "an agency literally has no power to act ... unless and until Congress confers power upon it." Louisiana Pub. Serv. Comm'n v. FCC, 476 U.S. 355, 374, 106 S.Ct. 1890, 90 L.Ed.2d 369 (1986). Whether Congress has conferred such power is the "relevant question[ ] of law" that must be answered before affording Chevron deference. 5 U.S.C. § 706.

III

A

Our precedents confirm this conclusion-beginning with Chevron itself. In Chevron, the EPA promulgated a regulation interpreting the term *318"stationary sources" in the Clean Air Act. 467 U.S., at 840, 104 S.Ct. 2778 (quoting 42 U.S.C. § 7502(b)(6)(1982 ed.) ). An environmental group petitioned for review of the rule, challenging it as an impermissible interpretation of the Act. 467 U.S., at 841, 859, 104 S.Ct. 2778. Finding the statutory text "not dispositive" and the legislative history "silent on the precise issue," we upheld the rule. Id., at 862, 866, 104 S.Ct. 2778. *1881In our view, the challenge to the agency's interpretation "center[ed] on the wisdom of the agency's policy, rather than whether it is a reasonable choice within a gap left open by Congress." Id., at 866, 104 S.Ct. 2778. Judges, we said, "are not experts in the field, and are not part of either political branch of the Government." Id., at 865, 104 S.Ct. 2778. Thus, because Congress had not answered the specific question at issue, judges had no business providing their own resolution on the basis of their "personal policy preferences." Ibid. Instead, the "agency to which Congress ha[d] delegated policymaking responsibilities" was the appropriate political actor to resolve the competing interests at stake, "within the limits of that delegation." Ibid.

Chevron 's rule of deference was based on-and limited by-this congressional delegation. And the Court did not ask simply whether Congress had delegated to the EPA the authority to administer the Clean Air Act generally. We asked whether Congress had "delegat[ed] authority to the agency to elucidate a specific provision of the statute by regulation." Id., at 843-844, 104 S.Ct. 2778 (emphasis added); see id., at 844, 104 S.Ct. 2778 (discussing "the legislative delegation to an agency on a particular question " (emphasis added)). We deferred to the EPA's interpretation of "stationary sources" based on our conclusion that the agency had been "charged with responsibility for administering the provision. " Id., at 865, 104 S.Ct. 2778 (emphasis added).

B

We have never faltered in our understanding of this straightforward principle, that whether a particular agency *319interpretation warrants Chevron deference turns on the court's determination whether Congress has delegated to the agency the authority to interpret the statutory ambiguity at issue.

We made the point perhaps most clearly in Adams Fruit Co. v. Barrett, 494 U.S. 638, 110 S.Ct. 1384, 108 L.Ed.2d 585 (1990). In that case, the Department of Labor contended the Court should defer to its interpretation of the scope of the private right of action provided by the Migrant and Seasonal Agriculture Worker Protection Act (AWPA), 29 U.S.C. § 1854, against employers who intentionally violated the Act's motor vehicle safety provisions. We refused to do so. Although "as an initial matter" we rejected the idea that Congress left a "statutory 'gap' " for the agency to fill, we reasoned that if the "AWPA's language establishing a private right of action is ambiguous," the Secretary of Labor's interpretation of its scope did not warrant Chevron deference. 494 U.S., at 649, 110 S.Ct. 1384.

In language directly applicable to the question before us, we explained that "[a] precondition to deference under Chevron is a congressional delegation of administrative authority." Ibid. Although "Congress clearly envisioned, indeed expressly mandated, a role for the Department of Labor in administering the statute by requiring the Secretary to promulgate standards implementing AWPA's motor vehicle provisions, " we found "[n]o such delegation regarding AWPA's enforcement provisions ." Id., at 650, 110 S.Ct. 1384 (emphasis added). It would therefore be "inappropriate," we said, "to consult executive interpretations" of the enforcement provisions to resolve ambiguities "surrounding the scope of AWPA's judicially enforceable remedy." Ibid. Without questioning the principle that agency determinations "within the scope of delegated authority are entitled to deference," we explained that "it is fundamental 'that an agency *1882may not bootstrap itself into an area in which it has no jurisdiction.' " Ibid. (quoting Federal Maritime Comm'n v. Seatrain Lines, Inc., 411 U.S. 726, 745, 93 S.Ct. 1773, 36 L.Ed.2d 620 (1973) ). *320Our subsequent cases follow the same approach. In United States v. Mead Corp., supra, for example, Chevron deference turned on whether Congress had delegated to the agency authority to interpret the statutory ambiguity by a particular means. The Customs Service had issued a "classification ruling," interpreting the term "diaries" in a tariff schedule to include "day planners" of the type Mead imported, and on that basis subjected the planners to a four-percent tariff. Mead protested the imposition of the tariff, the Customs Service claimed Chevron deference for its interpretation, and the controversy made its way to our Court. Id., at 224-226, 121 S.Ct. 2164.

In Mead, we again made clear that the "category of interpretative choices" to which Chevron deference applies is defined by congressional intent. Id., at 229, 121 S.Ct. 2164.Chevron deference, we said, rests on a recognition that Congress has delegated to an agency the interpretive authority to implement "a particular provision" or answer " 'a particular question.' " Ibid. (quoting Chevron, 467 U.S., at 844, 104 S.Ct. 2778). An agency's interpretation of "a particular statutory provision" thus qualifies for Chevron deference only "when it appears that Congress delegated authority to the agency generally to make rules carrying the force of law, and that the agency interpretation claiming deference was promulgated in the exercise of that authority." 533 U.S., at 226-227, 121 S.Ct. 2164.

The Court did not defer to the agency's views but instead determined that Congress had not delegated interpretive authority to the Customs Service to definitively construe the tariff schedule through classification rulings. Neither the statutory authorization for the classification rulings, nor the Customs Service's practice in issuing such rulings, "reasonably suggest[ed] that Congress ever thought of [such] classification rulings as deserving the deference claimed for them." Id., at 231, 121 S.Ct. 2164. And in the absence of such a delegation, we concluded the interpretations adopted in those rulings were "beyond the Chevron pale." Id., at 234, 121 S.Ct. 2164.

*321Gonzales v. Oregon, 546 U.S. 243, 126 S.Ct. 904, 163 L.Ed.2d 748 (2006), is in the same line of precedent. In that case, as here, deference turned on whether a congressional delegation of interpretive authority reached a particular statutory ambiguity. The Attorney General claimed Chevron deference for his interpretation of the phrase " legitimate medical purpose" in the Controlled Substances Act (CSA) to exclude the prescribing and dispensing of controlled substances for the purpose of assisting suicide. Id., at 254, 258, 126 S.Ct. 904. No one disputed that "legitimate medical purpose" was "ambiguous in the relevant sense." Id., at 258, 126 S.Ct. 904. Nor did any Justice dispute that the Attorney General had been granted the power in the CSA to promulgate rules with the force of law. Ibid. ; see id., at 281, 126 S.Ct. 904 (SCALIA, J., dissenting). Nevertheless, the Court explained, "Chevron deference ... is not accorded merely because the statute is ambiguous and an administrative official is involved." Id., at 258, 126 S.Ct. 904. The regulation advancing the interpretation, we said, "must be promulgated pursuant to authority Congress has delegated to the official." Ibid. (citing Mead,supra, at 226-227, 121 S.Ct. 2164).

In the CSA, Congress delegated to the Attorney General the authority to promulgate *1883regulations "relating to the registration and control of the manufacture, distribution, and dispensing of controlled substances," 21 U.S.C. § 821, or "for the efficient execution of his functions under [the CSA]," § 871(b). After considering the text, structure, and purpose of the Act, the Court concluded on its own that interpreting "legitimate medical purpose" fell under neither delegation. Gonzales, 546 U.S., at 258-269, 126 S.Ct. 904. Because the regulation "was not promulgated pursuant to the Attorney General's authority, its interpretation of 'legitimate medical purpose' d[id] not receive Chevron deference." Id., at 268, 126 S.Ct. 904.

Adams Fruit, Mead, and Gonzales thus confirm that Chevron deference is based on, and finds legitimacy as, a congressional delegation of interpretive authority. An agency interpretation warrants such deference only if Congress *322has delegated authority to definitively interpret a particular ambiguity in a particular manner. Whether Congress has done so must be determined by the court on its own before Chevron can apply. See H. Edwards, L. Elliott, & M. Levy, Federal Courts Standards of Review 168 (2d ed. 2013) ("a court decides de novo whether an agency has acted within the bounds of congressionally delegated authority" (citing Mead, supra, at 226-227, 121 S.Ct. 2164, and Gonzales, supra, at 258, 126 S.Ct. 904) ); Sales & Adler, The Rest is Silence: Chevron Deference, Agency Jurisdiction, and Statutory Silences, 2009 U. Ill. L.Rev. 1497, 1564 (2009) ("if delegation really is antecedent to deference, as Mead insists, it cannot be that courts should defer to an agency's views on whether a delegation has taken place").

In other words, we do not defer to an agency's interpretation of an ambiguous provision unless Congress wants us to, and whether Congress wants us to is a question that courts, not agencies, must decide. Simply put, that question is "beyond the Chevron pale." Mead,supra, at 234, 121 S.Ct. 2164

IV

Despite these precedents, the FCC argues that a court need only locate an agency and a grant of general rulemaking authority over a statute. Chevron deference then applies, it contends, to the agency's interpretation of any ambiguity in the Act, including ambiguity in a provision said to carve out specific provisions from the agency's general rulemaking authority. If Congress intends to exempt part of the statute from the agency's interpretive authority, the FCC says, Congress "can ordinarily be expected to state that intent explicitly." Brief for Federal Respondents 30 (citing American Hospital Assn. v. NLRB, 499 U.S. 606, 111 S.Ct. 1539, 113 L.Ed.2d 675 (1991) ).

If a congressional delegation of interpretive authority is to support Chevron deference, however, that delegation must extend to the specific statutory ambiguity at issue. The appropriate question is whether the delegation covers the "specific *323provision" and "particular question" before the court. Chevron, 467 U.S., at 844, 104 S.Ct. 2778. A congressional grant of authority over some portion of a statute does not necessarily mean that Congress granted the agency interpretive authority over all its provisions. See Adams Fruit, 494 U.S., at 650, 110 S.Ct. 1384.

An example that might highlight the point concerns statutes that parcel out authority to multiple agencies, which "may be the norm, rather than an exception." Gersen, Overlapping and Underlapping Jurisdiction in Administrative Law, 2006 S.Ct. Rev. 201, 208; see, e.g., Gonzales, 546 U.S., at 250-251, 126 S.Ct. 904 (describing shared authority over the CSA

*1884between the Attorney General and the Secretary of Health and Human Services); Sutton v. United Air Lines, Inc., 527 U.S. 471, 478, 119 S.Ct. 2139, 144 L.Ed.2d 450 (1999) (authority to issue regulations implementing the Americans with Disabilities Act "is split primarily among three Government agencies"). The Dodd-Frank Wall Street Reform and Consumer Protection Act, for example, authorizes rulemaking by at least eight different agencies. See Congressional Research Service, C. Copeland, Rulemaking Requirements and Authorities in the Dodd-Frank Wall Street Reform and Consumer Protection Act 7 (2010). When presented with an agency's interpretation of such a statute, a court cannot simply ask whether the statute is one that the agency administers; the question is whether authority over the particular ambiguity at issue has been delegated to the particular agency.

By the same logic, even when Congress provides interpretive authority to a single agency, a court must decide if the ambiguity the agency has purported to interpret with the force of law is one to which the congressional delegation extends. A general delegation to the agency to administer the statute will often suffice to satisfy the court that Congress has delegated interpretive authority over the ambiguity at issue. But if Congress has exempted particular provisions from that authority, that exemption must be respected, and *324the determination whether Congress has done so is for the courts alone.

The FCC's argument that Congress "can ordinarily be expected to state that intent explicitly," Brief for Federal Respondents 30 (citing American Hospital,supra ), goes to the merits of that determination, not to whether a court should decide the question de novo or defer to the agency. Indeed, that is how the Court in American Hospital considered it. It was in the process of "employing the traditional tools of statutory construction" that the Court said it would have expected Congress to speak more clearly if it had intended to exclude an entire subject area-employee units for collecting bargaining-from the NLRB's general rulemaking authority. Id., at 613, 614, 111 S.Ct. 1539. The Court concluded, after considering the language, structure, policy, and legislative history of the Act on its own-without deferring to the agency-that the meaning of the statute was "clear and contrary to the meaning advanced by petitioner." Id., at 609-614, 111 S.Ct. 1539. To be sure, the Court also noted that "[e]ven if we could find any ambiguity in [the provision] after employing the traditional tools of statutory construction, we would still defer to Board's reasonable interpretation." Id., at 614, 111 S.Ct. 1539 (emphasis added). But that single sentence of dictum cannot carry the day for the FCC here.

V

As the preceding analysis makes clear, I do not understand petitioners to ask the Court-nor do I think it necessary-to draw a "specious, but scary-sounding" line between "big, important" interpretations on the one hand and "humdrum, run-of-the-mill" ones on the other. Ante, at 1868, 1879. Drawing such a line may well be difficult. Distinguishing between whether an agency's interpretation of an ambiguous term is reasonable and whether that term is for the agency to interpret is not nearly so difficult. It certainly did not confuse the FCC in this proceeding. Compare In re Petition for *325Declaratory Ruling, 24 FCC Rcd. 13994, 14000-14003 (2009) (addressing the latter question), with id., at 14003-14015 (addressing the former). Nor did it confound the Fifth Circuit. Compare 668 F.3d, at 247-254 (deciding "whether the FCC possessed statutory authority to administer § 332(c)(7)(B)(ii)"), *1885with id., at 254-260 (considering "whether the 90- and 150-day time frames themselves also pass muster under Chevron "). More importantly, if the legitimacy of Chevron deference is based on a congressional delegation of interpretive authority, then the line is one the Court must draw.

The majority's hypothetical Common Carrier Acts do not demonstrate anything different. Ante, at 1880 - 1881. The majority states that in its second Common Carrier Act, Section 2 makes clear that Congress " 'conferred interpretative power on the agency' " to interpret the ambiguous terms "common carrier" and "unreasonable condition." Ante, at 1880 - 1881 (quoting Brief for Petitioners in No. 1545, p. 14). Thus, it says, under anyone's theory a court must defer to the agency's reasonable interpretations of those terms. Correct.

The majority claims, however, that "petitioners' theory would accord the agency no deference" in its interpretation of the same ambiguous terms in the first Common Carrier Act. Ante, at 1880 - 1881. But as I understand petitioners' argument-and certainly in my own view-a court, in both cases, need only decide for itself whether Congress has delegated to the agency authority to interpret the ambiguous terms, before affording the agency's interpretation Chevron deference.

For the second Common Carrier Act, the answer is easy. The majority's hypothetical Congress has spoken clearly and specifically in Section 2 of the Act about its delegation of authority to interpret Section 1. As for the first Act, it is harder to analyze the question, given only one section of a presumably much larger statute. But if the first Common Carrier Act is like most agencies' organic statutes, I have no *326reason to doubt that the agency would likewise have interpretive authority over the same ambiguous terms, and therefore be entitled to deference in construing them, just as with the second Common Carrier Act. There is no new "test" to worry about, cf. ante, at 1885 - 1886; courts would simply apply the normal rules of statutory construction.

That the question might be harder with respect to the first Common Carrier Act should come as no surprise. The second hypothetical Congress has more carefully defined the agency's authority than the first. Whatever standard of review applies, it is more difficult to interpret an unclear statute than a clear one. My point is simply that before a court can defer to the agency's interpretation of the ambiguous terms in either Act, it must determine for itself that Congress has delegated authority to the agency to issue those interpretations with the force of law.

The majority also expresses concern that adopting petitioners' position would undermine Chevron 's stable background rule against which Congress legislates. Ante, at 1879 - 1880. That, of course, begs the question of what that stable background rule is. See Merrill & Hickman, Chevron 's Domain, 89 Geo. L.Rev. 833, 910 (2001) ("Courts have never deferred to agencies with respect to questions such as whether Congress has delegated to an agency the power to act with the force of law through either legislative rules or binding adjudications. Similarly, it has never been maintained that Congress would want courts to give Chevron deference to an agency's determination that it is entitled to Chevron deference, or should give Chevron deference to an agency's determination of what types of interpretations are entitled to Chevron deference" (footnote omitted)).

VI

The Court sees something nefarious behind the view that courts must decide on *1886their own whether Congress has delegated interpretative authority to an agency, before deferring to that agency's interpretation of law. What is afoot, according *327to the Court, is a judicial power-grab, with nothing less than "Chevron itself" as "the ultimate target." Ante, at 1873.

The Court touches on a legitimate concern: Chevron importantly guards against the Judiciary arrogating to itself policymaking properly left, under the separation of powers, to the Executive. But there is another concern at play, no less firmly rooted in our constitutional structure. That is the obligation of the Judiciary not only to confine itself to its proper role, but to ensure that the other branches do so as well.

An agency's interpretive authority, entitling the agency to judicial deference, acquires its legitimacy from a delegation of lawmaking power from Congress to the Executive. Our duty to police the boundary between the Legislature and the Executive is as critical as our duty to respect that between the Judiciary and the Executive. See Zivotofsky v. Clinton, 566 U.S. ----, ----, 132 S.Ct. 1421, 1428, 182 L.Ed.2d 423 (2012). In the present context, that means ensuring that the Legislative Branch has in fact delegated lawmaking power to an agency within the Executive Branch, before the Judiciary defers to the Executive on what the law is. That concern is heightened, not diminished, by the fact that the administrative agencies, as a practical matter, draw upon a potent brew of executive, legislative, and judicial power. And it is heightened, not diminished, by the dramatic shift in power over the last 50 years from Congress to the Executive-a shift effected through the administrative agencies.

We reconcile our competing responsibilities in this area by ensuring judicial deference to agency interpretations under Chevron -but only after we have determined on our own that Congress has given interpretive authority to the agency. Our "task is to fix the boundaries of delegated authority," Monaghan, 83 Colum. L.Rev., at 27; that is not a task we can delegate to the agency. We do not leave it to the agency to decide when it is in charge.

* * *

*328In these cases, the FCC issued a declaratory ruling interpreting the term "reasonable period of time" in 47 U.S.C. § 332(c)(7) (B)(ii). The Fifth Circuit correctly recognized that it could not apply Chevron deference to the FCC's interpretation unless the agency "possessed statutory authority to administer § 332(c)(7)(B)(ii)," but it erred by granting Chevron deference to the FCC's view on that antecedent question. See 668 F.3d, at 248. Because the court should have determined on its own whether Congress delegated interpretive authority over § 332(c)(7) (B)(ii) to the FCC before affording Chevron deference, I would vacate the decision below and remand the cases to the Fifth Circuit to perform the proper inquiry in the first instance.

I respectfully dissent.

Justice KAGAN announced the judgment of the Court and delivered the opinion of the Court with respect to Parts I, II-B, III-B, and IV, and an opinion with respect to Parts II-A and III-A, in which Justice GINSBURG, Justice BREYER, and Justice SOTOMAYOR join.

*2408This Court has often deferred to agencies' reasonable readings of genuinely ambiguous regulations. We call that practice Auer deference, or sometimes Seminole Rock deference, after two cases in which we employed it. See Auer v. Robbins , 519 U.S. 452, 117 S.Ct. 905, 137 L.Ed.2d 79 (1997) ; Bowles v. Seminole Rock & Sand Co. , 325 U.S. 410, 65 S.Ct. 1215, 89 L.Ed. 1700 (1945). The only question presented here is whether we should overrule those decisions, discarding the deference they give to agencies. We answer that question no. Auer deference retains an important role in construing agency regulations. But even as we uphold it, we reinforce its limits. Auer deference is sometimes appropriate and sometimes not. Whether to apply it depends on a range of considerations that we have noted now and again, but compile and further develop today. The deference doctrine we describe is potent in its place, but cabined in its scope. On remand, the Court of Appeals should decide whether it applies to the agency interpretation at issue.

I

We begin by summarizing how petitioner James Kisor's case made its way to this Court. Truth be told, nothing recounted in this Part has much bearing on the rest of *2409our decision. The question whether to overrule Auer does not turn on any single application, whether right or wrong, of that decision's deference doctrine. But a recitation of the facts and proceedings below at least shows how the question presented arose.

Kisor is a Vietnam War veteran seeking disability benefits from the Department of Veterans Affairs (VA). He first applied in 1982, alleging that he had developed post-traumatic stress disorder (PTSD) as a result of his participation in a military action called Operation Harvest Moon. The report of the agency's evaluating psychiatrist noted Kisor's involvement in that battle, but found that he "d[id] not suffer from PTSD." App. 12, 14. The VA thus denied Kisor benefits. There matters stood until 2006, when Kisor moved to reopen his claim. Based on a new psychiatric report, the VA this time agreed that Kisor suffered from PTSD. But it granted him benefits only from the date of his motion to reopen, rather than (as he requested) from the date of his first application.

The Board of Veterans' Appeals-a part of the VA, represented in Kisor's case by a single administrative judge-affirmed that timing decision, based on its interpretation of an agency rule. Under the VA's regulation, the agency could grant Kisor retroactive benefits if it found there were "relevant official service department records" that it had not considered in its initial denial. See 38 C.F.R. § 3.156(c)(1) (2013). The Board acknowledged that Kisor had come up with two new service records, both confirming his participation in Operation Harvest Moon. But according to the Board, those records were not "relevant" because they did not go to the reason for the denial-that Kisor did not have PTSD. See App. to Pet. for Cert. 43a ("[The] documents were not relevant to the decision in May 1983 because the basis of the denial was that a diagnosis of PTSD was not warranted, not a dispute as to whether or not the Veteran engaged in combat"). The Court of Appeals for Veterans Claims, an independent Article I court that initially reviews the Board's decisions, affirmed for the same reason.

The Court of Appeals for the Federal Circuit also affirmed, but it did so based on deference to the Board's interpretation of the VA rule. See Kisor v. Shulkin , 869 F.3d 1360, 1368 (2017). Kisor had argued to the Federal Circuit that to count as "relevant," a service record need not (as the Board thought) "counter[ ] the basis of the prior denial"; instead, it could relate to some other criterion for obtaining disability benefits. Id., at 1366 (internal quotation marks omitted). The Federal Circuit found the regulation "ambiguous" as between the two readings. Id., at 1367. The rule, said the court, does not specifically address "whether 'relevant' records are those casting doubt on the agency's prior [rationale or] those relating to the veteran's claim more broadly." Ibid. So how to choose between the two views? The court continued: "Both parties insist that the plain regulatory language supports their case, and neither party's position strikes us as unreasonable." Id. , at 1368. Because that was so, the court believed Auer deference appropriate: The agency's construction of its own regulation would govern unless "plainly erroneous or inconsistent with the VA's regulatory framework." Ibid. (internal quotation marks omitted). Applying that standard, the court upheld the Board's reading-and so approved the denial of retroactive benefits.

We then granted certiorari to decide whether to overrule Auer and (its predecessor) Seminole Rock . 586 U. S. ----, 139 S.Ct. 657, 202 L.Ed.2d 491 (2018).

*2410II

Before addressing that question directly, we spend some time describing what Auer deference is, and is not, for. You might view this Part as "just background" because we have made many of its points in prior decisions. But even if so, it is background that matters. For our account of why the doctrine emerged-and also how we have limited it-goes a long way toward explaining our view that it is worth preserving.

A

Begin with a familiar problem in administrative law: For various reasons, regulations may be genuinely ambiguous. They may not directly or clearly address every issue; when applied to some fact patterns, they may prove susceptible to more than one reasonable reading. Sometimes, this sort of ambiguity arises from careless drafting-the use of a dangling modifier, an awkward word, an opaque construction. But often, ambiguity reflects the well-known limits of expression or knowledge. The subject matter of a rule "may be so specialized and varying in nature as to be impossible"-or at any rate, impracticable-to capture in its every detail. SEC v. Chenery Corp. , 332 U.S. 194, 203, 67 S.Ct. 1760, 91 L.Ed. 1995 (1947). Or a "problem[ ] may arise" that the agency, when drafting the rule, "could not [have] reasonably foresee[n]." Id., at 202, 67 S.Ct. 1760. Whichever the case, the result is to create real uncertainties about a regulation's meaning.

Consider these examples:

• In a rule issued to implement the Americans with Disabilities Act (ADA), the Department of Justice requires theaters and stadiums to provide people with disabilities "lines of sight comparable to those for members of the general public." 28 C.F.R. pt. 36, App. A, p. 563 (1996). Must the Washington Wizards construct wheelchair seating to offer lines of sight over spectators when they rise to their feet? Or is it enough that the facility offers comparable views so long as everyone remains seated? See Paralyzed Veterans of Am. v. D. C. Arena L. P. , 117 F.3d 579, 581-582 (CADC 1997).

• The Transportation Security Administration (TSA) requires that liquids, gels, and aerosols in carry-on baggage be packed in containers smaller than 3.4 ounces and carried in a clear plastic bag. Does a traveler have to pack his jar of truffle pâté in that way? See Laba v. Copeland , 2016 WL 5958241, *1 (WDNC, Oct. 13, 2016).

• The Mine Safety and Health Administration issues a rule requiring employers to report occupational diseases within two weeks after they are "diagnosed." 30 C.F.R. § 50.20(a) (1993). Do chest X-ray results that "scor[e]" above some level of opacity count as a "diagnosis"? What level, exactly? See American Min. Congress v. Mine Safety and Health Admin. , 995 F.2d 1106, 1107-1108 (CADC 1993).

• An FDA regulation gives pharmaceutical companies exclusive rights to drug products if they contain "no active moiety that has been approved by FDA in any other" new drug application. 21 C.F.R. § 314.108(a) (2010). Has a company created a new "active moiety" by joining a previously approved moiety to lysine through a non-ester covalent bond? See Actavis Elizabeth LLC v. FDA , 625 F.3d 760, 762-763 (CADC 2010) ; Tr. of Oral Arg. 12, 35.1

*2411• Or take the facts of Auer itself. An agency must decide whether police captains are eligible for overtime under the Fair Labor Standards Act. According to the agency's regulations, employees cannot receive overtime if they are paid on a "salary basis." 29 C.F.R. § 541.118(a) (1996). And in deciding whether an employee is salaried, one question is whether his pay is "subject to reduction" based on performance. Ibid. A police department's manual informs its officers that their pay might be docked if they commit a disciplinary infraction. Does that fact alone make them "subject to" pay deductions? Or must the department have a practice of docking officer pay, so that the possibility of that happening is more than theoretical? 519 U.S. at 459-462, 117 S.Ct. 905.

In each case, interpreting the regulation involves a choice between (or among) more than one reasonable reading. To apply the rule to some unanticipated or unresolved situation, the court must make a judgment call. How should it do so?

In answering that question, we have often thought that a court should defer to the agency's construction of its own regulation. For the last 20 or so years, we have referred to that doctrine as Auer deference, and applied it often.2 But the name is something of a misnomer. Before the doctrine was called Auer deference, it was called Seminole Rock deference-for the 1945 decision in which we declared that when "the meaning of [a regulation] is in doubt," the agency's interpretation "becomes of controlling weight unless it is plainly erroneous or inconsistent with the regulation." 325 U.S. at 414, 65 S.Ct. 1215.3

*2412And Seminole Rock itself was not built on sand. Deference to administrative agencies traces back to the late nineteenth century, and perhaps beyond. See United States v. Eaton , 169 U.S. 331, 343, 18 S.Ct. 374, 42 L.Ed. 767 (1898) ("The interpretation given to the regulations by the department charged with their execution ... is entitled to the greatest weight"); see Brief for Administrative Law Scholars as Amici Curiae 5, n. 3 (collecting early cases); Brief for AFL-CIO as Amicus Curiae 8 (same).

We have explained Auer deference (as we now call it) as rooted in a presumption about congressional intent-a presumption that Congress would generally want the agency to play the primary role in resolving regulatory ambiguities. See Martin v. Occupational Safety and Health Review Comm'n , 499 U.S. 144, 151-153, 111 S.Ct. 1171, 113 L.Ed.2d 117 (1991). Congress, we have pointed out, routinely delegates to agencies the power to implement statutes by issuing rules. See id., at 151, 111 S.Ct. 1171. In doing so, Congress knows (how could it not?) that regulations will sometimes contain ambiguities. See supra, at 2410. But Congress almost never explicitly assigns responsibility to deal with that problem, either to agencies or to courts. Hence the need to presume, one way or the other, what Congress would want. And as between those two choices, agencies have gotten the nod. We have adopted the presumption-though it is always rebuttable-that "the power authoritatively to interpret its own regulations is a component of the agency's delegated lawmaking powers." Martin , 499 U.S. at 151, 111 S.Ct. 1171. Or otherwise said, we have thought that when granting rulemaking power to agencies, Congress usually intends to give them, too, considerable latitude to interpret the ambiguous rules they issue.

In part, that is because the agency that promulgated a rule is in the "better position [to] reconstruct" its original meaning. Id., at 152, 111 S.Ct. 1171. Consider that if you don't know what some text (say, a memo or an e-mail) means, you would probably want to ask the person who wrote it. And for the same reasons, we have thought, Congress would too (though the person is here a collective actor). The agency that "wrote the regulation" will often have direct insight into what that rule was intended to mean. Mullins Coal Co. of Va. v. Director, Office of Workers' Compensation Programs , 484 U.S. 135, 159, 108 S.Ct. 427, 98 L.Ed.2d 450 (1987). The drafters will know what it was supposed to include or exclude or how it was supposed to apply to some problem. To be sure, this justification has its limits. It does not work so well, for example, when the agency failed to anticipate an issue in crafting a rule (e.g., if the agency never thought about whether and when chest X-rays would count as a "diagnosis"). See supra, at 2410. Then, the agency will not be uncovering a specific intention; at most (though this is not nothing), it will be offering insight into the analogous issues the drafters considered and the purposes they designed the regulation to serve. And the defense works yet less well when lots of time has passed between the rule's issuance and its interpretation-especially if the interpretation differs from one that has come before. All that said, the point holds good for a significant category of "contemporaneous" readings. Lyng v. Payne , 476 U.S. 926, 939, 106 S.Ct. 2333, 90 L.Ed.2d 921 (1986). Want to know what a rule means? Ask its author.

*2413In still greater measure, the presumption that Congress intended Auer deference stems from the awareness that resolving genuine regulatory ambiguities often "entail[s] the exercise of judgment grounded in policy concerns." Thomas Jefferson Univ. v. Shalala , 512 U.S. 504, 512, 114 S.Ct. 2381, 129 L.Ed.2d 405 (1994) (internal quotation marks omitted). Return to our TSA example. See supra, at 2410. In most of their applications, terms like "liquids" and "gels" are clear enough. (Traveler checklist: Pretzels OK; water not.) But resolving the uncertain issues-the truffle pâtés or olive tapenades of the world-requires getting in the weeds of the rule's policy: Why does TSA ban liquids and gels in the first instance? What makes them dangerous? Can a potential hijacker use pâté jars in the same way as soda cans? Or take the less specialized-seeming ADA example. See supra, at 2410. It is easy enough to know what "comparable lines of sight" means in a movie theater-but more complicated when, as in sports arenas, spectators sometimes stand up. How costly is it to insist that the stadium owner take that sporadic behavior into account, and is the viewing value received worth the added expense? That cost-benefit calculation, too, sounds more in policy than in law. Or finally, take the more technical "moiety" example. See supra, at 2410 - 2411. Or maybe, don't. If you are a judge, you probably have no idea of what the FDA's rule means, or whether its policy is implicated when a previously approved moiety is connected to lysine through a non-ester covalent bond.

And Congress, we have thought, knows just that: It is attuned to the comparative advantages of agencies over courts in making such policy judgments. Agencies (unlike courts) have "unique expertise," often of a scientific or technical nature, relevant to applying a regulation "to complex or changing circumstances." Martin , 499 U.S. at 151, 111 S.Ct. 1171 ; see Thomas Jefferson , 512 U.S. at 512, 114 S.Ct. 2381. Agencies (unlike courts) can conduct factual investigations, can consult with affected parties, can consider how their experts have handled similar issues over the long course of administering a regulatory program. See Long Island Care at Home, Ltd. v. Coke , 551 U.S. 158, 167-168, 127 S.Ct. 2339, 168 L.Ed.2d 54 (2007). And agencies (again unlike courts) have political accountability, because they are subject to the supervision of the President, who in turn answers to the public. See Free Enterprise Fund v. Public Company Accounting Oversight Bd. , 561 U.S. 477, 499, 130 S.Ct. 3138, 177 L.Ed.2d 706 (2010) ; Pauley v. BethEnergy Mines, Inc. , 501 U.S. 680, 696, 111 S.Ct. 2524, 115 L.Ed.2d 604 (1991) (discussing as a matter of democratic accountability the "proper roles of the political and judicial branches" in filling regulatory gaps). It is because of those features that Congress, when first enacting a statute, assigns rulemaking power to an agency and thus authorizes it to fill out the statutory scheme. And so too, when new issues demanding new policy calls come up within that scheme, Congress presumably wants the same agency, rather than any court, to take the laboring oar.

Finally, the presumption we use reflects the well-known benefits of uniformity in interpreting genuinely ambiguous rules. We have noted Congress's frequent "preference for resolving interpretive issues by uniform administrative decision, rather than piecemeal by litigation." Ford Motor Credit Co. v. Milhollin , 444 U.S. 555, 568, 100 S.Ct. 790, 63 L.Ed.2d 22 (1980). That preference may be strongest when the interpretive issue arises in the context of a "complex and highly technical regulatory *2414program." Thomas Jefferson , 512 U.S. at 512, 114 S.Ct. 2381. After all, judges are most likely to come to divergent conclusions when they are least likely to know what they are doing. (Is there anything to be said for courts all over the country trying to figure out what makes for a new active moiety?) But the uniformity justification retains some weight even for more accessible rules, because their language too may give rise to more than one eminently reasonable reading. Consider Auer itself. See supra, at 2411 - 2412. There, four Circuits held that police captains were "subject to" pay deductions for disciplinary infractions if a police manual said they were, even if the department had never docked anyone. Two other Circuits held that captains were "subject to" pay deductions only if the department's actual practice made that punishment a realistic possibility. See Auer , 519 U.S. at 460, 117 S.Ct. 905. Had the agency issued an interpretation before all those rulings (rather than, as actually happened, in a brief in this Court), a deference rule would have averted most of that conflict and uncertainty. See Christopher v. SmithKline Beecham Corp. , 567 U.S. 142, 158, n. 17, 132 S.Ct. 2156, 183 L.Ed.2d 153 (2012) (noting for this reason that Auer deference imparts "predictability to the administrative process" (internal quotation marks omitted)). Auer deference thus serves to ensure consistency in federal regulatory law, for everyone who needs to know what it requires.

B

But all that said, Auer deference is not the answer to every question of interpreting an agency's rules. Far from it. As we explain in this section, the possibility of deference can arise only if a regulation is genuinely ambiguous. And when we use that term, we mean it-genuinely ambiguous, even after a court has resorted to all the standard tools of interpretation. Still more, not all reasonable agency constructions of those truly ambiguous rules are entitled to deference. As just explained, we presume that Congress intended for courts to defer to agencies when they interpret their own ambiguous rules. See supra, at 2411 - 2414. But when the reasons for that presumption do not apply, or countervailing reasons outweigh them, courts should not give deference to an agency's reading, except to the extent it has the "power to persuade." Christopher , 567 U.S. at 159, 132 S.Ct. 2156 (quoting Skidmore v. Swift & Co. , 323 U.S. 134, 140, 65 S.Ct. 161, 89 L.Ed. 124 (1944) ). We have thus cautioned that Auer deference is just a "general rule"; it "does not apply in all cases." Christopher , 567 U.S. at 155, 132 S.Ct. 2156. And although the limits of Auer deference are not susceptible to any rigid test, we have noted various circumstances in which such deference is "unwarranted." Ibid . In particular, that will be so when a court concludes that an interpretation does not reflect an agency's authoritative, expertise-based, "fair[, or] considered judgment." Ibid. (quoting Auer , 519 U.S. at 462, 117 S.Ct. 905 ); cf. United States v. Mead Corp. , 533 U.S. 218, 229-231, 121 S.Ct. 2164, 150 L.Ed.2d 292 (2001) (adopting a similar approach to Chevron deference).

We take the opportunity to restate, and somewhat expand on, those principles here to clear up some mixed messages we have sent. At times, this Court has applied Auer deference without significant analysis of the underlying regulation. See, e.g., United States v. Larionoff , 431 U.S. 864, 872, 97 S.Ct. 2150, 53 L.Ed.2d 48 (1977) (stating that the Court "need not tarry" over the regulation's language given Seminole Rock ). At other times, the Court has given Auer deference without careful attention to the nature and context of the interpretation.

*2415See, e.g., Thorpe v. Housing Authority of Durham , 393 U.S. 268, 276, and nn. 22-23, 89 S.Ct. 518, 21 L.Ed.2d 474 (1969) (deferring to an agency's view as expressed in letters to third parties). And in a vacuum, our most classic formulation of the test-whether an agency's construction is "plainly erroneous or inconsistent with the regulation," Seminole Rock , 325 U.S. at 414, 65 S.Ct. 1215 -may suggest a caricature of the doctrine, in which deference is "reflexive." Pereira v. Sessions , 585 U. S. ----, ----, 138 S.Ct. 2105, 2120, 201 L.Ed.2d 433 (2018) (KENNEDY, J., concurring). So we cannot deny that Kisor has a bit of grist for his claim that Auer "bestows on agencies expansive, unreviewable" authority. Brief for Petitioner 25. But in fact Auer does no such thing: It gives agencies their due, while also allowing-indeed, obligating-courts to perform their reviewing and restraining functions. So before we turn to Kisor's specific grievances, we think it worth reinforcing some of the limits inherent in the Auer doctrine.4

First and foremost, a court should not afford Auer deference unless the regulation is genuinely ambiguous. See Christensen v. Harris County , 529 U.S. 576, 588, 120 S.Ct. 1655, 146 L.Ed.2d 621 (2000) ; Seminole Rock , 325 U.S. at 414, 65 S.Ct. 1215 (deferring only "if the meaning of the words used is in doubt"). If uncertainty does not exist, there is no plausible reason for deference. The regulation then just means what it means-and the court must give it effect, as the court would any law. Otherwise said, the core theory of Auer deference is that sometimes the law runs out, and policy-laden choice is what is left over. See supra, at 2412 - 2413. But if the law gives an answer-if there is only one reasonable construction of a regulation-then a court has no business deferring to any other reading, no matter how much the agency insists it would make more sense. Deference in that circumstance would "permit the agency, under the guise of interpreting a regulation, to create de facto a new regulation." See Christensen , 529 U.S. at 588, 120 S.Ct. 1655. Auer does not, and indeed could not, go that far.

And before concluding that a rule is genuinely ambiguous, a court must exhaust all the "traditional tools" of construction. Chevron U. S. A. Inc. v. Natural Resources Defense Council, Inc. , 467 U.S. 837, 843, n. 9, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984) (adopting the same approach for ambiguous statutes). For again, only when that legal toolkit is empty and the interpretive question still has no single right answer can a judge conclude that it is "more [one] of policy than of law." Pauley , 501 U.S. at 696, 111 S.Ct. 2524. That means a court cannot wave the ambiguity flag just because it found the regulation impenetrable on first read. Agency regulations can sometimes make the eyes glaze over. But hard interpretive conundrums, even relating to complex rules, can often be solved. See id., at 707, 111 S.Ct. 2524 (SCALIA, J., dissenting) (A regulation is not ambiguous merely because "discerning the only possible interpretation requires a taxing inquiry"). To make that effort, a court must "carefully consider[ ]" the text, structure, history, and purpose of a regulation, in all the ways it would if it had no agency to fall back on. Ibid. Doing so will resolve many seeming ambiguities out of the box, without resort to Auer deference.

If genuine ambiguity remains, moreover, the agency's reading must still be "reasonable." Thomas Jefferson , 512 U.S. at 515, 114 S.Ct. 2381. In other words, *2416it must come within the zone of ambiguity the court has identified after employing all its interpretive tools. (Note that serious application of those tools therefore has use even when a regulation turns out to be truly ambiguous. The text, structure, history, and so forth at least establish the outer bounds of permissible interpretation.) Some courts have thought (perhaps because of Seminole Rock 's "plainly erroneous" formulation) that at this stage of the analysis, agency constructions of rules receive greater deference than agency constructions of statutes. See, e.g., Ohio Dept. of Medicaid v. Price , 864 F.3d 469, 477 (CA6 2017). But that is not so. Under Auer , as under Chevron , the agency's reading must fall "within the bounds of reasonable interpretation." Arlington v. FCC , 569 U.S. 290, 296, 133 S.Ct. 1863, 185 L.Ed.2d 941 (2013). And let there be no mistake: That is a requirement an agency can fail.

Still, we are not done-for not every reasonable agency reading of a genuinely ambiguous rule should receive Auer deference. We have recognized in applying Auer that a court must make an independent inquiry into whether the character and context of the agency interpretation entitles it to controlling weight. See Christopher , 567 U.S. at 155, 132 S.Ct. 2156 ; see also Mead , 533 U.S. at 229-231, 236-237, 121 S.Ct. 2164 (requiring an analogous though not identical inquiry for Chevron deference). As explained above, we give Auer deference because we presume, for a set of reasons relating to the comparative attributes of courts and agencies, that Congress would have wanted us to. See supra, at 2411 - 2414. But the administrative realm is vast and varied, and we have understood that such a presumption cannot always hold. Cf. Mead , 533 U.S. at 236, 121 S.Ct. 2164 ("tailor[ing] deference to [the] variety" of administrative action); Arlington , 569 U.S. at 309-310, 133 S.Ct. 1863 (BREYER, J., concurring in part and concurring in judgment) (noting that "context-specific[ ] factors" may show that "Congress would [not] have intended the agency to resolve [some] ambiguity"). The inquiry on this dimension does not reduce to any exhaustive test. But we have laid out some especially important markers for identifying when Auer deference is and is not appropriate.

To begin with, the regulatory interpretation must be one actually made by the agency. In other words, it must be the agency's "authoritative" or "official position," rather than any more ad hoc statement not reflecting the agency's views. Mead , 533 U.S. at 257-259, and n. 6, 121 S.Ct. 2164 (SCALIA, J., dissenting). That constraint follows from the logic of Auer deference-because Congress has delegated rulemaking power, and all that typically goes with it, to the agency alone. Of course, the requirement of "authoritative" action must recognize a reality of bureaucratic life: Not everything the agency does comes from, or is even in the name of, the Secretary or his chief advisers. So, for example, we have deferred to "official staff memoranda" that were "published in the Federal Register," even though never approved by the agency head. Ford Motor Credit , 444 U.S. at 566, n. 9, 567, n. 10, 100 S.Ct. 790 (declining to "draw a radical distinction between" agency heads and staff for Auer deference). But there are limits. The interpretation must at the least emanate from those actors, using those vehicles, understood to make authoritative policy in the relevant context. See, e.g., Paralyzed Veterans , 117 F.3d at 587 (refusing to consider a "speech of a mid-level official" as an "authoritative departmental position"); N. Y. State Dept. of Social Servs. v. Bowen , 835 F.2d 360, 365-366 (CADC 1987) (rejecting the idea that an *2417"informal memorandum" recounting a telephone conversation between employees could count as an "authoritative pronouncement"); Exelon Generation Co. v. Local 15, Int'l Brotherhood of Elec. Workers, AFL-CIO , 676 F.3d 566, 576-578 (CA7 2012) (declining deference when the agency had itself "disclaimed the use of regulatory guides as authoritative"). If the interpretation does not do so, a court may not defer.

Next, the agency's interpretation must in some way implicate its substantive expertise. Administrative knowledge and experience largely "account [for] the presumption that Congress delegates interpretive lawmaking power to the agency." Martin , 499 U.S. at 153, 111 S.Ct. 1171. So the basis for deference ebbs when "[t]he subject matter of the [dispute is] distan[t] from the agency's ordinary" duties or "fall[s] within the scope of another agency's authority." Arlington , 569 U.S. at 309, 133 S.Ct. 1863 (opinion of BREYER, J.). This Court indicated as much when it analyzed a "split enforcement" scheme, in which Congress divided regulatory power between two entities. Martin , 499 U.S. at 151, 111 S.Ct. 1171. To decide "whose reasonable interpretation" of a rule controlled, we "presum[ed] Congress intended to invest interpretive power" in whichever actor was "best position[ed] to develop" expertise about the given problem. Id. , at 149, 153, 111 S.Ct. 1171. The same idea holds good as between agencies and courts. "Generally, agencies have a nuanced understanding of the regulations they administer." Brief for Respondent 33. That point is most obvious when a rule is technical; think back to our "moiety" or "diagnosis" examples. See supra , at 2410 - 2411. But more prosaic-seeming questions also commonly implicate policy expertise; consider the TSA assessing the security risks of pâté or a disabilities office weighing the costs and benefits of an accommodation. See ibid. Once again, though, there are limits. Some interpretive issues may fall more naturally into a judge's bailiwick. Take one requiring the elucidation of a simple common-law property term, see Jicarilla Apache Tribe v. FERC , 578 F.2d 289, 292-293 (CA10 1978), or one concerning the award of an attorney's fee, see West Va. Highlands Conservancy, Inc. v. Norton , 343 F.3d 239 (CA4 2003). Cf. Adams Fruit Co. v. Barrett , 494 U.S. 638, 649-650, 110 S.Ct. 1384, 108 L.Ed.2d 585 (1990) (declining to award Chevron deference when an agency interprets a judicial-review provision). When the agency has no comparative expertise in resolving a regulatory ambiguity, Congress presumably would not grant it that authority.5

Finally, an agency's reading of a rule must reflect "fair and considered judgment" to receive Auer deference. Christopher , 567 U.S. at 155, 132 S.Ct. 2156 (quoting Auer , 519 U.S. at 462, 117 S.Ct. 905 ). That means, we have stated, that a court should decline to defer to a merely "convenient litigating position" or "post hoc rationalizatio[n] advanced" to "defend past agency action against attack." Christopher , 567 U.S. at 155, 132 S.Ct. 2156 (quoting Bowen v. Georgetown Univ. Hospital , 488 U.S. 204, 213, 109 S.Ct. 468, 102 L.Ed.2d 493 (1988) and Auer , 519 U.S. at 462, 117 S.Ct. 905 ).6 And a court may *2418not defer to a new interpretation, whether or not introduced in litigation, that creates "unfair surprise" to regulated parties. Long Island Care , 551 U.S. at 170, 127 S.Ct. 2339. That disruption of expectations may occur when an agency substitutes one view of a rule for another. We have therefore only rarely given Auer deference to an agency construction "conflict[ing] with a prior" one. Thomas Jefferson , 512 U.S. at 515, 114 S.Ct. 2381. Or the upending of reliance may happen without such an explicit interpretive change. This Court, for example, recently refused to defer to an interpretation that would have imposed retroactive liability on parties for longstanding conduct that the agency had never before addressed. See Christopher , 567 U.S. at 155-156, 132 S.Ct. 2156. Here too the lack of "fair warning" outweighed the reasons to apply Auer. Id., at 156, 132 S.Ct. 2156 (internal quotation marks omitted).

* * *

The upshot of all this goes something as follows. When it applies, Auer deference gives an agency significant leeway to say what its own rules mean. In so doing, the doctrine enables the agency to fill out the regulatory scheme Congress has placed under its supervision. But that phrase "when it applies" is important-because it often doesn't. As described above, this Court has cabined Auer 's scope in varied and critical ways-and in exactly that measure, has maintained a strong judicial role in interpreting rules. What emerges is a deference doctrine not quite so tame as some might hope, but not nearly so menacing as they might fear.

III

That brings us to the lone question presented here-whether we should abandon the longstanding doctrine just described. In contending that we should, Kisor raises statutory, policy, and constitutional claims (in that order). But he faces an uphill climb. He must first convince us that Auer deference is wrong. And even then, he must overcome stare decisis -the special care we take to preserve our precedents. In the event, Kisor fails at the first step: None of his arguments provide good reason to doubt Auer deference. And even if that were not so, Kisor does not offer the kind of special justification needed to overrule Auer, and Seminole Rock, and all our many other decisions deferring to reasonable agency constructions of ambiguous rules.

A

Kisor first attacks Auer as inconsistent with the judicial review provision of the Administrative Procedure Act (APA). See 5 U.S.C. § 706. As Kisor notes, Congress enacted the APA in 1946-the year after Seminole Rock -to serve as "the fundamental charter of the administrative state." Brief for Petitioner 26 (internal quotation marks omitted). Section 706 of the Act, governing judicial review of agency action, states (among other things) that reviewing courts shall "determine the meaning or applicability of the terms of an agency action" (including a regulation). According to Kisor, Auer violates that edict by thwarting "meaningful judicial review"

*2419of agency rules. Brief for Petitioner 29. Courts under Auer , he asserts (now in the language of Section 706 ), "abdicate their office of determining the meaning" of a regulation. Id., at 27 (internal quotation marks omitted).

To begin with, that argument ignores the many ways, discussed above, that courts exercise independent review over the meaning of agency rules. See supra, at 2415 - 2418. As we have explained, a court must apply all traditional methods of interpretation to any rule, and must enforce the plain meaning those methods uncover. There can be no thought of deference unless, after performing that thoroughgoing review, the regulation remains genuinely susceptible to multiple reasonable meanings and the agency's interpretation lines up with one of them. And even if that is the case, courts must on their own determine whether the nature or context of the agency's construction reverses the usual presumption of deference. Most notably, a court must consider whether the interpretation is authoritative, expertise-based, considered, and fair to regulated parties. All of that figures as "meaningful judicial review." Brief for Petitioner 29.

And even when a court defers to a regulatory reading, it acts consistently with Section 706. That provision does not specify the standard of review a court should use in "determin[ing] the meaning" of an ambiguous rule. 5 U.S.C. § 706. One possibility, as Kisor says, is to review the issue de novo . But another is to review the agency's reading for reasonableness. To see the point, assume that a regulatory (say, an employment) statute expressly instructed courts to apply Auer deference when reviewing an agency's interpretations of its ambiguous rules. Nothing in that statute would conflict with Section 706. Instead, the employment law would simply make clear how a court is to "determine the meaning" of such a rule-by deferring to an agency's reasonable reading. Ibid . Of course, that is not the world we know: Most substantive statutes do not say anything about Auer deference, one way or the other. But for all the reasons spelled out above, we have long presumed (subject always to rebuttal) that the Congress delegating regulatory authority to an agency intends as well to give that agency considerable latitude to construe its ambiguous rules. See supra, at 2411 - 2414. And that presumption operates just like the hypothesized statute above. Because of it, once again, courts do not violate Section 706 by applying Auer. To the contrary, they fulfill their duty to "determine the meaning" of a rule precisely by deferring to the agency's reasonable reading. See Sunstein & Vermeule, The Unbearable Rightness of Auer , 84 U. Chi. L. Rev. 297, 306 (2017) (If Congress intends "that the meaning of a regulation turns on the agency's interpretation of its meaning," then courts comply with Section 706 's command to " 'determine the meaning' [of the regulation] by deferring to that view"); cf. Arlington , 569 U.S. at 317, 133 S.Ct. 1863 (ROBERTS, C. J., dissenting) (similarly addressing why Chevron deference comports with Section 706 ). Section 706 and Auer thus go hand in hand.

That is especially so given the practice of judicial review at the time of the APA's enactment. Section 706 was understood when enacted to "restate[ ] the present law as to the scope of judicial review." See Dept. of Justice, Attorney General's Manual on the Administrative Procedure Act 108 (1947); see also Vermont Yankee Nuclear Power Corp. v. Natural Resources Defense Council, Inc. , 435 U.S. 519, 546, 98 S.Ct. 1197, 55 L.Ed.2d 460 (1978) (noting that this Court gives some deference to the Manual "because of the role played by the Department of Justice in drafting the legislation"). We have thus interpreted the *2420APA not to "significantly alter the common law of judicial review of agency action." Heckler v. Chaney , 470 U.S. 821, 832, 105 S.Ct. 1649, 84 L.Ed.2d 714 (1985) (internal quotation marks omitted). That pre-APA common law included Seminole Rock itself (decided the year before) along with prior decisions foretelling that ruling. See supra, at 2411. Even assume that the deference regime laid out in those cases had not yet fully taken hold. At a minimum, nothing in the law of that era required all judicial review of agency interpretations to be de novo . Cf. Manning, Constitutional Structure and Judicial Deference to Agency Interpretations of Agency Rules, 96 Colum. L. Rev. 612, 635-636 (1996) (arguing that courts before the APA used "flexible, common law methods to review administrative action"). And so nothing suggests that Section 706 imposes that requirement. Or otherwise said: If Section 706 did not change the law of judicial review (as we have long recognized), then it did not proscribe a deferential standard then known and in use.

Kisor next claims that Auer circumvents the APA's rulemaking requirements. Section 553, as Kisor notes, mandates that an agency use notice-and-comment procedures before issuing legislative rules. See 5 U.S.C. §§ 553(b), (c). But the section allows agencies to issue "interpret[ive]" rules without notice and comment. See § 553(b)(A). A key feature of those rules is that (unlike legislative rules) they are not supposed to "have the force and effect of law"-or, otherwise said, to bind private parties. Perez v. Mortgage Bankers Assn. , 575 U. S. 92, ----, 135 S.Ct. 1199, 1204, 191 L.Ed.2d 186 (2015) (internal quotation marks omitted). Instead, interpretive rules are meant only to "advise the public" of how the agency understands, and is likely to apply, its binding statutes and legislative rules. Ibid. But consider, Kisor argues, what happens when a court gives Auer deference to an interpretive rule. The result, he asserts, is to make a rule that has never gone through notice and comment binding on the public. See Brief for Petitioner 21, 29. Or put another way, the interpretive rule ends up having the "force and effect of law" without ever paying the procedural cost. Mortgage Bankers , 575 U. S., at ----, 135 S.Ct., at 1204.

But this Court rejected the identical argument just a few years ago, and for good reason. In Mortgage Bankers , we held that interpretive rules, even when given Auer deference, do not have the force of law. See 575 U. S., at ----, and n. 4, 135 S.Ct., at 1208, and n. 4. An interpretive rule itself never forms "the basis for an enforcement action"-because, as just noted, such a rule does not impose any "legally binding requirements" on private parties. National Min. Assn. v. McCarthy , 758 F.3d 243, 251 (CADC 2014). An enforcement action must instead rely on a legislative rule, which (to be valid) must go through notice and comment. And in all the ways discussed above, the meaning of a legislative rule remains in the hands of courts, even if they sometimes divine that meaning by looking to the agency's interpretation. See supra, at 2415 - 2418. Courts first decide whether the rule is clear; if it is not, whether the agency's reading falls within its zone of ambiguity; and even if the reading does so, whether it should receive deference. In short, courts retain the final authority to approve-or not-the agency's reading of a notice-and-comment rule. See Mortgage Bankers, 575 U. S., at ----, n. 4, 135 S.Ct., at 1208, n. 4 ("[I]t is the court that ultimately decides whether a given regulation means what the agency says"). No binding of anyone occurs merely by the agency's say-so.

And indeed, a court deciding whether to give Auer deference must heed the same procedural values as Section 553 reflects.

*2421Remember that a court may defer to only an agency's authoritative and considered judgments. See supra, at 2416 - 2418. No ad hoc statements or post hoc rationalizations need apply. And recall too that deference turns on whether an agency's interpretation creates unfair surprise or upsets reliance interests. See supra, at 2417 - 2418. So an agency has a strong incentive to circulate its interpretations early and widely. In such ways, the doctrine of Auer deference reinforces, rather than undermines, the ideas of fairness and informed decisionmaking at the core of the APA.

To supplement his two APA arguments, Kisor turns to policy, leaning on a familiar claim about the incentives Auer creates. According to Kisor, Auer encourages agencies to issue vague and open-ended regulations, confident that they can later impose whatever interpretation of those rules they prefer. See Brief for Petitioner 37-41. That argument received its fullest elaboration in a widely respected law review article pre-dating Auer . See Manning, 96 Colum. L. Rev., at 654-669. More recently, the concern about such self-delegation has appeared in opinions from this Court, starting with several from Justice SCALIA calling for Auer 's reconsideration. See, e.g., Christopher, 567 U.S. at 158, 132 S.Ct. 2156 (citing Manning, supra, at 655-668 ); Decker v. Northwest Environmental Defense Center , 568 U.S. 597, 620-621, 133 S.Ct. 1326, 185 L.Ed.2d 447 (2013) (SCALIA, J., concurring in part and dissenting in part) (citing Manning, supra ); Talk America, Inc. v. Michigan Bell Telephone Co. , 564 U.S. 50, 69, 131 S.Ct. 2254, 180 L.Ed.2d 96 (2011) (SCALIA, J., concurring) (principally relying on Manning, supra ).

But the claim has notable weaknesses, empirical and theoretical alike. First, it does not survive an encounter with experience. No real evidence-indeed, scarcely an anecdote-backs up the assertion. As two noted scholars (one of whom reviewed thousands of rules during four years of government service) have written: "[W]e are unaware of, and no one has pointed to, any regulation in American history that, because of Auer , was designed vaguely." Sunstein & Vermeule, 84 U. Chi. L. Rev., at 308. And even the argument's theoretical allure dissipates upon reflection. For strong (almost surely stronger) incentives and pressures cut in the opposite direction. "[R]egulators want their regulations to be effective, and clarity promotes compliance." Brief for Administrative Law Scholars as Amici Curiae 18-19. Too, regulated parties often push for precision from an agency, so that they know what they can and cannot do. And ambiguities in rules pose risks to the long-run survival of agency policy. Vagueness increases the chance of adverse judicial rulings. And it enables future administrations, with different views, to reinterpret the rules to their own liking. Add all of that up and Kisor's ungrounded theory of incentives contributes nothing to the case against Auer .

Finally, Kisor goes big, asserting (though fleetingly) that Auer deference violates "separation-of-powers principles." See Brief for Petitioner 43. In his view, those principles prohibit "vest[ing] in a single branch the law-making and law-interpreting functions." Id., at 45. If that objection is to agencies' usurping the interpretive role of courts, this opinion has already met it head-on. Properly understood and applied, Auer does no such thing. In all the ways we have described, courts retain a firm grip on the interpretive function. See supra, at 2415 - 2418; Mortgage Bankers, 575 U. S., at ----, n. 4, 135 S.Ct., at 1208, n. 4. If Kisor's objection is instead to the supposed commingling of functions (that is, the legislative and judicial) within an agency, this Court has answered it often before. See, *2422e.g., Withrow v. Larkin , 421 U.S. 35, 54, 95 S.Ct. 1456, 43 L.Ed.2d 712 (1975) (permitting such a combination of functions); FTC v. Cement Institute , 333 U.S. 683, 702, 68 S.Ct. 793, 92 L.Ed. 1010 (1948) (same). That sort of mixing is endemic in agencies, and has been "since the beginning of the Republic." Arlington , 569 U.S. at 304-305, n. 4, 133 S.Ct. 1863. It does not violate the separation of powers, we have explained, because even when agency "activities take 'legislative' and 'judicial' forms," they continue to be "exercises of[ ] the 'executive Power' "-or otherwise said, ways of executing a statutory plan. Ibid. (quoting U. S. Const., Art. II, § 1, cl. 1 ). So Kisor's last argument to dispatch Auer deference fails as roundly as the rest.

B

If all that were not enough, stare decisis cuts strongly against Kisor's position. "Overruling precedent is never a small matter." Kimble v. Marvel Entertainment, LLC , 576 U. S. ----, ----, 135 S.Ct. 2401, 2409, 192 L.Ed.2d 463 (2015). Adherence to precedent is "a foundation stone of the rule of law." Michigan v. Bay Mills Indian Community , 572 U.S. 782, 798, 134 S.Ct. 2024, 188 L.Ed.2d 1071 (2014). "[I]t promotes the evenhanded, predictable, and consistent development of legal principles, fosters reliance on judicial decisions, and contributes to the actual and perceived integrity of the judicial process." Payne v. Tennessee , 501 U.S. 808, 827, 111 S.Ct. 2597, 115 L.Ed.2d 720 (1991). To be sure, stare decisis is "not an inexorable command." Id., at 828, 111 S.Ct. 2597. But any departure from the doctrine demands "special justification"-something more than "an argument that the precedent was wrongly decided." Halliburton Co. v. Erica P. John Fund, Inc. , 573 U.S. 258, 266, 134 S.Ct. 2398, 189 L.Ed.2d 339 (2014).

And that is even more than usually so in the circumstances here. First, Kisor asks us to overrule not a single case, but a "long line of precedents"-each one reaffirming the rest and going back 75 years or more. Bay Mills , 572 U.S. at 798, 134 S.Ct. 2024 ; see nn. 2, 3, supra . This Court alone has applied Auer or Seminole Rock in dozens of cases, and lower courts have done so thousands of times. Deference to reasonable agency interpretations of ambiguous rules pervades the whole corpus of administrative law. Second, because that is so, abandoning Auer deference would cast doubt on many settled constructions of rules. As Kisor acknowledged at oral argument, a decision in his favor would allow relitigation of any decision based on Auer , forcing courts to "wrestle [with] whether or not Auer " had actually made a difference. Tr. of Oral Arg. 30; see id., at 47 (Solicitor General agreeing that "every single regulation that's currently on the books whose interpretation has been established under Seminole Rock now [would have] to be relitigated anew"). It is the rare overruling that introduces so much instability into so many areas of law, all in one blow.

And third, even if we are wrong about Auer , "Congress remains free to alter what we have done." Patterson v. McLean Credit Union , 491 U.S. 164, 172-173, 109 S.Ct. 2363, 105 L.Ed.2d 132 (1989) (stating that when that is so, "[c]onsiderations of stare decisis have special force"). In a constitutional case, only we can correct our error. But that is not so here. Our deference decisions are "balls tossed into Congress's court, for acceptance or not as that branch elects." Kimble , 576 U. S., at ----, 135 S.Ct., at 2409. And so far, at least, Congress has chosen acceptance. It could amend the APA or any specific statute to require the sort of de novo review of regulatory interpretations *2423that Kisor favors. Instead, for approaching a century, it has let our deference regime work side-by-side with both the APA and the many statutes delegating rulemaking power to agencies. It has done so even after we made clear that our deference decisions reflect a presumption about congressional intent. See Martin , 499 U.S. at 151, 111 S.Ct. 1171 ; supra , at 2411 - 2412. And it has done so even after Members of this Court began to raise questions about the doctrine. See, e.g., Talk America , 564 U.S. at 67-69, 131 S.Ct. 2254 (SCALIA, J., concurring). Given that history-and Congress's continuing ability to take up Kisor's arguments-we would need a particularly "special justification" to now reverse Auer.

Kisor offers nothing of that ilk. Nearly all his arguments about abandoning precedent are variants of his merits claims. We hear again, if in different parts of his briefs, that Auer deference frustrates "the policies embodied in the APA" and violates the separation of powers. Reply Brief 13, and n. 5; Brief for Petitioner 47-48. More generally, we learn that Seminole Rock was "wrong on its own terms" and "badly reasoned." Id., at 47 (internal quotation marks omitted). Of course, it is good-and important-for our opinions to be right and well-reasoned. But that is not the test for overturning precedent. Kisor does not claim that Auer deference is "unworkable," a traditional basis for overruling a case. Patterson , 491 U.S. at 173, 109 S.Ct. 2363. Nor does he point to changes in legal rules that make Auer a "doctrinal dinosaur." Kimble , 576 U. S., at ----, 135 S.Ct., at 2411. All he can muster is that "[t]he administrative state has evolved substantially since 1945." Brief for Petitioner 53. We do not doubt the point (although we note that Auer and other key deference decisions came along after most of that evolution took place). Still more, we agree with Kisor that administrative law doctrines must take account of the far-reaching influence of agencies and the opportunities such power carries for abuse. That is one reason we have taken care today to reinforce the limits of Auer deference, and to emphasize the critical role courts retain in interpreting rules. But it is no answer to the growth of agencies for courts to take over their expertise-based, policymaking functions. Who knows? Maybe in 1945, the FDA was not thinking about "active moieties." See supra, at 2410 - 2411. But still, today-just as Seminole Rock and Auer held-it should have leeway to say what that term means.

IV

With that, we can finally return to Kisor's own case. You may remember that his retroactive benefits depend on the meaning of the term "relevant" records in a VA regulation. See supra, at 2408 - 2409. The Board of Veterans' Appeals, through a single judge's opinion, understood records to be relevant only if they relate to the basis of the VA's initial denial of benefits. By contrast, Kisor argued that records are relevant if they go to any benefits criterion, even one that was uncontested. The Federal Circuit upheld the Board's interpretation based on Auer deference.

Applying the principles outlined in this opinion, we hold that a redo is necessary for two reasons. First, the Federal Circuit jumped the gun in declaring the regulation ambiguous. We have insisted that a court bring all its interpretive tools to bear before finding that to be so. See supra, at 2415 - 2416. It is not enough to casually remark, as the court did here, that "[b]oth parties insist that the plain regulatory language supports their case, and neither party's position strikes us as unreasonable." 869 F.3d at 1368 ; see supra, at 2415 - 2416. Rather, the court must *2424make a conscientious effort to determine, based on indicia like text, structure, history, and purpose, whether the regulation really has more than one reasonable meaning. The Solicitor General argued in this Court that the Board's reading is the only reasonable one. See Brief for Respondent 49-50. Perhaps Kisor will make the converse claim below. Before even considering deference, the court must seriously think through those positions.

And second, the Federal Circuit assumed too fast that Auer deference should apply in the event of genuine ambiguity. As we have explained, that is not always true. A court must assess whether the interpretation is of the sort that Congress would want to receive deference. See supra, at 2416 - 2418. The Solicitor General suggested at oral argument that the answer in this case might be no. He explained that all 100 or so members of the VA Board act individually (rather than in panels) and that their roughly 80,000 annual decisions have no "precedential value." Tr. of Oral Arg. 64. He thus questioned whether a Board member's ruling "reflects the considered judgment of the agency as a whole." Ibid. ; cf. Mead , 533 U.S. at 233, 121 S.Ct. 2164 (declining to give Chevron deference to rulings "being churned out at a rate of 10,000 a year at an agency's 46 scattered offices"). We do not know what position the Government will take on that issue below. But the questions the Solicitor General raised are exactly the kind the court must consider in deciding whether to award Auer deference to the Board's interpretation.

We accordingly vacate the judgment below and remand the case for further proceedings.

It is so ordered.

Chief Justice ROBERTS, concurring in part.

I join Parts I, II-B, III-B, and IV of the Court's opinion. We took this case to consider whether to overrule Auer v. Robbins , 519 U.S. 452, 117 S.Ct. 905, 137 L.Ed.2d 79 (1997), and Bowles v. Seminole Rock & Sand Co. , 325 U.S. 410, 65 S.Ct. 1215, 89 L.Ed. 1700 (1945). For the reasons the Court discusses in Part III-B, I agree that overruling those precedents is not warranted. I also agree with the Court's treatment in Part II-B of the bounds of Auer deference.

I write separately to suggest that the distance between the majority and Justice GORSUCH is not as great as it may initially appear. The majority catalogs the prerequisites for, and limitations on, Auer deference: The underlying regulation must be genuinely ambiguous; the agency's interpretation must be reasonable and must reflect its authoritative, expertise-based, and fair and considered judgment; and the agency must take account of reliance interests and avoid unfair surprise. Justice GORSUCH, meanwhile, lists the reasons that a court might be persuaded to adopt an agency's interpretation of its own regulation: The agency thoroughly considered the problem, offered a valid rationale, brought its expertise to bear, and interpreted the regulation in a manner consistent with earlier and later pronouncements. Accounting for variations in verbal formulation, those lists have much in common.

That is not to say that Auer deference is just the same as the power of persuasion discussed in Skidmore v. Swift & Co. , 323 U.S. 134, 65 S.Ct. 161, 89 L.Ed. 124 (1944) ; there is a difference between holding that a court ought to be persuaded by an agency's interpretation and holding that it should defer to that interpretation under certain conditions. But it is to say that the *2425cases in which Auer deference is warranted largely overlap with the cases in which it would be unreasonable for a court not to be persuaded by an agency's interpretation of its own regulation.

One further point: Issues surrounding judicial deference to agency interpretations of their own regulations are distinct from those raised in connection with judicial deference to agency interpretations of statutes enacted by Congress. See Chevron U. S. A. Inc. v. Natural Resources Defense Council, Inc. , 467 U.S. 837, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984). I do not regard the Court's decision today to touch upon the latter question.

Justice GORSUCH, with whom Justice THOMAS joins, with whom Justice KAVANAUGH joins as to Parts I, II, III, IV, and V, and with whom Justice ALITO joins as to Parts I, II, and III, concurring in the judgment.

It should have been easy for the Court to say goodbye to Auer v. Robbins .1 In disputes involving the relationship between the government and the people, Auer requires judges to accept an executive agency's interpretation of its own regulations even when that interpretation doesn't represent the best and fairest reading. This rule creates a "systematic judicial bias in favor of the federal government, the most powerful of parties, and against everyone else."2 Nor is Auer 's biased rule the product of some congressional mandate we are powerless to correct: This Court invented it, almost by accident and without any meaningful effort to reconcile it with the Administrative Procedure Act or the Constitution. A legion of academics, lower court judges, and Members of this Court-even Auer 's author-has called on us to abandon Auer . Yet today a bare majority flinches, and Auer lives on.

Still, today's decision is more a stay of execution than a pardon. The Court cannot muster even five votes to say that Auer is lawful or wise. Instead, a majority retains Auer only because of stare decisis . And yet, far from standing by that precedent, the majority proceeds to impose so many new and nebulous qualifications and limitations on Auer that THE CHIEF JUSTICE claims to see little practical difference between keeping it on life support in this way and overruling it entirely. So the doctrine emerges maimed and enfeebled-in truth, zombified.

Respectfully, we owe our colleagues on the lower courts more candid and useful guidance than this. And judges owe the people who come before them nothing less than a fair contest, where every party has an equal chance to persuade the court of its interpretation of the law's demands. One can hope that THE CHIEF JUSTICE is right, and that whether we formally overrule Auer or merely neuter it, the results in most cases will prove the same. But means, not just ends, matter, and retaining even this debilitated version of Auer threatens to force litigants and lower courts to jump through needless and perplexing new hoops and in the process deny the people the independent judicial decisions they deserve. All to what end? So that we may pretend to abide stare decisis ?

Consider this case. Mr. Kisor is a Marine who lost out on benefits for post-traumatic stress disorder when the court of appeals deferred to a regulatory interpretation *2426advanced by the Department of Veterans Affairs. The court of appeals was guilty of nothing more than faithfully following Auer . But the majority today invokes stare decisis , of all things, to vacate that judgment and tell the court of appeals to try again using its newly retooled, multi-factored, and far less determinate version of Auer . Respectfully, I would stop this business of making up excuses for judges to abdicate their job of interpreting the law, and simply allow the court of appeals to afford Mr. Kisor its best independent judgment of the law's meaning.

The Court's failure to be done with Auer , and its decision to adorn Auer with so many new and ambiguous limitations, all but guarantees we will have to pass this way again. When that day comes, I hope this Court will find the nerve it lacks today and inter Auer at last. Until then, I hope that our judicial colleagues on other courts will take courage from today's ruling and realize that it has transformed Auer into a paper tiger.

I. How We Got Here

Where did Auer come from? Not from the Constitution, some ancient common law tradition, or even a modern statute. Instead, it began as an unexplained aside in a decision about emergency price controls at the height of the Second World War. Even then, the dictum sat on the shelf, little noticed, for years. Only in the last few decades of the 20th century did lawyers and courts really begin to dust it off and shape it into the reflexive rule of deference to regulatory agencies we know today. And they did so without ever pausing to consider whether a rule like that could be legally justified or even made sense. Auer is really little more than an accident.

A

Before the mid-20th century, few federal agencies engaged in extensive rulemaking, and those that did rarely sought deference for their regulatory interpretations.3 But when the question arose, this Court did not hesitate to say that judges reviewing administrative action should decide all questions of law, including questions concerning the meaning of regulations. As Justice BRANDEIS put it, "[t]he inexorable safeguard which the due process clause assures is ... that there will be opportunity for a court to determine whether the applicable rules of law ... were observed."4 Unsurprisingly, the government's early, longstanding, and consistent interpretation of a statute, regulation, or other legal instrument could count as powerful evidence of its original public meaning.5 But courts respected executive interpretations only because and to the extent "they embodied understandings made roughly contemporaneously with ... enactment and stably maintained and practiced since that time," not "because they were executive as such."6

*2427Writing for four Members of the Court, Justice KAGAN suggests that Auer 's very different approach to the interpretation of agency regulations was foreshadowed as early as this Court's 1898 decision in United States v. Eaton .7 Ante , at ----. But this is mistaken. The question in that case was whether Mr. Eaton's appointment as temporary vice-consul to Siam was consistent with State Department regulations. After several pages of careful and independent legal analysis, the Court held that the regulations did authorize the appointment. That conclusion, the Court explained, was "rendered necessary by a consideration of the text."8 Only after reaching this conclusion did the Court observe that the State Department had previously adopted the same construction, noting along the way that the Department's views were "entitled to the greatest weight" and that the Court saw "no reason in this case to doubt [their] correctness."9 Eaton thus simply followed the well-worn path of acknowledging that an agency's interpretation of a regulation can supply evidence of its meaning.10 Nowhere did the Court even hint that it would have deferred to the State Department's views about the meaning of the law if its own independent textual analysis had not led it to the same conclusion.

All this is borne out by the Court's later teachings in Skidmore v. Swift & Co. in 1944.11 The question there was whether the time overnight employees spent waiting to respond to fire alarms could amount to compensable overtime under the Fair Labor Standards Act. The lower courts had held as a matter of law that it could not. In an opinion by Justice JACKSON, this Court reversed. The Court first held, based on its own independent analysis, that "no principle of law found either in the statute or in Court decisions precludes waiting time from also being working time."12 Only then did the Court consider "what, if any, deference courts should pay" to the views of the Administrator of the Labor Department's Wage and Hour Division.13 And on that question the Court reaffirmed the traditional rule that an agency's interpretation of the law is "not controlling upon the courts" and is entitled only to a weight proportional to "the thoroughness evident in its consideration, the validity of its reasoning, its consistency with earlier and later pronouncements, and all those factors which give it power to persuade."14 At the time, the influential *2428administrative law scholar Kenneth Culp Davis considered this "[a]n entirely reliable statement" of the law.15

B

In truth, the seeds of the Auer doctrine were first planted only in 1945, in Bowles v. Seminole Rock & Sand Co.16 That case involved regulations issued by the Office of Price Administration (OPA), which Congress had tasked with stabilizing the national economy during the Second World War through the use of emergency price controls. It was in that context that the Court declared-for the first time and without citing any authority-that "if the meaning of [the regulation were] in doubt," the agency's interpretation would merit "controlling weight unless it is plainly erroneous or inconsistent with the regulation."17

Yet even then it was far from clear how much weight the Court really placed on the agency's interpretation. As it had in Eaton , the Court in Seminole Rock began with an extended discussion of "the plain words of the regulation," which led it to conclude that the text "clearly" supported the government's position.18 Only after reaching that conclusion based on its own independent analysis did the Court proceed to add that "[a]ny doubts ... are removed by reference to the administrative construction."19

So confused was all this that readers at the time didn't perceive Seminole Rock 's dictum as changing anything. Professor Davis observed that the Court's discussion about giving "controlling weight" to the agency's interpretation was an unexplained aside that made no difference to the case's outcome.20 The dictum, too, was readily explained as reflecting the unusual factual context in which the case arose, involving an emergency government program created to deal with "unique circumstances of war and economic depression."21 And the Court decided Seminole Rock the same Term it issued Skidmore , where it reaffirmed the traditional rule that an agency's views about the law may persuade a court but can never control its judgment. In fact, the Court in Seminole Rock was careful to note that the OPA interpretation before it bore many of the characteristics Skidmore would have recognized as increasing its persuasive force: It had been announced concurrently with the regulation, disseminated widely to the regulated community, and adhered to consistently by the agency.22

*2429No wonder, then, that for many years after the decision, courts "connected Seminole Rock more closely with the deference framework ... under Skidmore " and generally engaged in a Skidmore -type analysis, accepting the agency's interpretation "only after independently examining the regulation and concluding that the agency interpretation was sound."23 If Seminole Rock 's "controlling weight" dictum was afforded any force, it was usually only in the price control context; even then it was ordinarily extended only to "official" agency interpretations that were published contemporaneously with the regulation and widely distributed.24 The Fourth Circuit exemplified the early understanding of Seminole Rock when it observed-citing both Seminole Rock and Skidmore -that "under settled principles" an official agency interpretation in an opinion letter was entitled only to "respectful consideration."25 The letter, the court stressed, did not "have the effect of law," and "[i]t would be absurd to hold that the courts must subordinate their judgment as to the meaning of a ... regulation to the mere unsupported opinion of an associate counsel in an administrative department."26

C

This Court did not cite Seminole Rock 's "controlling weight" dictum again until 1965, in Udall v. Tallman .27 And though Tallman "did very little to advance the jurisprudential understanding of Seminole Rock ," it certainly helped fuel the expansion of so-called " Seminole Rock deference."28 From the 1960s on, this Court and lower courts began to cite the Seminole Rock dictum with increasing frequency and in a wider variety of circumstances, but still without much explanation. They also increasingly divorced Seminole Rock from Skidmore .29

Auer represents the apotheosis of this line of cases. In the name of what some now call the Auer doctrine, courts have in recent years "mechanically applied and reflexively treated" Seminole Rock 's dictum "as a constraint upon the careful inquiry that one might ordinarily expect of courts engaged in textual analysis."30 Under Auer , judges are forced to subordinate their own views about what the law means to those of a political actor, one who may even be a party to the litigation before the court. After all, if the court agrees that the agency's reading is the best one, Auer does no real work; the doctrine matters only when a court would conclude that the agency's interpretation is not the best or fairest reading of the regulation.

To be sure, Justice KAGAN paints a very different picture of Auer , asking us to imagine it riding to the rescue only in cases where the scales of justice are evenly balanced between two equally persuasive readings. But that's a fantasy: "If nature knows of such equipoise in legal arguments, *2430the courts at least do not."31 In the real world the judge uses his traditional interpretive toolkit, full of canons and tiebreaking rules, to reach a decision about the best and fairest reading of the law. Of course, there are close cases and reasonable judges will sometimes disagree. But every day, in courts throughout this country, judges manage with these traditional tools to reach conclusions about the meaning of statutes, rules of procedure, contracts, and the Constitution. Yet when it comes to interpreting federal regulations, Auer displaces this process and requires judges instead to treat the agency's interpretation as controlling even when it is "not ... the best one."32

If that were not troubling enough, Auer has also become "a doctrine of uncertain scope and application."33 This Court has never offered meaningful guidance on how to decide whether the agency's reading is "reasonable" enough to demand judicial deference-and lower courts have drawn that line in wildly different places.34 Deepening the confusion, this Court and lower courts have, over time, tried to soften Auer 's rigidity by declaring that it "might" not apply in some ill-defined circumstances, such as when the agency's interpretation "conflicts with a prior interpretation" or reflects a "convenient litigating position" or a "post hoc rationalization" for past agency action.35 All this has resulted in "widespread confusion" about when and how to apply Auer deference.36

In light of Auer 's many problems, it should come as no surprise that several Members of this Court,37 along with a great many lower court judges38 and *2431members of the legal academy,39 have questioned Auer 's validity and pleaded with this Court to reconsider it.

D

That's where things stood when James Kisor asked the Department of Veterans Affairs to reopen his disability benefits claim. Mr. Kisor served as a United States Marine from 1962 through 1966 and saw combat in Vietnam. In the early 1980s, a VA counselor observed that Mr. Kisor was battling depression and suicidal thoughts and suggested he might be suffering from post-traumatic stress disorder. In light of this, Mr. Kisor filed a claim for disability benefits in 1982. But, in the end, the VA denied the claim.

In 2006, Mr. Kisor sought to reopen the matter. In connection with that request, he presented new evidence, including a psychiatrist's report diagnosing him with PTSD and additional records documenting his service in Vietnam. The VA reopened Mr. Kisor's claim and granted him disability benefits effective June 5, 2006, the date he had submitted his new request. Mr. Kisor argued that a VA regulation entitled him to an earlier effective date for disability benefits, one tracing back to his original submission in 1982. But the Board of Veterans Appeals concluded that the applicable regulation didn't authorize that relief.

Mr. Kisor appealed the Board's ruling all the way to the Federal Circuit, arguing that the Board had misinterpreted the relevant regulation. The Federal Circuit affirmed. Relying on the Auer doctrine, the court held that it had no choice but to treat the Board's interpretation as " 'controlling' " unless that interpretation was " 'plainly erroneous or inconsistent with the regulatio[n].' "40 Without even trying to determine who had the better reading of the regulation, the Board or Mr. Kisor, the court declared that "[t]he Board's interpretation does not strike us as either plainly erroneous or inconsistent with the VA's regulatory framework."41 Case closed.

Mr. Kisor sought and was denied rehearing en banc. Three judges dissented and joined those who have questioned "the logic behind continued adherence to the [ Auer ] doctrine"; they argued that, without Auer deference, Mr. Kisor's reading of the regulation would likely prevail.42 Mr. Kisor then asked us to grant certiorari to reconsider Auer . Thinking it past time to do so, we granted the petition.43

*2432II. The Administrative Procedure Act

When this Court speaks about the rules governing judicial review of federal agency action, we are not (or shouldn't be) writing on a blank slate or exercising some common-law-making power. We are supposed to be applying the Administrative Procedure Act. The APA is a "seminal" statute that Congress wrote to define the relationship between courts and agencies.44 Some have even described it as a kind of constitution for our "administrative state." Yet, remarkably, until today this Court has never made any serious effort to square the Auer doctrine with the APA. Even now, only four Justices make the attempt. And for at least two reasons, their arguments are wholly unpersuasive.

A

The first problem lies in § 706. That provision instructs reviewing courts to "decide all relevant questions of law" and "set aside agency action ... found to be ... not in accordance with law."45 Determining the meaning of a statute or regulation, of course, presents a classic legal question. But in case these directives were not clear enough, the APA further directs courts to "determine the meaning" of any relevant "agency action," including any rule issued by the agency.46 The APA thus requires a reviewing court to resolve for itself any dispute over the proper interpretation of an agency regulation. A court that, in deference to an agency, adopts something other than the best reading of a regulation isn't "decid[ing]" the relevant "questio[n] of law" or "determin[ing] the meaning" of the regulation. Instead, it's allowing the agency to dictate the answer to that question. In doing so, the court is abdicating the duty Congress assigned to it in the APA.47

Justice KAGAN seeks to address the glaring inconsistency between our judge-made rule and the controlling statute this way. On her account, the APA tells a reviewing court to "determine the meaning" of regulations, but it does not tell the court "how " to do that. Thus, we are told, reading the regulation for itself and deferring to the agency's reading are just two equally valid ways for a court to fulfill its statutory duty to "determine the meaning" of the regulation. Ante , at 2419 - 2420.

But the APA isn't as anemic as that. Its unqualified command requires the court to determine legal questions-including questions about a regulation's meaning-by its own lights, not by those of political appointees or bureaucrats who may even be self-interested litigants in the case at hand. Nor can there be any doubt that, when Congress wrote the APA, it knew perfectly well how to require judicial deference to an agency when it wished-in fact, Congress repeatedly specified deferential standards for judicial review elsewhere in the statute.48 But when it comes to the business of *2433interpreting regulations, no such command exists; instead, Congress told courts to "determine" those matters for themselves. Though one hardly needs to be an academic to recognize the point, "commentators in administrative law have 'generally acknowledged' that Section 706 seems to require de novo review on questions of law."49

What the statutory language suggests, experience confirms. If Auer deference were really just another way for courts to "determine the meaning" of regulations under § 706, you might expect that a final judicial "determination" would at least settle, as a matter of precedent, the question of what the regulation "means." Of course, even after one court has spoken on a regulation's meaning, that court or another might properly give weight to a new agency interpretation as part of the court's own decision-making process. See supra , at 6. But in light of National Cable & Telecommunications Assn. v. Brand X Internet Services ,50 courts have interpreted Auer as forbidding a court from ever "determin[ing] the meaning" of a regulation with the force that normally attaches to precedent, because an agency is always free to adopt a different view and insist on judicial deference to its new judgment.51 And if an agency can not only control the court's initial decision but also revoke that decision at any time, how can anyone honestly say the court, rather than the agency, ever really "determine[s]" what the regulation means?

To test the point further, consider a statute that tells a court to "determin[e]" an appropriate sentence in a criminal case.52 If the judge said he was sending a defendant to prison for longer than he believed appropriate only in deference to the government's "reasonable" sentencing recommendation, would anyone really think that complied with the law? Or take a statute that instructs a court to "determine" whether a consent judgment proposed by the government in a civil antitrust *2434case "is in the public interest."53 If a court thought the proposed judgment harmful to the public but decided to defer to the government's "reasonable" contrary view anyway, would anyone suggest the court had complied with Congress's instruction?

Nor does Justice KAGAN's reading of § 706 offer any logical stopping point. If courts can "determine the meaning" of a regulation by deferring to any "reasonable" agency reading, then why not by deferring to any agency reading? If it were really true that the APA has nothing to say about how courts decide what regulations mean, then it would follow that the APA tolerates a rule that "the agency is always right." And if you find yourself in a place as absurd as that, you might want to consider whether you've taken a wrong turn along the way.

B

The problems don't end there. Auer is also incompatible with the APA's instructions in § 553. That provision requires agencies to follow notice-and-comment procedures when issuing or amending legally binding regulations (what the APA calls "substantive rules"), but not when offering mere interpretations of those regulations.54 An agency wishing to adopt or amend a binding regulation thus must publish a proposal in the Federal Register, give interested members of the public an opportunity to submit written comments on the proposal, and consider those comments before issuing the final regulation. Under the APA, that regulation then carries the force of law unless and until it is amended or repealed.55 By contrast, an agency can announce an interpretation of an existing substantive regulation without advance warning and in pretty much whatever form it chooses.

Auer effectively nullifies the distinction Congress drew here. Under Auer , courts must treat as "controlling" not only an agency's duly promulgated rules but also its mere interpretations-even ones that appear only in a legal brief, press release, or guidance document issued without affording the public advance notice or a chance to comment. For all practical purposes, "the new interpretation might as well be a new regulation."56 Auer thus obliterates a distinction Congress thought vital and supplies agencies with a shortcut around the APA's required procedures for issuing and amending substantive rules that bind the public with the full force and effect of law.57

Think of it this way. We've held that the Constitution's specification of a "single, finely wrought" procedure for the enactment of statutes (bicameralism and presentment) necessarily implies that Congress cannot amend an enacted statute without following that procedure-say, by allowing a single House to change what the law requires.58 By the same logic, Congress's specification in the APA of procedures for the creation of new substantive rules (like notice and comment) necessarily implies that an agency cannot amend a substantive *2435rule without following those procedures. To hold otherwise, as Auer demands, subverts the APA's design.

Certain amici contend this argument is "out of place" in this particular case because the VA happened to issue the interpretation challenged here in an adjudicative proceeding.59 But the premise on which they proceed-that the APA permits agencies to issue "controlling" amendments to their regulations in adjudicative proceedings-is not correct. Once an agency issues a substantive rule through notice and comment, it can amend that rule only by following the same notice-and-comment procedures.60 Whether an agency issues its interpretation in a press release or something it chooses to call an "adjudication," all we have is the agency's opinion about what an existing rule means, something that the APA tells us is not binding in a court of law or on the American people.

If that won't work, Justice KAGAN tries an alternative argument from nearly the opposite direction. She replies that affording Auer deference to an agency's interpretation of its own rules never offends the APA because the agency's interpretation lacks "the force of law" associated with substantive rules. Agency interpretations lack this force, we are told, because a court always retains the power to decide at least whether the interpretation is entitled to deference. Ante , at 2420 - 2421. But this argument rests on an implausibly narrow understanding of what it means for an agency action to bear the force of law. Under Justice KAGAN's logic, even a binding substantive rule would lack the force of law because a court retains the power to decide whether the rule is arbitrary and capricious and thus invalid under the APA. But no one believes that. While an agency interpretation, just like a substantive rule, "must meet certain conditions before it gets deference," "once it does so [ Auer makes it] every bit as binding as a substantive rule."61 To suggest that Auer does not make an agency's interpretive guidance "binding o[n] anyone," ante , at 2420 - 2421, is linguistic hocus-pocus.

C

If Auer cannot be squared with the text of the APA, Justice KAGAN suggests it at least conforms to a reasonable "presumption about congressional intent." Ante , at 2412. The theory seems to be that whenever Congress grants an agency "rulemaking power," it also implicitly gives the agency " 'the power authoritatively to interpret' " whatever rules the agency chooses to adopt. Ante , at 2412. But against the clear statutory commands Congress gave us in the APA, what sense does it make to "presume" that Congress really, secretly, wanted courts to treat agency interpretations as binding? Normally, this Court does not allow hidden legislative intentions to "muddy" such plainly expressed statutory directives.62

Even on its own terms, too, this argument proves pretty muddy. It goes something like this: The drafters of the APA did not intend to " 'significantly alter' " established law governing judicial review of agency action as of 1946; the Auer doctrine was part of that established law; therefore, the APA implicitly requires *2436courts to afford agencies Auer deference. Ante , at 2419 - 2420. But neither of this syllogism's essential premises stands on solid ground.

Take the major premise-that those who adopted the APA intended to work no change in the established law of judicial review of agency action. Justice KAGAN is right, of course, that Attorney General Clark claimed as much shortly after the APA's passage. Ante , at 2419 - 2420. But his view, which reflected the interests of the executive branch, was far from universally shared. Others, including many members of Congress, thought the APA would clarify, if not expand, the scope of judicial review. For example, Senator McCarran, the Chairman of the Judiciary Committee, wrote that it would be "hard ... for anyone to argue that this Act did anything other than cut down the 'cult of discretion' so far as federal law is concerned."63 And both the House and Senate reports on the APA said it was intended to "provid[e] that questions of law are for courts rather than agencies to decide in the last analysis."64

Just five years after the APA's passage, this Court seemed to side with those who thought the APA was intended to do more than just summarize existing law. In an opinion by Justice FRANKFURTER, the Court opined that the APA required courts to assume "more responsibility" for reviewing agency decisions "than some courts ha[d] shown in the past."65 One early commentator likewise observed that the APA seemed designed to eliminate all doubt that questions of law "shall be decided by the reviewing Court for itself, and in the exercise of its own independent judgment"; "[m]ore explicit words to impose this mandate," he thought, "could hardly be found."66

Justice KAGAN's syllogism runs into even more trouble with its minor premise-that the Auer doctrine was a well-established part of the common law background when Congress enacted the APA in 1946. As we've seen, this Court planted the seeds of Auer deference for the first time in dictum in Seminole Rock , just a year before Congress passed the APA. See Part I-B, supra . And that dictum did not somehow immediately become an entrenched part of the common law: For years following Seminole Rock , courts and "commentators largely ignored" it,67 and those who took notice weren't sure what to make of it. Professor Davis, for example, doubted that the dictum could be "taken at face value" given that it seemed "irreconcilable" with the Court's approach in other cases.68 In truth, when Congress passed the APA the law of judicial review of agency action was in a confused state. During *2437the congressional hearings on the bill, one witness's suggestion that Congress should leave the scope of judicial review "as it now is" drew this fair reply from Representative Walter, chairman of the House Subcommittee on Administrative Law and author of the House Report on the APA: "You say 'as it now is.' Frankly, I do not know what it now is .... [T]he Supreme Court apparently changes its mind daily."69

III. The Constitution

Not only is Auer incompatible with the APA; it also sits uneasily with the Constitution. Article III, § 1 provides that the "judicial Power of the United States" is vested exclusively in this Court and the lower federal courts. A core component of that judicial power is " 'the duty of interpreting [the laws] and applying them in cases properly brought before the courts.' "70 As Chief Justice MARSHALL put it, "[i]t is emphatically the province and duty of the judicial department to say what the law is."71 And never, this Court has warned, should the "judicial power ... be shared with [the] Executive Branch."72 Yet that seems to be exactly what Auer requires.

A

Our Nation's founders were painfully aware of the dangers of executive and legislative intrusion on judicial decision-making. One of the abuses of royal power that led to the American Revolution was King George's attempt to gain influence over colonial judges.73 Colonial legislatures, too, had interfered with the courts' independence "at the behest of private interests and factions."74 These experiences had taught the founders that " 'there is no liberty if the power of judgment be not separated from the legislative and executive powers.' "75 They knew that when political actors are left free not only to adopt and enforce written laws, but also to control the interpretation of those laws, the legal rights of "litigants with unpopular or minority causes or ... who belong to despised or suspect classes" count for little.76

*2438Maybe the powerful, well-heeled, popular, and connected can wheedle favorable outcomes from a system like that-but what about everyone else? They are left always a little unsure what the law is, at the mercy of political actors and the shifting winds of popular opinion, and without the chance for a fair hearing before a neutral judge. The rule of law begins to bleed into the rule of men.

Experiencing all this in their own time, the founders sought to ensure that those who came after them would not. Believing that "[n]o maxim was better established" than "that the power of making ought to be kept distinct from that of expounding, the laws,"77 they designed a judiciary that would be able to interpret the laws "free from potential domination by other branches of government."78 To that end, they resisted proposals that would have subjected judicial decisions to review by political actors.79 And they rejected the British tradition of using the upper house of the legislature as a court of last resort, out of fear that a body with "even a partial agency in passing bad laws" would operate under the "same spirit" in "interpreting them."80 Instead, they gave federal judges life tenure, subject only to removal by impeachment; and they guaranteed that the other branches could not reduce judges' compensation so long as they remained in office.

The founders afforded these extraordinary powers and protections not for the comfort of judges, but so that an independent judiciary could better guard the people from the arbitrary use of governmental power. And sitting atop the judicial branch, this Court has always carried a special duty to "jealously guar[d]" the Constitution's promise of judicial independence.81 So we have long resisted any effort by the other branches to " 'usurp a court's power to interpret and apply the law to the circumstances before it.' "82 The judicial power to interpret the law, this Court has held, "can no more be shared with another branch than the Chief Executive, for example, can share with the Judiciary the veto power, or the Congress share with the Judiciary the power to override a Presidential veto."83

Auer represents no trivial threat to these foundational principles. Under the APA, substantive rules issued by federal agencies through notice-and-comment procedures bear "the 'force and effect of law' "84 and are part of the body of federal *2439law, binding on private individuals, that the Constitution charges federal judges with interpreting. Yet Auer tells the judge that he must interpret these binding laws to mean not what he thinks they mean, but what an executive agency says they mean. Unlike Article III judges, executive officials are not, nor are they supposed to be, "wholly impartial."85 They have their own interests, their own constituencies, and their own policy goals-and when interpreting a regulation, they may choose to "press the case for the side [they] represen[t]" instead of adopting the fairest and best reading.86 Auer thus means that, far from being "kept distinct," the powers of making, enforcing, and interpreting laws are united in the same hands-and in the process a cornerstone of the rule of law is compromised.

Consider an analogy. The Court has long held that Congress cannot " 'indirectly control the action of the courts, by requiring of them a construction of the law according to its own views.' "87 If Congress disagrees with how courts are interpreting an existing statute, it is free to amend the statute to establish a different rule going forward. What it cannot do is issue "a mandate ... to compel the courts to construe and apply [existing law], not according to the judicial, but according to the legislative judgment."88 As early as 1804, when a lawyer argued before this Court that an Act of the North Carolina legislature could not control the Court's construction of an earlier North Carolina statute because "[t]o declare what the law is, or has been, is a judicial power," not a legislative power, the Court stopped him, deeming the point too plain for argument.89

But if the legislature can't control a judge's interpretation of an existing statute, how can an executive agency control a judge's interpretation of an existing and equally binding regulation? Auer allows an agency to do exactly what this Court has always said a legislature cannot do: "compel the courts to construe and apply" a law on the books, "not according to the judicial ... judgment," but according to the judgment of another branch.90 When we defer to an agency interpretation that differs from what we believe to be the best interpretation of the law, we compromise our judicial independence and deny the people who come before us the impartial judgment that the Constitution guarantees them. And we mislead those whom we serve by placing a judicial imprimatur on what is, in fact, no more than an exercise of raw political executive power.91

B

What do our colleagues have to say about these concerns? A majority has *2440nothing to offer, and Justice KAGAN dismisses them out of hand. In fact, she barely mentions the Constitution, other than to assure us that Auer does not allow agencies to "usur[p] the interpretive role of courts" because "courts retain a firm grip on the interpretive function" through their ability to decide whether Auer deference applies. Ante , at 2421. But that is no assurance at all. The judicial power has always been understood to provide the people with a neutral arbiter who bears the responsibility and duty to "expound and interpret" the governing law, not just the power to say whether someone else's interpretation, let alone the interpretation of a self-interested political actor, is "reasonable."92

To be sure, it's conceivable that Congress might seek to limit the ability of judges to remedy an adverse agency action. It might, for example, provide that a court shall have power to set aside agency action pursuant to a regulation only if the action was based on an unreasonable interpretation of the regulation. But even assuming the constitutionality of a hypothetical statute like that, Auer is different. It does not limit the scope of the judicial power; instead, it seeks to coopt the judicial power by requiring an Article III judge to decide a case before him according to principles that he believes do not accurately reflect the law. Under Auer , a judge is required to lay aside his independent judgment and declare affirmatively that a regulation means what the agency says it means-and, thus, that the law is what the agency says it is. Then the judge is compelled to exercise his judicial authority to adjust private rights and obligations based on the agency's (mis)understanding of the law. If Auer were a statute, it would not be an exercise of Congress's "power (within limits) to tell the courts what classes of cases they may decide," or what relief they may supply, but a forbidden attempt "to prescribe or superintend how they decide those cases."93 And in the absence of any statute like that, this Court surely should not so freely give away to the executive branch its assigned responsibility to interpret the laws. "Abdication of responsibility is not part of the constitutional design."94

In the end, Justice KAGAN's only real reply is this: However misguided it may be to hand over our interpretive powers to executive agencies, at least there isn't a mountain of empirical evidence showing that agencies have used this power to deliberately write "vague and open-ended" regulations to maximize their interpretive leeway. Ante , at 2421. But even this misses the point. Whether or not regulations are " 'designed' " to be vague, ibid. , many can be read in different ways, especially when new and unanticipated applications arise; cases like that come before the courts all the time. Without Auer 's shadow hanging over them, parties would receive a fair hearing before an impartial judge. The agency's interpretation would sometimes be rejected; and that, in turn, might lead it to solicit public comment on possible amendments to the regulation, which would provide an opportunity for public input that might produce better policy. But with Auer , there is no fair hearing and no need for the agency to amend the regulation through notice and comment. Whether purposeful or not, the agency's failure to write a clear regulation winds up increasing *2441its power, allowing it to both write and interpret rules that bear the force of law-in the process uniting powers the Constitution deliberately separated and denying the people their right to an independent judicial determination of the law's meaning.

IV. Policy Arguments

Lacking support elsewhere, Justice KAGAN is forced to resort to policy arguments to defend Auer . But even the most sensible policy argument would not empower us to ignore the plain language of the APA or the demands of the Constitution. And as we've seen, those documents reflect a very different "policy" judgment by the people and their representatives. Besides, the policy arguments offered today are not just unpersuasive, they are troubling.

Take the first and boldest offering. Justice KAGAN suggests that determining the meaning of a regulation is largely a matter of figuring out what the "person who wrote it ... intended." Ante , at 2412. In this way, we're told, a legally binding regulation isn't all that different from "a memo or an e-mail"-if you "[w]ant to know what [it] means," you'd better "[a]sk its author." Ante , at 2412 - 2413. But the federal government's substantive rules are not like memos or e-mails; they are binding edicts that carry the force of law for all citizens. And if the rule of law means anything, it means that we are governed by the public meaning of the words found in statutes and regulations, not by their authors' private intentions. This is a vital part of what it means to have "a government of laws, and not of men."95 When judges interpret a regulation, what we are trying to get at, as Justice HOLMES explained long ago, is not the "particular intent" of those who wrote it, but "what [its] words would mean [to] a normal speaker of English ... in the circumstances in which they were used."96 If the best reading of the regulation turns out to be something other than what the agency claims to have intended, the agency is free to rewrite the regulation; but its secret intentions are not the law.

Nor does Justice KAGAN's account of the interpretive process even wind up supporting Auer . If a court's goal in interpreting a regulation really were to determine what its author "intended," Auer would be an almost complete mismatch with the goal. Agency personnel change over time, and an agency's policy priorities may shift dramatically from one presidential administration to another. Yet Auer tells courts that they must defer to the agency's current view of what the regulation ought to mean, which may or may not correspond to the views of those who actually wrote it. If interpreting a regulation really were just like reading an e-mail, Auer would be like seeking guidance about the e-mail's meaning, years or decades later, from the latest user of the computer from which the e-mail was sent. We've repeatedly rejected that approach in the context of statutory interpretation. While Members of this Court sometimes disagree about the usefulness of pre-enactment legislative history, we all agree that legislators' statements about the meaning of an already-enacted statute are not "a legitimate tool of statutory interpretation,' " much less a *2442controlling one.97 So why on earth would we give "controlling weight" to an agency's statements about the meaning of an already-promulgated regulation?

Proceeding farther down this doubtful path, Justice KAGAN asserts that resolving ambiguities in a regulation "sounds more in policy than in law" and is thus a task more suited to executive officials than judges. Ante , at 2413. But this claim, too, contradicts a basic premise of our legal order: that we are governed not by the shifting whims of politicians and bureaucrats, but by written laws whose meaning is fixed and ascertainable-if not by all members of the public, then at least by lawyers who can advise them and judges who must apply the law to individual cases guided by the neutral principles found in our traditional tools of interpretation. The text of the regulation is treated as the law, and the agency's policy judgment has the force of law only insofar as it is embodied in the regulatory text. If "new issues demanding new policy calls" arise that aren't addressed in existing regulations, ante , at 2413, the solution is for the agency to promulgate new regulations using the notice-and-comment procedures set forth in the APA. But an agency has no warrant to compel judges to change the law to conform with the agency's current policy preferences.

To be sure, during the period of Auer 's ascendancy some suggested that the meaning of written law is always "radically indeterminate" and that judges expounding it are "for the most part, guided by policy-not text."98 And in an environment like that it was perhaps thought a small step to conclude that, if legal disputes are going to be resolved on political grounds, then they ought to be resolved by real politicians in the executive branch rather than ersatz politicians on the bench. But the proposed cure proved worse than the disease. Arguments like these surrendered the judgment embodied in our Constitution and the APA that courts owe the people they serve their independent legal judgment about the law's meaning. Besides, we've long since come to realize that the real cure doesn't lie in turning judges into rubber stamps for politicians, but in redirecting the judge's interpretive task back to its roots, away from open-ended policy appeals and speculation about legislative intentions and toward the traditional tools of interpretation judges have employed for centuries to elucidate the law's original public meaning. Today it is even said that we judges are, to one degree or another, "all textualists now."99

Pursuing a more modest tack, Justice KAGAN next suggests that Auer is justified by the respect due agencies' "technical" expertise. Ante , at 2413 - 2414. But no one doubts that courts should pay close attention to an expert agency's views on technical questions in its field. Just as a court "would want to know what John Henry Wigmore said about an issue of evidence law [or] what Arthur Corbin thought about a matter of contract law," so too should courts carefully consider what the Food and Drug Administration thinks about how its prescription drug safety regulations operate.100 The fact remains, however, *2443that even agency experts "can be wrong; even Homer nodded."101 Skidmore and the traditional approach it embodied recognized both of these facts of life long ago, explaining that, while courts should of course afford respectful consideration to the expert agency's views, they must remain open to competing expert and other evidence supplied in an adversarial setting. Respect for an agency's technical expertise demands no more.

Justice KAGAN's final policy argument is that Auer promotes "consistency" and "uniformity" in the interpretation of regulations. Ante , at 2413 - 2414. If we let courts decide what regulations mean, she warns, they might disagree, and it might take some time for higher courts to resolve those disagreements. But consistency and uniformity are hardly grounds on which Auer 's advocates should wish to fight. The judicial process is how we settle disputes about the meaning of written law, and our judicial system is more than capable of producing a single, uniform, and stable interpretation that will last until the regulation is amended or repealed. Meanwhile, under Auer courts often disagree about whether deference is warranted, see supra , at 2430 - 2431, and a regulation's "meaning" can be transformed with the stroke of a pen any time there is a new presidential administration. "Consistency," "uniformity," and stability in the law are hardly among Auer 's crowning achievements.

V. Stare Decisis

In the end, a majority declines to endorse Justice KAGAN's arguments and insists only that, even if Auer is not "right and well-reasoned," we're stuck with it because of the respect due precedent. Ante , at 2423.

But notice: While pretending to bow to stare decisis , the majority goes about reshaping our precedent in new and experimental ways. True, the majority admits, this Court has in the past accorded Auer deference " 'reflexive[ly],' " "without significant analysis of the underlying regulation" or "careful attention to [its] nature and context," and encouraged lower courts to do the same. Ante , at 2414. But no more. From now on, the majority says, not only must judges "exhaust all the 'traditional tools' of construction" to decide whether the agency's interpretation is "reasonable," they must also make "an independent inquiry into whether the character and context of the agency interpretation" justifies deference. Ante , at 2416. The majority candidly admits that it finds it impossible to "reduce" this new inquiry "to any exhaustive test," so it settles for laying out some "markers." Ante , at 2416 - 2417. What are the markers? We are told that courts should often-but not always-withhold deference from an interpretation offered by mid-level agency staff; often-but not always-withhold deference from a nontechnical, "prosaic-seeming" interpretation; often-but not always-withhold deference from an interpretation advanced for the first time in an amicus brief; and often-but not always-withhold deference from an interpretation that conflicts with an earlier one. See ante , at 2416 - 2418. The only certainty in all this is that the majority isn't really much moved by stare decisis ; everyone recognizes, to one degree or another, that Auer cannot stand. And between our remaining choices-continuing to make up new deference rules, or returning to the text of the APA and the approach to judicial review that prevailed for most of our history-the answer should have been easy.

A

There are serious questions about whether stare decisis should apply here at *2444all. To be sure, Auer 's narrow holding about the meaning of the regulation at issue in that case may be entitled to stare decisis effect. The same may be true for the specific holdings in other cases where this Court has applied Auer deference. But does stare decisis extend beyond those discrete holdings and bind future Members of this Court to apply Auer 's broader deference framework?

It seems doubtful that stare decisis demands that much. We are not dealing with a precedent that purported to settle the meaning of a single statute or regulation or resolve a particular case. The Auer doctrine claims to do much more than that-to prescribe an interpretive methodology governing every future dispute over the meaning of every regulation. In other contexts, we do not regard statements in our opinions about such generally applicable interpretive methods, like the proper weight to afford historical practice in constitutional cases or legislative history in statutory cases, as binding future Justices with the full force of horizontal stare decisis .102 Why, then, should we regard as binding Auer 's statements about the weight to afford agencies' interpretations in regulatory cases? To the extent Auer purports to dictate "the interpretive inferences that future Justices must draw in construing statutes and regulations that the Court has never engaged," it may well "exceed the limits of stare decisis."103

Even if our past expressions of support for Auer deference bear some precedential force, they certainly are not entitled (as the majority suggests, ante , at 2422 - 2423) to the special, heightened form of stare decisis we reserve for narrow statutory decisions. In contrast to precedents that fix the meaning of particular statutes and generate reliance interests in the process, the Auer doctrine is an abstract default rule of interpretive methodology that settles nothing of its own force. And this Court has recognized that it is "inconsistent with the Court's proper role" to insist that Congress exercise its legislative power to overturn such erroneous and judicially invented "default rule[s]."104 That should be especially so here because Auer 's default rule undermines judicial independence, which this Court has a special responsibility to defend.

Nor is it entirely clear that Congress could overturn the Auer doctrine legislatively. The majority describes Auer as a "presumption" about how courts should interpret statutes granting rulemaking power to agencies. Ante , at 2414 -1215. Congress can, of course, rebut the presumption on a statute-by-statute basis, or even for all past statutes. But can Congress eliminate the Auer presumption for future statutes? Perhaps-but legislation like that would raise questions, which the majority does not address, about the ability of one Congress to entrench its preferences by attempting to control the interpretation of legislation enacted by future Congresses. 105

*2445We should not be in the business of tossing " 'balls ... into Congress's court,' " ante , at 2422, that would explode with constitutional questions if Congress tried to pick them up.

B

Even assuming for argument's sake that standard stare decisis considerations apply, they still do not require us to retain Auer . Even the majority implicitly recognizes this much, as it proceeds to vacate a lower court judgment that faithfully applied Auer and instruct that court to try again using the majority's new directions. If stare decisis allows us so freely to remodel Auer , it's hard to see on what account it might require us to retain it.

We do not lightly overturn precedents, and we seek always to honor the thoughtful guidance of those who have preceded us. At the same time, everyone agrees that stare decisis is not an " 'inexorable command,' "106 and this Court should not always remain bound to decisions whose "rationale no longer withstands 'careful analysis.' "107 Recognizing the need for balance in this area, the Court has, over time, fashioned principles to guide our treatment of precedent. Those principles call on us to consider factors such as "the quality of [the precedent's] reasoning, the workability of the rule it established, its consistency with other related decisions, developments since the decision was handed down, and reliance on the decision."108 As applied to Auer , all of these considerations weigh strongly in favor of bidding farewell to the doctrine rather than keeping it on life support.

First , we've already seen that no persuasive rationale supports Auer . From its humble origins as an unexplained bit of dictum in a wartime case about emergency price controls, the Auer doctrine evolved into a rigid rule of deference-all without any serious attempt by this Court to rationalize it or reconcile it with the APA, the Constitution, or traditional modes of judicial review. See Part I, supra . Even its fiercest defenders acknowledge that " Auer deference has not remained static over time" and urge the Court to continue to "shape" and "refin[e]" the doctrine.109 Today's decision attempts just such a "refinement" by hedging Auer with new qualifications and limitations. See ante , at 2414 - 2418. This shifting ground "undermin[es] the force of stare decisis ."110

Second , today's ruling all but admits that Auer has not proved to be a workable standard. Even before this latest overhaul, uncertainty surrounding Auer 's scope and application had caused many to question whether there was any "practical benefit" in continuing to apply Auer "rather than a less deferential but more flexible and *2446open-ended standard like Skidmore ."111 See supra , at 2430 - 2431. Nor does the majority's kinder, gentler version of Auer promise to solve the problem. On the contrary, its newly mandated inquiry into the "character and context of the agency interpretation," which it admits cannot be reduced "to any exhaustive test," ante , at 2416, seems destined only to compound the confusion. See supra , at 2444 - 2445. Many words come to mind to describe the tasks we assign lower court judges today, but "workable" is not among them.

Third , the Auer doctrine is, as we have also already seen, out of step with how courts normally interpret written laws. When we interpret a regulation, we typically (at least when there is no agency say-so) proceed in the same way we would when interpreting any other written law: We "begin our interpretation of the regulation with its text" and, if the text is unclear, we "turn to other canons of interpretation" and tie-breaking rules to resolve the ambiguity.112 And when we interpret an ambiguous statute , we never ask what current members of Congress think it means; in fact, we've held unanimously that legislators' post-enactment views about a statute's meaning are not even a " 'legitimate tool of statutory interpretation.' "113 Affording "controlling weight" to regulators' post-promulgation views about the meaning of an ambiguous regulation is hard to square with these usual judicial practices.114

Fourth , the explosive growth of the administrative state over the last half-century has exacerbated Auer 's potential for mischief. When the Court first uttered its dictum in Seminole Rock , the administrative state was new and the APA was only a gleam in Congress's eye. Even 20 years later, when the Court began reviving the Seminole Rock dictum and turning it into a new deference doctrine, it was not yet apparent how pervasive the administrative state would become in the lives of ordinary Americans. Now, in the 21st century, "[t]he administrative state wields vast power and touches almost every aspect of daily life."115 Among other things, it produces " 'reams of regulations' "116 -so *2447many that they dwarf the statutes enacted by Congress. As of 2018, the Code of Federal Regulations filled 242 volumes and was about 185,000 pages long, almost quadruple the length of the most recent edition of the U. S. Code.117 And agencies add thousands more pages of regulations every year. Whether you think this administrative fecundity is a good or a bad thing, it surely means that the cost of continuing to deny citizens an impartial judicial hearing on the meaning of disputed regulations has increased dramatically since this Court started down this road.

Fifth , Auer has generated no serious reliance interests. The only parties that might have relied on Auer 's promise of deference are agencies that use post hoc interpretations to bypass the APA's notice-and-comment procedures. But this Court has never suggested that the convenience of government officials should count in the balance of stare decisis , especially when weighed against the interests of citizens in a fair hearing before an independent judge and a stable and knowable set of laws. In short, " '[t]he fact that [agencies] may view [ Auer deference] as an entitlement does not establish the sort of reliance interest that could outweigh the countervailing interest' " of all citizens " 'in having their constitutional rights fully protected.' "118

Coming closer to the mark, the majority worries that "abandoning Auer deference would cast doubt on many settled constructions" of regulations on which regulated parties might have relied. Ante , at 2406. But, again, decisions construing particular regulations might retain stare decisis effect even if the Court announced that it would no longer adhere to Auer 's interpretive methodology. After all, decisions construing particular statutes continue to command respect even when the interpretive methods that led to those constructions fall out of favor. Besides, if the majority is correct that abandoning Auer would require revisiting regulatory constructions that were upheld based on Auer deference, the majority's revision of Auer will yield exactly the same result. There are innumerable lower court decisions that have followed this Court's lead and afforded Auer deference mechanically, without conducting the inquiry the Court now holds is required. Today's ruling casts no less doubt on the continuing validity of those decisions than we would if we simply moved on from Auer .

*

Overruling Auer would have taken us directly back to Skidmore , liberating courts to decide cases based on their independent judgment and "follow [the] agency's [view] only to the extent it is persuasive."119 By contrast, the majority's attempt to remodel Auer 's rule into a multi-step, multi-factor inquiry guarantees more uncertainty and much litigation. Proceeding in this convoluted way burdens our colleagues on the lower courts, who will have to spend time debating *2448deference that they could have spent interpreting disputed regulations. It also continues to deny the people who come before us the neutral forum for their disputes that they rightly expect and deserve.

But this cloud may have a silver lining: The majority leaves Auer so riddled with holes that, when all is said and done, courts may find that it does not constrain their independent judgment any more than Skidmore . As reengineered, Auer requires courts to "exhaust all the 'traditional tools' of construction" before they even consider deferring to an agency. Ante , at 2415 - 2416. And those tools include all sorts of tie-breaking rules for resolving ambiguity even in the closest cases. Courts manage to make do with these tools in many other areas of the law, so one might hope they will hardly ever find them inadequate here. And if they do, they will now have to conduct a further inquiry that includes so few firm guides and so many cryptic "markers" that they will rarely, if ever, have to defer to an agency regulatory interpretation that differs from what they believe is the best and fairest reading.

But whatever happens, this case hardly promises to be this Court's last word on Auer . If today's opinion ends up reducing Auer to the role of a tin god-officious, but ultimately powerless-then a future Court should candidly admit as much and stop requiring litigants and lower courts to pay token homage to it. Alternatively, if Auer proves more resilient, this Court should reassert its responsibility to say what the law is and afford the people the neutral forum for their disputes that they expect and deserve.

Justice KAVANAUGH, with whom Justice ALITO joins, concurring in the judgment.

I agree with Justice GORSUCH's conclusion that the Auer deference doctrine should be formally retired. I write separately to emphasize two points.

First , I agree with THE CHIEF JUSTICE that "the distance between the majority and Justice GORSUCH is not as great as it may initially appear." Ante, at 2424 (opinion concurring in part). The majority's approach in Part II-B of its opinion closely resembles the argument advanced by the Solicitor General to "clarif[y] and narro[w]" Auer . Brief for Respondent 15. Importantly, the majority borrows from footnote 9 of this Court's opinion in Chevron to say that a reviewing court must "exhaust all the 'traditional tools' of construction" before concluding that an agency rule is ambiguous and deferring to an agency's reasonable interpretation. Ante, at 2443 (quoting Chevron U. S. A. Inc. v. Natural Resources Defense Council, Inc. , 467 U.S. 837, 843, n. 9, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984) ). If a reviewing court employs all of the traditional tools of construction, the court will almost always reach a conclusion about the best interpretation of the regulation at issue. After doing so, the court then will have no need to adopt or defer to an agency's contrary interpretation. In other words, the footnote 9 principle, taken seriously, means that courts will have no reason or basis to put a thumb on the scale in favor of an agency when courts interpret agency regulations.

Formally rejecting Auer would have been a more direct approach, but rigorously applying footnote 9 should lead in most cases to the same general destination. Umpires in games at Wrigley Field do not defer to the Cubs manager's in-game interpretation of Wrigley's ground rules. So too here.

To be sure, some cases involve regulations that employ broad and open-ended terms like "reasonable," "appropriate," "feasible," or "practicable." Those kinds of terms afford agencies broad policy discretion, *2449and courts allow an agency to reasonably exercise its discretion to choose among the options allowed by the text of the rule. But that is more State Farm than Auer . See Motor Vehicle Mfrs. Assn. of United States, Inc. v. State Farm Mut. Automobile Ins. Co. , 463 U.S. 29, 103 S.Ct. 2856, 77 L.Ed.2d 443 (1983).

In short, after today's decision, a judge should engage in appropriately rigorous scrutiny of an agency's interpretation of a regulation, and can simultaneously be appropriately deferential to an agency's reasonable policy choices within the discretion allowed by a regulation.

Second , I also agree with THE CHIEF JUSTICE that "[i]ssues surrounding judicial deference to agency interpretations of their own regulations are distinct from those raised in connection with judicial deference to agency interpretations of statutes enacted by Congress." Ante, at 2425. Like THE CHIEF JUSTICE, "I do not regard the Court's decision" not to formally overrule Auer " to touch upon the latter question." Ibid.

Justice White

delivered the opinion of the Court.

The development of the automobile gave Americans unprecedented freedom to travel, but exacted a high price for *33enhanced mobility. Since 1929, motor vehicles have been the leading cause of accidental deaths and injuries in the United States. In 1982, 46,300 Americans died in motor vehicle accidents and hundreds of thousands more were maimed and injured.1 While a consensus exists that the current loss of life on our highways is unacceptably high, improving safety does not admit to easy solution. In 1966, Congress decided that at least part of the answer lies in improving the design and safety features of the vehicle itself.2 But much of the technology for building safer cars was undeveloped or untested. Before changes in automobile design could be mandated, the effectiveness of these changes had to be studied, their costs examined, and public acceptance considered. This task called for considerable expertise and Congress responded by enacting the National Traffic and Motor Vehicle Safety Act of 1966 (Act), 80 Stat. 718, as amended, 15 U. S. C. § 1381 et seq. (1976 ed. and Supp. V). The Act, created for the purpose of “reducing] traffic accidents and deaths and injuries to persons resulting from traffic accidents,” 15 U. S. C. § 1381, directs the Secretary of Transportation or his delegate to issue motor vehicle safety standards that “shall be practicable, shall meet the need for motor vehicle safety, and shall be stated in objective terms.” 15 U. S. C. § 1392(a) (1976 ed., Supp. V). In issuing these standards, the Secretary is directed to consider “relevant available motor vehicle safety data,” whether the proposed standard “is reasonable, practicable and appropriate” for the particular type of motor vehicle, and the “extent to which *34such standards will contribute to carrying out the purposes” of the Act. 15 U. S. C. §§ 1392(f)(1), (3), (4).3

The Act also authorizes judicial review under the provisions of the Administrative Procedure Act (APA), 5 U. S. C. §706, of all “orders establishing, amending, or revoking a Federal motor vehicle safety standard,” 15 U. S. C. § 1392(b). Under this authority, we review today whether NHTSA acted arbitrarily and capriciously in revoking the requirement in Motor Vehicle Safety Standard 208 that new motor vehicles produced after September 1982 be equipped with passive restraints to protect the safety of the occupants of the vehicle in the event of a collision. Briefly summarized, we hold that the agency failed to present an adequate basis and explanation for rescinding the passive restraint requirement and that the agency must either consider the matter further or adhere to or amend Standard 208 along lines which its analysis supports. •

I

The regulation whose rescission is at issue bears a complex and convoluted history. Over the course of approximately 60 rulemaking notices, the requirement has been imposed, amended, rescinded, reimposed, and now rescinded again.

As originally issued by the Department of Transportation in 1967, Standard 208 simply required the installation of seatbelts in all automobiles. 32 Fed. Reg. 2415. It soon became apparent that the level of seatbelt use was too low to reduce traffic injuries to an acceptable level. The Department therefore began consideration of “passive occupant restraint systems” — devices that do not depend for their effec*35tiveness upon any action taken by the occupant except that necessary to operate the vehicle. Two types of automatic crash protection emerged: automatic seatbelts and airbags. The automatic seatbelt is a traditional safety belt, which when fastened to the interior of the door remains attached without impeding entry or exit from the vehicle, and deploys automatically without any action on the part of the passenger. The airbag is an inflatable device concealed in the dashboard and steering column. It automatically inflates when a sensor indicates that deceleration forces from an accident have exceeded a preset minimum, then rapidly deflates to dissipate those forces. The lifesaving potential of these devices was immediately recognized, and in 1977, after substantial on-the-road experience with both devices, it was estimated by NHTSA that passive restraints could prevent approximately 12,000 deaths and over 100,000 serious injuries annually. 42 Fed. Reg. 34298.

In 1969, the Department formally proposed a standard requiring the installation of passive restraints, 34 Fed. Reg. 11148, thereby commencing a lengthy series of proceedings. In 1970, the agency revised Standard 208 to include passive protection requirements, 35 Fed. Reg. 16927, and in 1972, the agency amended the Standard to require full passive protection for all front seat occupants of vehicles manufactured after August 15, 1975. 37 Fed. Reg. 3911. In the interim, vehicles built between August 1973 and August 1975 were to carry either passive restraints or lap and shoulder belts coupled with an “ignition interlock” that would prevent starting the vehicle if the belts were not connected.4 On review, the *36agency’s decision to require passive restraints was found to be supported by “substantial evidence” and upheld. Chrysler Corp. v. Department of Transportation, 472 F. 2d 659 (CA6 1972).5

In preparing for the upcoming model year, most car makers chose the “ignition interlock” option, a decision which was highly unpopular, and led Congress to amend the Act to prohibit a motor vehicle safety standard from requiring or permitting compliance by means of an ignition interlock or a continuous buzzer designed to indicate that safety belts were not in use. Motor Vehicle and Schoolbus Safety Amendments of 1974, Pub. L. 93-492, §109, 88 Stat. 1482, 15 U. S. C. § 1410b(b). The 1974 Amendments also provided that any safety standard that could be satisfied by a system other than seatbelts would have to be submitted to Congress where it could be vetoed by concurrent resolution of both Houses. 15 U. S. C. § 1410b(b)(2).6

The effective date for mandatory passive restraint systems was extended for a year until August 31,1976. 40 Fed. Reg. 16217 (1975); id., at 33977. But in June 1976, Secretary of Transportation William T. Coleman, Jr., initiated a new rulemaking on the issue, 41 Fed. Reg. 24070. After hearing testimony and reviewing written comments, Coleman extended the optional alternatives indefinitely and suspended the passive restraint requirement. Although he found pas*37sive restraints technologically and economically feasible, the Secretary based his decision on the expectation that there would be widespread public resistance to the new systems. He instead proposed a demonstration project involving up to 500,000 cars installed with passive restraints, in order to smooth the way for public acceptance of mandatory passive restraints at a later date. Department of Transportation, The Secretary’s Decision Concerning Motor Vehicle Occupant Crash Protection (Dec. 6, 1976), App. 2068.

Coleman’s successor as Secretary of Transportation disagreed. Within months of assuming office, Secretary Brock Adams decided that the demonstration project was unnecessary. He issued a new mandatory passive restraint regulation, known as Modified Standard 208. 42 Fed. Reg. 34289 (1977); 49 CFR § 571.208 (1978). The Modified Standard mandated the phasing in of passive restraints beginning with large cars in model year 1982 and extending to all cars by model year 1984. The two principal systems that would satisfy the Standard were airbags and passive belts; the choice of which system to install was left to the manufacturers. In Pacific Legal Foundation v. Department of Transportation, 193 U. S. App. D. C. 184, 593 F. 2d 1338, cert. denied, 444 U. S. 830 (1979), the Court of Appeals upheld Modified Standard 208 as a rational, nonarbitrary regulation consistent with the agency’s mandate under the Act. The Standard also survived scrutiny by Congress, which did not exercise its authority under the legislative veto provision of the 1974 Amendments.7

Over the next several years, the automobile industry geared up to comply with Modified Standard 208. As late as July 1980, NHTSA reported:

*38“On the road experience in thousands of vehicles equipped with air bags and automatic safety belts has confirmed agency estimates of the life-saving and injury-preventing benefits of such systems. When all cars are equipped with automatic crash protection systems, each year an estimated 9,000 more lives will be saved, and tens of thousands of serious injuries will be prevented.” NHTSA, Automobile Occupant Crash Protection, Progress Report No. 3, p. 4; App. in No. 81-2220 (CADC), p. 1627 (hereinafter App.).

In February 1981, however, Secretary of Transportation Andrew Lewis reopened the rulemaking due to changed economic circumstances and, in particular, the difficulties of the automobile industry. 46 Fed. Reg. 12033. Two months later, the agency ordered a one-year delay in the application of the Standard to large cars, extending the deadline to September 1982, id., at 21172, and at the same time, proposed the possible rescission of the entire Standard. Id., at 21205. After receiving written comments and holding public hearings, NHTSA issued a final rule (Notice 25) that rescinded the passive restraint requirement contained in Modified Standard 208.

II

In a statement explaining the rescission, NHTSA maintained that it was no longer able to find, as it had in 1977, that the automatic restraint requirement would produce significant safety benefits. Notice 25, id., at 53419. This judgment reflected not a change of opinion on the effectiveness of the technology, but a change in plans by the automobile industry. In 1977, the agency had assumed that airbags would be installed in 60% of all new cars and automatic seatbelts in 40%. By 1981 it became apparent that automobile manufacturers planned to install the automatic seatbelts in approximately 99% of the new cars. For this reason, the lifesaving potential of airbags would not be realized. Moreover, it now appeared that the overwhelming majority of passive belts *39planned to be installed by manufacturers could be detached easily and left that way permanently. Passive belts, once detached, then required “the same type of affirmative action that is the stumbling block to obtaining high usage levels of manual belts.” Id., at 53421. For this reason, the agency concluded that there was no longer a basis for reliably predicting that the Standard would lead to any significant increased usage of restraints at all.

In view of the possibly minimal safety benefits, the automatic restraint requirement no longer was reasonable or practicable in the agency’s view. The requirement would require approximately $1 billion to implement and the agency did not believe it would be reasonable to impose such substantial costs on manufacturers and consumers without more adequate assurance that sufficient safety benefits would accrue. In addition, NHTSA concluded that automatic restraints might have an adverse effect on the public’s attitude toward safety. Given the high expense and limited benefits of detachable belts, NHTSA feared that many consumers would regard the Standard as an instance of ineffective regulation, adversely affecting the public’s view of safety regulation and, in particular, “poisoning . . . popular sentiment toward efforts to improve occupant restraint systems in the future.” Id., at 53424.

State Farm Mutual Automobile Insurance Co. and the National Association of Independent Insurers filed petitions for review of NHTSA’s rescission of the passive restraint Standard. The United States Court of Appeals for the District of Columbia Circuit held that the agency’s rescission of the passive restraint requirement was arbitrary and capricious. 220 U. S. App. D. C. 170, 680 F. 2d 206 (1982). While observing that rescission is not unrelated to an agency’s refusal to take action in the first instance, the court concluded that, in this case, NHTSA’s discretion to rescind the passive restraint requirement had been restricted by various forms of congressional “reaction” to the passive restraint issue. It then *40proceeded to find that the rescission of Standard 208 was arbitrary and capricious for three reasons. First, the court found insufficient as a basis for rescission NHTSA’s conclusion that it could not reliably predict an increase in belt usage under the Standard. The court held that there was insufficient evidence in the record to sustain NHTSA’s position on this issue, and that, “only a well justified refusal to seek more evidence could render rescission non-arbitrary.” Id., at 196, 680 F. 2d, at 232. Second, a majority of the panel8 concluded that NHTSA inadequately considered the possibility of requiring manufacturers to install nondetachable rather than detachable passive belts. Third, the majority found that the agency acted arbitrarily and capriciously by failing to give any consideration whatever to requiring compliance with Modified Standard 208 by the installation of airbags.

The court allowed NHTSA 30 days in which to submit a schedule for “resolving the questions raised in th[e] opinion.” Id., at 206, 680 F. 2d, at 242. Subsequently, the agency filed a Notice of Proposed Supplemental Rulemaking setting forth a schedule for complying with the court’s mandate. On August 4, 1982, the Court of Appeals issued an order staying the compliance date for the passive restraint requirement until September 1, 1983, and requested NHTSA to inform the court whether that compliance date was achievable. NHTSA informed the court on October 1,1982, that based on representations by manufacturers, it did not appear that practicable compliance could be achieved before September 1985. On November 8, 1982, we granted certiorari, 459 U. S. 987, and on November 18, the Court of Appeals entered an order recalling its mandate.

r-H h-I

Unlike the Court of Appeals, we do not find the appropriate scope of judicial review to be the “most troublesome *41question” in these cases. Both the Act and the 1974 Amendments concerning occupant crash protection standards indicate that motor vehicle safety standards are to be promulgated under the informal rulemaking procedures of the Administrative Procedure Act. 5 U. S. C. § 553. The agency’s action in promulgating such standards therefore may be set aside if found to be “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” 5 U. S. C. § 706(2)(A); Citizens to Preserve Overton Park v. Volpe, 401 U. S. 402, 414 (1971); Bowman Transportation, Inc. v. Arkansas-Best Freight System, Inc., 419 U. S. 281 (1974). We believe that the rescission or modification of an occupant-protection standard is subject to the same test. Section 103(b) of the Act, 15 U. S. C. § 1392(b), states that the procedural and judicial review provisions of the Administrative Procedure Act “shall apply to all orders establishing, amending, or revoking a Federal motor vehicle safety standard,” and suggests no difference in the scope of judicial review depending upon the nature of the agency’s action.

Petitioner Motor Vehicle Manufacturers Association (MVMA) disagrees, contending that the rescission of an agency rule should be judged by the same standard a court would use to judge an agency’s refusal to promulgate a rule in the first place — a standard petitioner believes considerably narrower than the traditional arbitrary-and-capricious test. We reject this view. The Act expressly equates orders “revoking” and “establishing” safety standards; neither that Act nor the APA suggests that revocations are to be treated as refusals to promulgate standards. Petitioner’s view would render meaningless Congress’ authorization for judicial review of orders revoking safety rules. Moreover, the revocation of an extant regulation is substantially different than a failure to act. Revocation constitutes a reversal of the agency’s former views as to the proper course. A “settled course of behavior embodies the agency’s informed judgment that, by pursuing that course, it will carry out the policies *42committed to it by Congress. There is, then, at least a presumption that those policies will be carried out best if the settled rule is adhered to.” Atchison, T. & S. F. R. Co. v. Wichita Bd. of Trade, 412 U. S. 800, 807-808 (1973). Accordingly, an agency changing its course by rescinding a rule is obligated to supply a reasoned analysis for the change beyond that which may be required when an agency does not act in the first instance.

In so holding, we fully recognize that “[rjegulatory agencies do not establish rules of conduct to last forever,” American Trucking Assns., Inc. v. Atchison, T. & S. F. R. Co., 387 U. S. 397, 416 (1967), and that an agency must be given ample latitude to “adapt their rules and policies to the demands of changing circumstances.” Permian Basin Area Rate Cases, 390 U. S. 747, 784 (1968). But the forces of change do not always or necessarily point in the direction of deregulation. In the abstract, there is no more reason to presume that changing circumstances require the rescission of prior action, instead of a revision in or even the extension of current regulation. If Congress established a presumption from which judicial review should start, that presumption — contrary to petitioners’ views — is not against safety regulation, but against changes in current policy that are not justified by the rulemaking record. While the removal of a regulation may not entail the monetary expenditures and other costs of enacting a new standard, and, accordingly, it may be easier for an agency to justify a deregulatory action, the direction in which an agency chooses to move does not alter the standard of judicial review established by law.

The Department of Transportation accepts the applicability of the “arbitrary and capricious” standard. It argues that under this standard, a reviewing court may not set aside an agency rule that is rational, based on consideration of the relevant factors, and within the scope of the authority delegated to the agency by the statute. We do not disagree with *43this formulation.9 The scope of review under the “arbitrary and capricious” standard is narrow and a court is not to substitute its judgment for that of the agency. Nevertheless, the agency must examine the relevant data and articulate a satisfactory explanation for its action including a “rational connection between the facts found and the choice made.” Burlington Truck Lines, Inc. v. United States, 371 U. S. 156, 168 (1962). In reviewing that explanation, we must “consider whether the decision was based on a consideration of the relevant factors and whether there has been a clear error of judgment.” Bowman Transportation, Inc. v. Arkansas-Best Freight System, Inc., supra, at 285; Citizens to Preserve Overton Park v. Volpe, supra, at 416. Normally, an agency rule would be arbitrary and capricious if the agency has relied on factors which Congress has not intended it to consider, entirely failed to consider an important aspect of the problem, offered an explanation for its decision that runs counter to the evidence before the agency, or is so implausible that it could not be ascribed to a difference in view or the product of agency expertise. The reviewing court should not attempt itself to make up for such deficiencies; we may not supply a reasoned basis for the agency’s action that the agency itself has not given. SEC v. Chenery Corp., 332 U. S. 194, 196 (1947). We will, however, “uphold a decision of less than ideal clarity if the agency’s path may reasonably be discerned.” Bowman Transportation, Inc. v. Arkansas-Best Freight System, Inc., supra, at 286. See also Camp v. Pitts, 411 U. S. 138, 142-143 (1973) (per curiam). For purposes of these cases, it is also relevant that Congress required a record of the rulemaking proceedings to be compiled *44and submitted to a reviewing court, 15 U. S. C. § 1394, and intended that agency findings under the Act would be supported by “substantial evidence on the record considered as a whole.” S. Rep. No. 1301, 89th Cong., 2d Sess., 8 (1966); H. R. Rep. No. 1776, 89th Cong., 2d Sess., 21 (1966).

>

The Court of Appeals correctly found that the arbitrary- and-capricious test applied to rescissions of prior agency regulations, but then erred in intensifying the scope of its review based upon its reading of legislative events. It held that congressional reaction to various versions of Standard 208 “raise[d] doubts” that NHTSA’s rescission “necessarily demonstrates an effort to fulfill its statutory mandate,” and therefore the agency was obligated to provide “increasingly clear and convincing reasons” for its action. 220 U. S. App. D. C., at 186, 193, 680 F. 2d, at 222, 229. Specifically, the Court of Appeals found significance in three legislative occurrences:

“In 1974, Congress banned the ignition interlock but did not foreclose NHTSA’s pursuit of a passive restraint standard. In 1977, Congress allowed the standard to take effect when neither of the concurrent resolutions needed for disapproval was passed. In 1980, a majority of each house indicated support for the concept of mandatory passive restraints and a majority of each house supported the unprecedented attempt to require some installation of airbags.” Id., at 192, 680 F. 2d, at 228.

From these legislative acts and nonacts the Court of Appeals derived a “congressional commitment to the concept of automatic crash protection devices for vehicle occupants.” Ibid.

This path of analysis was misguided and the inferences it produced are questionable. It is noteworthy that in this Court respondent State Farm expressly agrees that the post-enactment legislative history of the Act does not heighten the *45standard of review of NHTSA’s actions. Brief for Respondent State Farm Mutual Automobile Insurance Co. 13. State Farm’s concession is well taken for this Court has never suggested that the standard of review is enlarged or diminished by subsequent congressional action. While an agency’s interpretation of a statute may be confirmed or ratified by subsequent congressional failure to change that interpretation, Bob Jones University v. United States, 461 U. S. 574, 599-602 (1983); Haig v. Agee, 453 U. S. 280, 291-300 (1981), in the cases before us, even an unequivocal ratification — short of statutory incorporation — of the passive restraint standard would not connote approval or disapproval of an agency’s later decision to rescind the regulation. That decision remains subject to the arbitrary-and-capricious standard.

That we should not be so quick to infer a congressional mandate for passive restraints is confirmed by examining the postenactment legislative events cited by the Court of Appeals. Even were we inclined to rely on inchoate legislative action, the inferences to be drawn fail to suggest that NHTSA acted improperly in rescinding Standard 208. First, in 1974 a mandatory passive restraint standard was technically not in effect, see n. 6, supra; Congress had no reason to foreclose that course. Moreover, one can hardly infer support for a mandatory standard from Congress’ decision to provide that such a regulation would be subject to disapproval by resolutions of disapproval in both Houses. Similarly, no mandate can be divined from the tabling of resolutions of disapproval which were introduced in 1977. The failure of Congress to exercise its veto might reflect legislative deference to the agency’s expertise and does not indicate that Congress would disapprove of the agency’s action in 1981. And even if Congress favored the Standard in 1977, it — like NHTSA — may well reach a different judgment, given changed circumstances four years later. Finally, the Court of Appeals read too much into floor action on the 1980 authorization bill, a bill which was not enacted into law. Other *46contemporaneous events could be read as showing equal congressional hostility to passive restraints.10

V

The ultimate question before us is whether NHTSA’s rescission of the passive restraint requirement of Standard 208 was arbitrary and capricious. We conclude, as did the Court of Appeals, that it was. We also conclude, but for somewhat different reasons, that further consideration of the issue by the agency is therefore required. We deal separately with the rescission as it applies to airbags and as it applies to seatbelts.

A

The first and most obvious reason for finding the rescission arbitrary and capricious is that NHTSA apparently gave no consideration whatever to modifying the Standard to require that airbag technology be utilized. Standard 208 sought to achieve automatic crash protection by requiring automobile manufacturers to install either of two passive restraint devices: airbags or automatic seatbelts. There was no suggestion in the long rulemaking process that led to Standard 208 that if only one of these options were feasible, no passive restraint standard should be promulgated. Indeed, the agency’s original proposed Standard contemplated the installation of inflatable restraints in all cars.11 Automatic belts *47were added as a means of complying with the Standard because they were believed to be as effective as airbags in achieving the goal of occupant crash protection. 36 Fed. Reg. 12859 (1971). At that time, the passive belt approved by the agency could not be detached.12 Only later, at a manufacturer’s behest, did the agency approve of the detach-ability feature — and only after assurances that the feature would not compromise the safety benefits of the restraint.13 Although it was then foreseen that 60% of the new cars would contain airbags and 40% would have automatic seatbelts, the ratio between the two was not significant as long as the passive belt would also assure greater passenger safety.

The agency has now determined that the detachable automatic belts will not attain anticipated safety benefits because so many individuals will detach the mechanism. Even if this conclusion were acceptable in its entirety, see infra, at 51-54, standing alone it would not justify any more than an amendment of Standard 208 to disallow compliance by means of the one technology which will not provide effective passenger protection. It does not cast doubt on the need for a passive restraint standard or upon the efficacy of airbag technology. In its most recent rulemaking, the agency again acknowledged the lifesaving potential of the airbag:

*48“The agency has no basis at this time for changing its earlier conclusions in 1976 and 1977 that basic air bag technology is sound and has been sufficiently demonstrated to be effective in those vehicles in current use . . . NHTSA Final Regulatory Impact Analysis (RIA) XI-4 (Oct. 1981), App. 264.

Given the effectiveness ascribed to airbag technology by the agency, the mandate of the Act to achieve traffic safety would suggest that the logical response to the faults of detachable seatbelts would be to require the installation of airbags. At the very least this alternative way of achieving the objectives of the Act should have been addressed and adequate reasons given for its abandonment. But the agency not only did not require compliance through airbags, it also did not even consider the possibility in its 1981 rulemaking. Not one sentence of its rulemaking statement discusses the airbags-only option. Because, as the Court of Appeals stated, “NHTSA’s . . . analysis of airbags was nonexistent,” 220 U. S. App. D. C., at 200, 680 F. 2d, at 236, what we said in Burlington Truck Lines, Inc. v. United States, 371 U. S., at 167, is apropos here:

“There are no findings and no analysis here to justify the choice made, no indication of the basis on which the [agency] exercised its expert discretion. We are not prepared to and the Administrative Procedure Act will not permit us to accept such . . . practice. . . . Expert discretion is the lifeblood of the administrative process, but ‘unless we make the requirements for administrative action strict and demanding, expertise, the strength of modern government, can become a monster which rules with no practical limits on its discretion.’ New York v. United States, 342 U. S. 882, 884 (dissenting opinion)” (footnote omitted).

We have frequently reiterated that an agency must cogently explain why it has exercised its discretion in a given manner, *49Atchison, T. & S. F. R. Co. v. Wichita Bd. of Trade, 412 U. S., at 806; FTC v. Sperry & Hutchinson Co., 405 U. S. 233, 249 (1972); NLRB v. Metropolitan Life Ins. Co., 380 U. S. 438, 443 (1965); and we reaffirm this principle again today.

The automobile industry has opted for the passive belt over the airbag, but surely it is not enough that the regulated industry has eschewed a given safety device. For nearly a decade, the automobile industry waged the regulatory equivalent of war against the airbag14 and lost — the inflatable restraint was proved sufficiently effective. Now the automobile industry has decided to employ a seatbelt system which will not meet the safety objectives of Standard 208. This hardly constitutes cause to revoke the Standard itself. Indeed, the Act was necessary because the industry was not sufficiently responsive to safety concerns. The Act intended that safety standards not depend on current technology and could be “technology-forcing” in the sense of inducing the development of superior safety design. See Chrysler Corp. v. Department of Transportation, 472 F. 2d, at 672-673. If, under the statute, the agency should not defer to the industry’s failure to develop safer cars, which it surely should not do, a fortiori it may not revoke a safety standard which can be satisfied by current technology simply because the industry has opted for an ineffective seatbelt design.

Although the agency did not address the mandatory airbag option and the Court of Appeals noted that “airbags seem to have none of the problems that NHTSA identified in passive seatbelts,” 220 U. S. App. D. C., at 201, 680 F. 2d, at 237, petitioners recite a number of difficulties that they *50believe would be posed by a mandatory airbag standard. These range from questions concerning the installation of airbags in small cars to that of adverse public reaction. But these are not the agency’s reasons for rejecting a mandatory airbag standard. Not having discussed the possibility, the agency submitted no reasons at all. The short — and sufficient — answer to petitioners’ submission is that the courts may not accept appellate counsel’s post hoc rationalizations for agency action. Burlington Truck Lines, Inc. v. United States, 371 U. S., at 168. It is well established that an agency’s action must be upheld, if at all, on the basis articulated by the agency itself. Ibid.; SEC v. Chenery Corp., 332 U. S., at 196; American Textile Mfrs. Institute, Inc. v. Donovan, 452 U. S. 490, 539 (1981).15

Petitioners also invoke our decision in Vermont Yankee Nuclear Power Corp. v. Natural Resources Defense Council, Inc., 435 U. S. 519 (1978), as though it were a talisman under which any agency decision is by definition unimpeachable. Specifically, it is submitted that to require an agency to consider an airbags-only alternative is, in essence, to dictate to the agency the procedures it is to follow. Petitioners both misread Vermont Yankee and misconstrue the nature of the remand that is in order. In Vermont Yankee, we held that a court may not impose additional procedural requirements upon an agency. We do not require today any specific proce*51dures which NHTSA must follow. Nor do we broadly require an agency to consider all policy alternatives in reaching decision. It is true that rulemaking “cannot be found wanting simply because the agency failed to include every alternative device and thought conceivable by the mind of man . . . regardless of how uncommon or unknown that alternative may have been . . . .” Id., at 551. But the airbag is more than a policy alternative to the passive restraint Standard; it is a technological alternative within the ambit of the existing Standard. We hold only that given the judgment made in 1977 that airbags are an effective and cost-beneficial lifesaving technology, the mandatory passive restraint rule may not be abandoned without any consideration whatsoever of an airbags-only requirement.

B

Although the issue is closer, we also find that the agency was too quick to dismiss the safety benefits of automatic seatbelts. NHTSA’s critical finding was that, in light of the industry’s plans to install readily detachable passive belts, it could not reliably predict “even a 5 percentage point increase as the minimum level of expected usage increase.” 46 Fed. Reg. 53423 (1981). The Court of Appeals rejected this finding because there is “not one iota” of evidence that Modified Standard 208 will fail to increase nationwide seatbelt use by at least 13 percentage points, the level of increased usage necessary for the Standard to justify its cost. Given the lack of probative evidence, the court held that “only a well justified refusal to seek more evidence could render rescission non-arbitrary.” 220 U. S. App. D. C., at 196, 680 F. 2d, at 232.

Petitioners object to this conclusion. In their view, “substantial uncertainty” that a regulation will accomplish its intended purpose is sufficient reason, without more, to rescind a regulation. We agree with petitioners that just as an agency reasonably may decline to issue a safety standard if it is uncertain about its efficacy, an agency may also revoke a *52standard on the basis of serious uncertainties if supported by the record and reasonably explained. Rescission of the passive restraint requirement would not be arbitrary and capricious simply because there was no evidence in direct support of the agency’s conclusion. It 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. Recognizing that policymaking in a complex society must account for uncertainty, however, does not imply that it is sufficient for an agency to merely recite the terms “substantial uncertainty” as a justification for its actions. As previously noted, the agency must explain the evidence which is available, and must offer a “rational connection between the facts found and the choice made.” Burlington Truck Lines, Inc. v. United States, supra, at 168. Generally, one aspect of that explanation would be a justification for rescinding the regulation before engaging in a search for further evidence.

In these cases, the agency’s explanation for rescission of the passive restraint requirement is not sufficient to enable us to conclude that the rescission was the product of reasoned decisionmaking. To reach this conclusion, we do not upset the agency’s view of the facts, but we do appreciate the limitations of this record in supporting the agency’s decision. We start with the accepted ground that if used, seatbelts unquestionably would save many thousands of lives and would prevent tens of thousands of crippling injuries. Unlike recent regulatory decisions we have reviewed, Industrial Union Dept. v. American Petroleum Institute, 448 U. S. 607 (1980); American Textile Mfrs. Institute, Inc. v. Donovan, 452 U. S. 490 (1981), the safety benefits of wearing seatbelts are not in doubt, and it is not challenged that were those benefits to accrue, the monetary costs of implementing the Standard would be easily justified. We move next to the fact that there is no direct evidence in support of the agency’s finding that detachable automatic belts cannot be predicted *53to yield a substantial increase in usage. The empirical evidence on the record, consisting of surveys of drivers of automobiles equipped with passive belts, reveals more than a doubling of the usage rate experienced with manual belts.16 Much of the agency’s rulemaking statement — and much of the controversy in these cases — centers on the conclusions that should be drawn from these studies. The agency maintained that the doubling of seatbelt usage in these studies could not be extrapolated to an across-the-board mandatory standard because the passive seatbelts were guarded by ignition interlocks and purchasers of the tested cars are somewhat atypical.17 Respondents insist these studies demonstrate that Modified Standard 208 will substantially increase seatbelt usage. We believe that it is within the agency’s discretion to pass upon the generalizability of these field studies. This is precisely the type of issue which rests within the expertise of NHTSA, and upon which a reviewing court must be most hesitant to intrude.

But accepting the agency’s view of the field tests on passive restraints indicates only that there is no reliable real-world experience that usage rates will substantially increase. To be sure, NHTSA opines that “it cannot reliably predict even a 5 percentage point increase as the minimum level of *54expected increased usage.” Notice 25, 46 Fed. Reg. 53423 (1981). But this and other statements that passive belts will not yield substantial increases in seatbelt usage apparently take no account of the critical difference between detachable automatic belts and current manual belts. A detached passive belt does require an affirmative act to reconnect it, but— unlike a manual seatbelt — the passive belt, once reattached, will continue to function automatically unless again disconnected. Thus, inertia — a factor which the agency’s own studies have found significant in explaining the current low usage rates for seatbelts18 — works in favor of, not against, use of the protective device. Since 20% to 50% of motorists currently wear seatbelts on some occasions,19 there would seem to be grounds to believe that seatbelt use by occasional users will be substantially increased by the detachable passive belts. Whether this is in fact the case is a matter for the agency to decide, but it must bring its expertise to bear on the question.

The agency is correct to look at the costs as well as the benefits of Standard 208. The agency’s conclusion that the incremental costs of the requirements were no longer reasonable was predicated on its prediction that the safety benefits of the regulation might be minimal. Specifically, the *55agency’s fears that the public may resent paying more for the automatic belt systems is expressly dependent on the assumption that detachable automatic belts will not produce more than “negligible safety benefits.” Id., at 53424. When the agency reexamines its findings as to the likely increase in seatbelt usage, it must also reconsider its judgment of the reasonableness of the monetary and other costs associated with the Standard. In reaching its judgment, NHTSA should bear in mind that Congress intended safety to be the pre-eminent factor under the Act:

“The Committee intends that safety shall be the overriding consideration in the issuance of standards under this bill. The Committee recognizes . . . that the Secretary will necessarily consider reasonableness of cost, feasibility and adequate leadtime.” S. Rep. No. 1301, 89th Cong., 2d Sess., 6 (1966).

“In establishing standards the Secretary must conform to the requirement that the standard be practicable. This would require consideration of all relevant factors, including technological ability to achieve the goal of a particular standard as well as consideration of economic factors.

“Motor vehicle safety is the paramount purpose of this bill and each standard must be related thereto.” H. R. Rep. No. 1776, 89th Cong., 2d Sess., 16 (1966).

The agency also failed to articulate a basis for not requiring nondetachable belts under Standard 208. It is argued that the concern of the agency with the easy detachability of the currently favored design would be readily solved by a continuous passive belt, which allows the occupant to “spool out” the belt and create the necessary slack for easy extrication from the vehicle. The agency did not separately consider the continuous belt option, but treated it together with the ignition interlock device in a category it titled “Option of Adopting Use-Compelling Features.” 46 Fed. Reg. 53424 *56(1981). The agency was concerned that use-compelling devices would “complicate the extrication of [an] occupant from his or her car.” Ibid. “[T]o require that passive belts contain use-compelling features,” the agency observed, “could be counterproductive [, given]. . . widespread, latent and irrational fear in many members of the public that they could be trapped by the seat belt after a crash.” Ibid. In addition, based on the experience with the ignition interlock, the agency feared that use-compelling features might trigger adverse public reaction.

By failing to analyze the continuous seatbelts option in its own right, the agency has failed to offer the rational connection between facts and judgment required to pass muster under the arbitrary-and-capricious standard. We agree with the Court of Appeals that NHTSA did not suggest that the emergency release mechanisms used in nondetachable belts are any less effective for emergency egress than the buckle release system used in detachable belts. In 1978, when General Motors obtained the agency’s approval to install a continuous passive belt, it assured the agency that nondetachable belts with spool releases were as safe as detachable belts with buckle releases. 43 Fed. Reg. 21912, 21913-21914 (1978). NHTSA was satisfied that this belt design assured easy extricability: “[t]he agency does not believe that the use of [such] release mechanisms will cause serious occupant egress problems . . . .” Id., at 52493, 52494. While the agency is entitled to change its view on the acceptability of continuous passive belts, it is obligated to explain its reasons for doing so.

The agency also failed to offer any explanation why a continuous passive belt would engender the same adverse public reaction as the ignition interlock, and, as the Court of Appeals concluded, “every indication in the record points the other way.” 220 U. S. App. D. C., at 198, 680 F. 2d, at 234.20

*57We see no basis for equating the two devices: the continous belt, unlike the ignition interlock, does not interfere with the operation of the vehicle. More importantly, it is the agency’s responsibility, not this Court’s, to explain its decision.

VI

“An agency’s view of what is in the public interest may change, either with or without a change in circumstances. But an agency changing its course must supply a reasoned analysis . ...” Greater Boston Television Corp. v. FCC, 143 U. S. App. D. C. 383, 394, 444 F. 2d 841, 852 (1970) (footnote omitted), cert. denied, 403 U. S. 923 (1971). We do not accept all of the reasoning of the Court of Appeals but we do conclude that the agency has failed to supply the requisite “reasoned analysis” in this case. Accordingly, we vacate the judgment of the Court of Appeals and remand the cases to that court with directions to remand the matter to the NHTSA for further consideration consistent with this opinion.21

So ordered.

Justice Rehnquist,

with whom The Chief Justice, Justice Powell, and Justice O’Connor join, concurring in part and dissenting in part.

I join Parts I, II, III, IV, and V-A of the Court’s opinion. In particular, I agree that, since the airbag and continuous *58spool automatic seatbelt were explicitly approved in the Standard the agency was rescinding, the agency should explain why it declined to leave those requirements intact. In this case, the agency gave no explanation at all. Of course, if the agency can provide a rational explanation, it may adhere to its decision to rescind the entire Standard.

I do not believe, however, that NHTSA’s view of detachable automatic seatbelts was arbitrary and capricious. The agency adequately explained its decision to rescind the Standard insofar as it was satisfied by detachable belts.

The statute that requires the Secretary of Transportation to issue motor vehicle safety standards also requires that “[e]ach such . . . standard shall be practicable [and] shall meet the need for motor vehicle safety.” 15 U. S. C. § 1392(a) (1976 ed., Supp. V). The Court rejects the agency’s explanation for its conclusion that there is substantial uncertainty whether requiring installation of detachable automatic belts would substantially increase seatbelt usage. The agency chose not to rely on a study showing a substantial increase in seatbelt usage in cars equipped with automatic seatbelts and an ignition interlock to prevent the car from being operated when the belts were not in place and which were voluntarily purchased with this equipment by consumers. See ante, at 53, n. 16. It is reasonable for the agency to decide that this study does not support any conclusion concerning the effect of automatic seatbelts that are installed in all cars whether the consumer wants them or not and are not linked to an ignition interlock system.

The Court rejects this explanation because “there would seem to be grounds to believe that seatbelt use by occasional users will be substantially increased by the detachable passive belts,” ante, at 54, and the agency did not adequately explain its rejection of these grounds. It seems to me that the agency’s explanation, while by no means a model, is adequate. The agency acknowledged that there would probably be some increase in belt usage, but concluded that the increase would be small and not worth the cost of manda*59tory detachable automatic belts. 46 Fed. Reg. 53421-53423 (1981). The agency’s obligation is to articulate a “‘rational connection between the facts found and the choice made.’” Ante, at 42, 52, quoting Burlington Truck Lines, Inc. v. United States, 371 U. S. 156, 168 (1962). I believe it has met this standard.

The agency explicitly stated that it will increase its educational efforts in an attempt to promote public understanding, acceptance, and use of passenger restraint systems. 46 Fed. Reg. 53425 (1981). It also stated that it will “initiate efforts with automobile manufacturers to ensure that the public will have [automatic crash protection] technology available. If this does not succeed, the agency will consider regulatory action to assure that the last decade’s enormous advances in crash protection technology will not be lost.” Id., at 53426.

The agency’s changed view of the standard seems to be related to the election of a new President of a different political party. It is readily apparent that the responsible members of one administration may consider public resistance and uncertainties to be more important than do their counterparts in a previous administration. A change in administration brought about by the people casting their votes is a perfectly reasonable basis for an executive agency’s reappraisal of the costs and benefits of its programs and regulations. As long as the agency remains within the bounds established by Congress,* it is entitled to assess administrative records and evaluate priorities in light of the philosophy of the administration.

Chief Justice ROBERTS delivered the opinion of the Court.

*2561The 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 *2562recent 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.

*2563The 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 *2564under 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 *2565court 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 *2566even 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 *2567have 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 *2568commits 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, *2569see § 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 *2570between 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.

*2571The 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.

*2572At 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 *2573in § 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 *2575particular 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 *2576decisions, 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 *2577that "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 *2578accordance 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 *2579reasons" 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 *2580an 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 *2581the 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 *2582that 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 *2583regularity 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 *2584enables 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 *2586the 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 *2587fears 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 *2588between 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 *2589arguments 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 *2590officials 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 *2591further 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 *2592would 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 *2593the 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

*2594enforce 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.

*2595Taken 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.

Justice ALITO, concurring in part and dissenting in part.

*2596It 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 *2597to 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 *2598another 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 *2599deference 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 *2600forth 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 § 1418 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,"

*2602i.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 *2603Executive 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).

*2604Here, 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.

*2605This 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,"

*2606and 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-C15 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.

Opinion for the Court filed by Circuit. Judge SCALIA.

SCALIA, Circuit Judge:

The Association of Data Processing Service Organizations, Inc. (“ADAPSO”), a national trade association representing the data processing industry, and two of its members petition this court for review of two orders of the Board of Governors of the Federal Reserve System, pursuant to 12 U.S.C. § 1848 (1982). In No. 82-1910, they seek review of the Board’s July 9, 1982 order approving Citicorp’s application to establish a subsidiary, Citishare, to engage in certain data processing and transmission services. Order Approving Engaging in Data Processing and Data Transmission Activities, 68 Fed.Res.Bull. 505 (1982) (“Citicorp Order”). In No. 82-2108, they seek review of the Board’s August 23, 1982 order, entered after notice and comment rulemaking, amending those portions of Regulation Y which dealt with the performance of data processing activities by bank holding companies. Data *305 Processing and Electronic Funds Transfer Activities, 47 Fed.Reg. 37,368 (1982) (as set forth at 12 C.F.R. §§ 225.4(a)(8), 225.-123(e) (1983) (“Regulation Y Order”).1 We consolidated the two appeals'.

The Bank Holding Company Act of 1956, ch. 240, 70 Stat. 133 (codified as amended at 12 U.S.C. §§ 1841-50 (1982)) (the “Act”), requires all bank holding companies to seek prior regulatory approval before engaging in nonbanking activities. The restrictions do not apply to:

activities ... which the Board after due notice and opportunity for hearing has determined (by order or regulation) to be so closely related to banking or managing or controlling banks as to be a proper incident thereto____ In determining whether a particular activity is a proper incident to banking or managing or controlling banks the Board shall consider whether its performance by an affiliate of a holding company can reasonably be expected to produce benefits to the public, such as greater convenience, increased competition, or gains in efficiency, that outweigh possible adverse effects, such as undue concentration of resources, decreased or unfair competition, conflicts of interests, or unsound banking practices.

12 U.S.C. § 1843(c)(8). Section 1848, the source of our review authority, provides that “[t]he findings of the Board as to the facts, if supported by substantial evidence, shall be conclusive.” Id. at § 1848.

On February 23, 1979, Citicorp applied for authority to engage, through its subsidiary Citishare, in the processing and transmission of banking, financial, and economic related data through timesharing, electronic funds transfer, home banking and other techniques. It also sought permission to sell its excess computing capacity and some computer hardware. The Board published notice of Citicorp’s application, which was protested by ADAPSO, and set it for formal hearing. 45 Fed.Reg. 41,533 (July 19, 1980). Before the hearing was held, Citicorp amended its application to add certain activities and to request amendment of Regulation Y to permit the activities it had specified. The Board published an Amended Order for Hearing and invited public comments and participation. 45 Fed.Reg. 76,515 (Nov. 19, 1980). A formal hearing was held before an Administrative Law Judge in which the merits of both the application and the proposed rule were considered. In addition, more than sixty companies and individuals submitted written comments on the proposed rule. On March 29, 1982, the AU decided that the activities proposed by Citicorp were closely related to banking and would produce benefits to the public which would outweigh their costs. In re: Application of Citicorp to Engage in Data Processing and Transmission Activities, AU Recommended Decision, J.A. B-68 to B-123 (“Recommended Decision”). The AU also recommended amendments to Regulation Y that would permit those activities contained in the Citicorp application. On July 9, 1982, the Board adopted the AU’s recommendation to approve the Citicorp application, with certain restrictions. On August 23, 1982, the Board adopted the AU’s recommended amendments to Regulation Y, again with certain restrictions. ADAPSO, and two of its members, participants in the actions below, filed these petitions for review.

I. Standard of Review

We are faced at the outset with a dispute regarding the proper standard of review. These consolidated appeals call for us to review both an on-the-record adjudication and an informal notice and comment rule-making. Petitioners contend that the substantial evidence standard, which presumably authorizes more rigorous judicial review, should govern our review of both orders. Petitioners’ Reply Brief at 26-28. The Board agrees, noting that § 1848 applies a substantial evidence standard to fac*306tual determinations. Respondent’s Brief at 22. Intervenor Citicorp contends that while the substantial evidence standard should govern review of the Citicorp order, Regulation Y should be upset only if arbitrary or capricious. Citicorp Brief at 16-17. Intervenor California Bankers Clearing House Association, addressing only Regulation Y, also advocates review under the arbitrary or capricious review standard. CBCHA Brief at 8-14. The parties’ submissions on this point reflect considerable confusion, which is understandable when one examines decisions defining the standard.of review under this statute.

Both of the Supreme Court's opinions reviewing action of the Board in amending Regulation Y noted that the Board’s determination “is entitled to the greatest deference,” Board of Governors of the Federal Reserve System v. Investment Company Institute, 450 U.S. 46, 56, 101 S.Ct. 973, 981-82, 67 L.Ed.2d 36 (1981) (“ICI ”); Securities Industry Association v. Board of Governors of the Federal Reserve System, — U.S. -, 104 S.Ct. 3003, 3009, 82 L.Ed.2d 158 (1984) (“SIA ”), but neither of them discussed the applicable standard of review, or even referred to § 1848.2 The courts of appeals, however, have applied the substantial evidence standard of § 1848 to Board adjudications such as the authorization in the first order here under review, Securities Industry Association v. Board of Governors of the Federal Reserve System, 716 F.2d 92, 101-02 (2d Cir.1983), aff'd, — U.S. -, 104 S.Ct. 3003, 82 L.Ed.2d 158 (1984), while applying the arbitrary or capricious standard, despite § 1848, to Board rules, including specifically amendments of Regulation Y, National Courier Association v. Board of Governors of the Federal Reserve System, 516 F.2d 1229 (D.C.Cir.1975); Association of Bank Travel Bureaus v. Board of Governors of the Federal Reserve System, 568 F.2d 549 (7th Cir.1978); see Investment Company Institute v. Board of Governors of the Federal Reserve System, 551 F.2d 1270, 1281 (D.C.Cir.1977) (dicta). In fact one appellate opinion has, like this one, addressed precisely the situation in which both an adjudicatory authorization and an amendment of Regulation Y were at issue in the same case—and applied the § 1848 substantial evidence standard to the former but the arbitrary or capricious standard to the latter. Compare Alabama Association of Insurance Agents v. Board of Governors of the Federal Reserve System, 533 F.2d 224, 246 (5th Cir.1976), with id. at 240. This would make a lot of sense if, as the Board has argued in some cases, § 1848 in its totality applies only to adjudication rather than rulemaking, since it is limited to “orders” of the Board, a word which the Administrative Procedure Act (“APA”) defines to mean the product of an adjudication. See 5 U.S.C. § 551(4), (6) (1982). Such a technical interpretation of the provision, however, has been uniformly and quite correctly rejected. See Investment Company Institute, supra, 551 F.2d at 1276-78; Alabama Association of Insurance Agents, supra, 533 F.2d at 234-35. That leaves the courts with the difficult task of explaining why the last sentence of § 1848, unlike all the rest of it, should be deemed to apply only to adjudication and not to rulemaking. Difficult, because there is nothing in either the text3 or the *307legislative history of the section to suggest such a result. The courts applying the arbitrary or capricious standard to Board rulemaking (which, as stated above, include all the courts that have confronted the issue) dispose of this problem either by totally ignoring it, see Alabama Association of Insurance Agents, supra, 533 F.2d at 240, 246; cf Investment Company Institute, supra, 551 F.2d at 1281, or by noting that the parties “do not appear to contest” the point, National Courier, supra, 4 516 F.2d at 1235 n. 8, or by the ipse dixit that “[w]e interpret [the last sentence of § 1848] to apply to findings of fact ‘on the record’ in an adjudicatory hearing as contrasted with a rulemaking proceeding,” Association of Bank Travel Bureaus, supra, 568 F.2d at 552 n. 5.

We think that there is no basis for giving the last sentence of § 1848 anything less than the general application given to the rest of the section. The Supreme Court’s pronouncement that the “greatest deference” is to be given to the determinations of the Board, and the court of appeals decisions applying the arbitrary or capricious test to Board rulemaking, seem to us explicable on quite different grounds— namely, that in their application to the requirement of factual support the substantial evidence test and the arbitrary or capricious test are one and the same. The former is only a specific application of the latter, separately recited in the APA not to establish a more rigorous standard of factual support but to emphasize that in the case of formal proceedings the factual support must be found in the closed record as opposed to elsewhere. We shall elaborate upon this point because it is not uncommon for parties to expend great effort in appeals before us to establish which of the two standards is applicable where in fact their operation is precisely the same.

The “scope of review” provisions of the APA, 5 U.S.C. § 706(2),5 are cumulative. Thus, an agency action which is supported by the required substantial evidence may in another regard be “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law”—for example, because it is an abrupt and unexplained departure from agency precedent. Paragraph (A) of subsection 706(2)—the “arbitrary or capricious” provision—is a catchall, picking up administrative misconduct not covered by the other more specific paragraphs. Thus, in those situations where paragraph (E) has no application (informal rulemaking, for example, which is not governed by §§ 556 and 557 to which paragraph (E) refers), paragraph (A) takes up the slack, so to speak, enabling the courts to strike down, as arbitrary, agency action that is devoid of needed factual support. When the arbitrary or capricious standard is performing that function of assuring factual support, there is no substantive difference between what it requires and what would be required by the *308substantial evidence test, since it is impossible to conceive of a “nonarbitrary” factual judgment supported only by evidence that is not substantial in the APA sense— i.e., not “ ‘enough to justify, if the trial were to a jury, a refusal to direct a verdict when the conclusion sought to be drawn ... is one of fact for the jury,’ ” Illinois Central R.R. v. Norfolk & Western Ry., 385 U.S. 57, 66, 87 S.Ct. 255, 260, 17 L.Ed.2d 162 (1966) (quoting NLRB v. Columbian Enameling & Stamping Co., 306 U.S. 292, 300, 59 S.Ct. 501, 505, 83 L.Ed. 660 (1939)).

We have noted on several occasions that the distinction between the substantial evidence test and the arbitrary or capricious test is “largely semantic,” Aircraft Owners and Pilots Association v. FAA, 600 F.2d 965, 971 n. 28 (D.C.Cir.1979); Pacific Legal Foundation v. Department of Transportation, 593 F.2d 1338, 1343 n. 35 (D.C.Cir.1979); American Public Gas Association v. FPC, 567 F.2d 1016, 1028-29 (D.C.Cir.1977), and have indeed described that view as “the emerging consensus of the Courts of Appeals,” Pacific Legal Foundation, supra, 593 F.2d at 1343 n. 35. See Associated Industries v. Department of Labor, 487 F.2d 342, 349-50 (2d Cir.1973) (Friendly, J.); National Nutritional Foods Association v. Weinberger, 512 F.2d 688, 705. (2d Cir.1975) (Lumbard, J., concurring); Paccar, Inc. v. NHTSA, 573 F.2d 632, 636 (9th Cir.1978) (purporting to avoid the issue, but seemingly not doing so). Leading commentators agree:

Does the extent of required factual support for rules depend in part on whether the standard for review is “substantial evidence” or “arbitrary and capricious”? Although from 1946 until some time during the 1970s the dominant answer probably was yes, a change to a no answer has probably occurred during the 1970s ....

1 K. Davis, Administrative Law Treatise § 6:13 at 512 (2d ed. 1978).

In review of rules of general applicability made after “notice and comment” rule-making, [substantial evidence and arbitrary or capricious] criteria converge into a test of reasonableness.

Review without an agency record thus comes down to review of reasonableness. [T]he question of reasonableness is also the one which the court must now ask itself in reviewing findings of fact under the post-APA substantial evidence rule.

B. Schwartz, Administrative Law 604, 606 (1976).

As noted earlier, this does not consign paragraph (E) of the APA’s judicial review section to pointlessness. The distinctive function of paragraph (E)—what it achieves that paragraph (A) does not—is to require substantial evidence to be found within the record of closed-record proceedings to which it exclusively applies. The importance of that requirement should not be underestimated. It is true that, as the Supreme Court said in Camp v. Pitts, 411 U.S. 138, 142, 93 S.Ct. 1241, 1244, 36 L.Ed.2d 106 (1973), even informal agency action (not governed by paragraph (E)) must be reviewed only on the basis of “the administrative record already in existence.” But that is quite a different and less onerous requirement, meaning only that whether the administrator was arbitrary must be determined on the basis of what he had before him when he acted, and not on the basis of “some new record made initially in the reviewing court,” id. That “administrative record” might well include crucial material that was neither shown to nor known by the private parties in the proceeding—as indeed appears to have been the situation in Camp v. Pitts itself. It is true that, in informal rulemaking, at least the most critical factual material that is used to support the agency’s position on review must have been made public in the proceeding and exposed to refutation. That requirement, however, does not extend to all data, see Administrative Conference of the United States, Recommendation No. 74-4, Preenforcement Judicial Review of Rules of General Applicability, 1 C.F.R. § 305.74-4 (1975); Verkuil, Judicial Review of Informal Rulemaking, 60 Va.L. *309Rev. 185 (1974); and it only applies in rule-making and not in other informal agency action, since it derives not from the arbitrary or capricious test but from the command of 5 U.S.C. § 553(c) that “the agency ... give interested persons an opportunity to participate in the rule making.” See Portland Cement Association v. Ruckelshaus, 486 F.2d 375, 393 n. 67 (D.C.Cir.1973).

Consolidated cases such as those before us here—involving simultaneous review of a rule (whose factual basis is governed only by paragraph (A)’s catch-all control against “arbitrary or capricious” action) and of a formal adjudication dealing with the same subject (whose factual basis is governed by paragraph (E)’s requirement of substantial evidence)—demonstrate why the foregoing interpretation of the two standards is the only interpretation that makes sense. If the standards were substantively different (and leaving aside for the moment consideration of any special effect of § 1848), the Citicorp order, authorizing one bank holding company’s data processing services, would be subject to more rigorous judicial review of factual support than the Regulation Y order which, due to its general applicability, would affect the operations of every bank holding company in the nation. Or, to put the point another way: If the Board had never issued any Regulation Y, and simply determined in the context of a particular application that the provision of timesharing services is “closely related” to banking, that determination, which could be reconsidered and revised in the context of the next adjudication, would require more factual support than the same determination in a rule-making, which would have immediate nationwide application and, until amended by further rulemaking, would have to be applied to all subsequent applications.

This seemingly upside-down application of varying standards is not an issue in the present case since, as we have observed, § 1848 makes it clear that only one standard—the substantial evidence test—applies to review of all Board actions. The relevance of the foregoing discussion here is to determine what that standard means. What we have said suggests that the normal (APA) meaning of the “substantial evidence” terminology connotes a substantive standard no different from the arbitrary or capricious test. One cannot dismiss out of hand, however, the possibility that, in this particular statute, a different meaning was intended—in which case that different standard would govern review of both rulemaking and adjudication. A number of “substantial evidence” review provisions have been attached to rulemaking authority, particularly in recent years. See, e.g., 29 U.S.C. § 655(f) (1982) (Occupational Safety and Health Act); 30 U.S.C. § 816(a) (1982) (Federal Coal Mine Health and Safety Act); 15 U.S.C. § 1193(e)(3) (1982) (Flammable Fabrics Act); 15 U.S.C. § 57a(e)(3)(A) (1982) (FTC Improvement Act of 1975). It is conceivable that some of these were intended, as the Fifth Circuit found with regard to such a provision in the Consumer Product Safety Act, 15 U.S.C. § 2060 (1982), to require the courts “to scrutinize [agency] actions more closely than an ‘arbitrary and capricious’ standard would allow.” Aqua Slide ‘N’ Dive Corp. v. CPSC, 569 F.2d 831, 837 (5th Cir.1978). Congress’s unpropitious use of the “substantial evidence” APA language for such a purpose is plausible, since the standard has acquired a reputation for being more stringent.6 One should not be too quick, however, to impute such a congressional intent. There is surely little appeal to an ineffable review standard that lies somewhere in-between the quantum of factual support required to go to a jury (the traditional “substantial evidence” test) and the “preponderance of the evidence” standard that would apply in de novo review. Moreover, 5 *310U.S.C. § 559 provides that a subsequent statute shall not be held to supersede or modify the APA provisions “except to the extent that it does so expressly.” While the provision for “substantial evidence” review where the APA would otherwise require only “arbitrary or capricious” review is unquestionably an “express” alteration, surely the import of the § 559 instruction is that Congress’s intent to make a substantive change be clear. This can lead to some fairly convoluted inquiries. Suppose, for example, that Congress clearly intended to switch to a stricter test, but was also clearly operating on the mistaken belief that the existing test (“arbitrary or capricious”) was more lenient than the “substantial evidence” standard. Should one give effect to the congressional intent to adopt a stricter standard, or rather to the congressional intent to adopt the “substantial evidence” standard (which is in fact, as we have discussed, no stricter)? Several decisions of this court stand for the proposition that a “substantial evidence” provision in the substantive statute under consideration did not have the effect of requiring increased factual support beyond that demanded by the normal “arbitrary or capricious” rulemaking standard of review. Public Systems v. FERC, 606 F.2d 973, 980 n. 34 (D.C.Cir.1979) (interpreting the predecessors of § 19(b) of the Natural Gas Act, 15 U.S.C. § 717r(b) (1982), and the corresponding provision of the Federal Power Act, 16 U.S.C. § 825i (b) (1982)); American Public Gas Association v. FPC, supra, 567 F.2d at 1028-29 (interpreting the predecessor of § 19(b) of the Natural Gas Act). See also Aircraft Owners and Pilots Association v. FA A, supra, 600 F.2d at 971 n. 28 (not deciding the issue with regard to the “substantial evidence” provision of the Federal Aviation Act, 49 U.S.C. § 1486(e) (1976)).

Fortunately, it is not necessary to engage in these speculations with regard to the “substantial evidence” provision of § 1848. The Supreme Court has evidently rejected the notion that it alters normal APA review requirements, since the Court’s opinions reviewing Board action deem the provision unworthy of mention, and specifically accord the Board “the greatest deference.” See, e.g., SIA, supra, 104 S.Ct. at 3009; ICI, supra, 450 U.S. at 56-57, 101 S.Ct. at 981-982. See also National Courier, supra, 516 F.2d at 1237. We hold, therefore, that the § 1848 “substantial evidence” requirement applicable to our review here demands a quantum of factual support no different from that demanded by the substantial evidence provision of the APA, which is in turn no different from that demanded by the arbitrary or capricious standard.

II. Closely Related to Banking

This appeal requires us, then, to decide whether the Board of Governors acted arbitrarily or capriciously—either because no substantial evidence existed to support its factual premises, or in any other respect—when it concluded, based on the record of the proceedings before it, that the new activities Citicorp proposed were closely related to banking. The test which the Board applied was that developed by this court in National Courier, supra, 516 F.2d at 1237:

As to what kinds of connections may qualify [as closely related to banking], at least the following seem to us within the statutory intent:

1. Banks generally have in fact provided the proposed services.

2. Banks generally provide services that are operationally or functionally so similar to the proposed services as to equip them particularly well to provide the proposed service.

3. Banks generally provide services that are so integrally related to the proposed services as to require their provision in a specialized form.

See, e.g., Citicorp Order, 68 Fed.Res.Bull. at 506 n. 3; Regulation Y Order, 47 Fed. Reg. 37,368 n. 1. These criteria are disjunctive rather than cumulative, see National Cornier, supra, 516 F.2d at 1237-38, and not exclusive, see 49 Fed.Reg. 806 (1984); SIA, supra, 104 S.Ct. at 3006 n. 5.

*311 The “Data Test”

Petitioners do not contend that the National Courier criteria are in themselves erroneous (they have subsequently been approved by the Supreme Court, see SIA, supra, 104 S.Ct. at 3006 n. 5), but they do contend, preliminarily, that the Board erroneously applied them to the kinds of data involved in the proposed services, rather than to the services themselves.

The depth of confusion that surrounds this argument is suggested by the fact that the Board’s brief denies that a data test was applied, Respondent’s Brief at 27-33, while the briefs of intervenors supporting the Board acknowledge that, at least as to the Regulation Y order, it was applied but assert that its application was lawful, Citicorp Brief at 27-32; CBCHA Brief at 27-32. We must address the Board’s contention first.

The Board is quite correct that the Citicorp order specifically examined each of the eight categories.of data processing the applicant proposed to provide. But some of those categories were only described with reference to (1) the data processing technology employed and (2) the nature of the data to be provided or processed, and the Board clearly held that the former “is not determinative of whether [the] activity is permissible,” Citicorp Order, 68 Fed.Res. Bull, at 507. The conclusion is therefore inescapable that, even in the Citicorp order, a data test was employed. With regard to timesharing services, for example, the applicant proposed to provide “data processing and transmission services for financial and non-financial institutions wherein the data being processed and transmitted are financial, banking or economic related.” J.A. C-54. The Board approved this, with the one change (which will be discussed further in another portion of this opinion) that “economic related” was altered to “economic.” Citicorp Order, 68 Fed.Res.Bull. at 507 n. 8. It is true that the application further provided some examples of the specific data processing uses to which the requested authority would be put:

Such packaged financial systems permit customers at various locations to obtain the benefit of Citicorp’s financial information systems and financial analysis expertise with respect to such applications associated with banking as financial modeling, loan analysis, accounting and bookkeeping, budget and profitability analysis, portfolio record-keeping and analysis, foreign exchange exposure, general ledger, bond analysis, international trade settlement, and economic forecasting.

J.A. C-54 to C-55. The Board’s opinion discussed several of these examples,7 concluding that they in fact represented services provided by banks or were so similar to such services that banking organizations are particularly well equipped to provide them. But that can only be understood as an attempted demonstration that its data test is as successful in concrete application as it is in theory.

It is even clearer that the Regulation Y order employs a data test. No more is needed to establish this than the text of the new regulation which the order adopted. Permissible activities of bank holding companies now include:

Providing to others data processing and data transmission services, facilities (including data processing and data transmission hardware, software, documentation or operating personnel), data bases, or access to such services, facilities, or data bases, by any technological means, if:

(i) The data to be processed or furnished are financial, banking, or economic, and the services are provided pursuant to a written agreement so describing and limiting the services;

(ii) The facilities are designed, marketed, and operated for the processing and *312transmission of financial, banking, or economic data; and

(iii) The hardware provided in connection therewith is offered only in conjunction with software designed and marketed for the processing and transmission of financial, banking, or economic data, and where the general purpose hardware does not constitute more than 30 percent of the cost of any packaged offering.

12 C.F.R. § 225.25(b)(7). The Regulation Y order states that the amendment it adopts “will make it permissible for bank holding companies to engage in the data processing and transmission services the Board has approved by order in the [Citicorp] case.” Regulation Y Order, 47 Fed.Reg. 37,369. But as the footnote accompanying that, statement shows, this is a reference to the eight broad technological categories of service {e.g., timesharing) rather than to the particular uses within those categories {e.g., budgeting, bookkeeping, accounting) which petitioners complain a data test avoids. And when the Regulation Y order incorporates by reference “[t]he Board’s findings on the permissibility of the services involved ... set forth in detail in the Board’s [Citicorp] order,” id., that incorporation, like the original findings themselves, must be for the purpose of exemplifying that in concrete application the data test will produce services closely related to banking.

As the insignificant nature of the textual changes in the portion of Regulation Y dealing with the current issue indicates, the present amendment does not fundamentally alter the approach of that regulation to data processing ■ services. With regard to the data test point it is significant, we think, that the Board’s staff recommendation proposing what in substance became the 1971 Regulation Y described its approach as follows:

In our view, the real issue is what kind of data should banking organizations be permitted to process. The technology employed is not the subject of the Act.

Accordingly, we recommend that the Board shift the emphasis of its proposal from the method of processing data to the kind of data being processed. Under this recommendation, bank holding companies would be permitted to process banking, financial, or other economic data, regardless of the tool used in the processing.

Legal Division Memo to Board of Governors (June 7, 1971), Applicant’s Exhibit M-2 at 6. While this is the staff’s description rather than the Board’s we think it an accurate representation of what the Board did in 1971 and perpetuated in the present orders.

The Board’s brief suggests that the ex: elusiveness and allegedly impermissible generality of the data test are avoided by the fact that

[bjefore any holding company may engage in data processing services, it must file an application with the Board and receive a Board order determining that the ‘public benefits’ test of section 4(c)(8) [of the Act] has been satisfied. 12 C.F.R. § 225.4(a), (b). Since data processing activities for a particular company are subject to Board approval upon specific application, the Board will have the opportunity to ensure that no bank holding company's activities exceed the statutory standard.

Respondent’s Brief at 32. It is indeed true, as noted in the Regulation Y Order, 47 Fed.Reg. 37,371, that the Board makes the separate “public benefits” determination required under the Act8 by examining individual applications rather than by regulation.9 But it would be unlawful, in the *313course of that case-by-case examination for that separate purpose, to deny an applicant the “closely related” qualification which the regulation generally confers. Such action would require amendment of the regulation—so that the limitation upon the data test which the Board’s brief puts forward amounts to no more than the possibility that the rule can be amended to forbid what it now impermissibly (according to petitioners) permits. Reliance upon such a possibility would validate every invalid rule.

We must confront, therefore, the stark question whether thé data test, as the determinant of whether data processing services are closely related to banking, is arbitrary or capricious. We think not. It would of course be preferable, from the point of view of accuracy alone, to make every “closely related” determination on a more narrow, specific, case-by-case basis— to ask, as petitioners would have the Board do, whether data processing for budget analysis, for bookkeeping, and for accounting each separately qualifies for the exemption. Indeed, it would be evén more accurate to get even more specific, and to ask whether budget analysis for manufacturing entities, budget analysis for retail sales entities, and budget analysis for personal service entities each separately qualifies. But the whole point of rulemaking as opposed to adjudication (or of statutory law as opposed to case-by-case common law development) is to incur a small possibility of inaccuracy in exchange for a large increase in efficiency and predictability. What the present controversy comes down to is simply whether there is reasonable assurance that the activities embraced within the data test—not all of which have been individually examined or even yet foreseen—will be closely related to banking under one or more of the broad National Courier tests. We think that there is.

As a theoretical matter, to begin with, the test is appealing. The record of this proceeding amply demonstrates, if any demonstration is needed, that banks regularly develop and process for their customers large amounts of banking, financial and economic data, and that they do so (and will presumably continue to do so) through the most advanced technological means. Once that is acknowledged, it is difficult to envision how any provision of data processing services dealing with data of that particular type would not meet at least the second of the National Courier tests:

Banks generally provide services that are operationally or functionally so similar to the proposed services as to equip them particularly well to provide the proposed service.

National Courier, supra, 516 F.2d at 1237. Perhaps it may not be in the public interest for them to provide one or another of such services, for anticompetitive or other reasons. But that relates to the “public benefits” determination, which is made case-by-case rather than in the amended Regulation Y, and whose resolution in the Citicorp order is not under challenge on this appeal..

In addition to its theoretical reasonableness, there is the fact that the Board, in the course of this proceeding, considered specific applications of the principle to various specific data processing uses proposed by Citicorp—finding all of them to be within one of the National Courier criteria, and many to be within the most rudimentary criterion that “[bjanks generally have in fact provided the proposed services,” id. We find adequate support in the record for those conclusions.10

*314It is significant that petitioners have not proposed any alternative to the data test, evidently demanding instead that each new data processing technology, and even each new use of data processing (such as budgeting, financial modeling, accounting) be separately examined and approved for its “closely relatedness.” In a field such as data processing, this will not do; the predictability of a rulemaking approach to the issue is vital. As the ALJ in the present proceeding found:

In view of rapid technological evolution, it is virtually impossible for any firm to predict the data processing services which will be available or offered in the coming years.

Recommended Decision at 10, J.A. B-81. And as the Comptroller of the Currency noted in a rulemaking proceeding dealing with the data processing activities of national banks:

In addition to being difficult to draft, acceptable specific lists or examples of permissible activities would be very difficult to keep current. This difficulty is due to the rapidly changing nature of the data processing field. The current rate of change virtually ensures that any acceptable list would quickly be rendered obsolete. Thus, any acceptable list or set of examples would require repeated updating. Perhaps more significantly, such a list may discourage the development by national banks of new data processing services which, although part of the business of banking, are not on a list of permitted activities or similar in nature to those in a set of examples of permitted activities.

Data Processing by National Banks, 47 Fed.Reg. 46,526, 46,529 (1982). We think these considerations amply justify the approach the Board has taken, an approach to which the petitioners have suggested no feasible alternative.

Economic Data

We turn next to an objection of petitioners that relates not to the data test as such, but to the types of data which the data test embraces.

The previous version of Regulation Y permitted bank holding companies to engage in “storing and processing other banking, financial, or related economic data, such as performing payroll, accounts receivable or payable or billing services.” 12 C.F.R. § 225.4(a)(8)(h) (1972) (emphasis added). In the combined adjudication-rule-making here at issue, the ALJ’s Recommended Decision accepted Citicorp’s proposal that this be expanded to include the processing and transmission of “banking, financial and economic related data.” Recommended Decision at 49, J.A. B-120 (emphasis added). The Board rejected that *315recommendation, agreeing with petitioners here that “economic related data” was “too broad a category and includes data that are not closely related to banking.” Regulation Y Order, 47 Fed.Reg. 37,369; Citicorp Order, 68 Fed.Res.Bull. at 507 n. 8. It adopted, however, an expansion of the previous language to “banking, financial and economic [as opposed to economic-related] data,” and permitted Citicorp to engage in such activities. Id. This expansion of prior authority was based upon its conclusion that “the record in this proceeding supports a finding that banks process economic data.” Regulation Y Order, 47 Fed.Reg. 37,369. See also Citicorp Order, 68 Fed.Res.Bull. at 507 n. 8 (“[t]he record supports a finding that banks engage in the processing and transmission of economic data”).

We agree. The testimony amply establishes the proposition that banks have long served their customers by developing and making available information regarding the national and international economy, including economic projections useful for investment decisions. The existing Regulation Y itself permitted the provision of “general economic information and advice, general economic statistical forecasting services and industry studies.” 12 C.F.R. § 225.4(a)(5)(iv) (1982). The current amendment of Regulation Y, which merely enables banks to make such information available in a new and more useful fashion, thus clearly meets the second of the National Courier criteria, and perhaps the first as well. Petitioners complain that “[njowhere in the record has there been any demonstration that banks have historically provided economic services even approaching the sophistication, scale and business mode approved by the Board in its Orders.” Petitioners’ Brief at 42. Perhaps so, but that is progress. It is not the purpose of the Bank Holding Company Act restrictions to confine banks to the same level, or crudeness, or technological simplicity of services previously provided—but merely to services of a closely related nature.

We also reject petitioners’ complaint that the Board has not defined “economic data.” Petitioners’ Brief at 35-36. In the context of these orders the meaning of the term is clear enough. It includes, as petitioners fear, “agricultural matters, retail sales matters, housing matters, corpo.rate profit matters and anything ‘of value in banking and financial decisions.’ ” Id. at 36 (quoting Citicorp Reply Brief before the Board at 13 & n. 17).

Timesharing Services

Until the 1970s, data processing was performed in the “batch” mode, that is, the application of a computer program was applied to data recorded (keypunched) on small cards, which were physically delivered to the computer. Batch processing requires expensive data pick-up and delivery, often involves lengthy overall turnaround time, and permits only one job to be performed at a time. Technological advances now enable data to be transmitted electronically, typically over a telephone line, to and from the central processing unit and the user’s terminal. That capability plus other technological advances permit “timesharing”—the simultaneous use of a computer by many users, each of whom can interact with the computer, i.e., ask yes-no questions and receive immediate responses. The AU found timesharing particularly well suited to economic, financial, and banking operations. Recommended Decision at 9, J.A. B-80.

Timesharing was approved in both the Citicorp and the Regulation Y Orders, 68 Fed.Res.Bull. at 507; 47 Fed.Reg. 37,369. Of course one of petitioners’ objections to this action is that the approval of this new service on the basis of the data test was impermissible; we have addressed that in a more general context above. Petitioners also claim, however, that timesharing, particularly when combined with customer use (which the orders permit) of applications software for such functions as modeling, forecasting and statistical analysis residing on holding company computers (as opposed to requiring customer creation and use of *316their own applications software) enables customers to use bank computer systems for “open-ended and general, non-financial services,” Petitioners’ Brief at 28, such as “race track handicapping or employee evaluation,” id. at 27.

This objection is unfounded.- To begin with, the Act places a limitation upon the services that bank holding companies can offer, not upon the uses to which others choose to put them. If the services proposed here were so well adapted to such uses unrelated to banking as employee evaluation that it could reasonably be thought that the services were offered for that purpose it would be one thing; but these services are no more invalidated by the mere possibility of such use, “unknown even to the holding company itself,”' Petitioners’ Brief at 29, than was the service which we approved in National Courier invalidated by the obvious possibility that a customer might include some nonfinancial material in the courier packages. And such a possibility (at most) is all that petitioners established. They brought forward not a single instance of actual use of such services for such purposes. Even their theoretical horrible was refuted by testimony asserting that it would be prohibitively expensive to use financial programs for such nonfinancial purposes. Nov. 17, 1981 Tr. at 1178-79, 1214-15.

Moreover, as undocumented and as unrealistic as the petitioners’ fears on this score appear to be, the Board nonetheless included provisions in its orders to calm this concern. In the Citicorp order the Board required that “all proposed data processing services provided by Citicorp to others outside the holding company for banking, financial and economic data must be provided pursuant to a written agreement so describing and limiting the services.” Citicorp Order, 68 Fed.Res.Bull. at 507. The approval conferred by the amended Regulation Y applies only when “the data to be processed or furnished are financial, banking, or economic, and the services are provided pursuant to a written agreement so describing and limiting the services.” 12 C.F.R. § 225.25(b)(7)(i). These agreements will be subject to scrutiny in connection with the Board’s case-by-case “public benefits” determination. In addition, the amended Regulation Y specifically requires that the offered facilities (by which the Board means “data processing and transmission hardware, systems software, documentation and operating personnel,” Citicorp Order, 68 Fed.Res.Bull. at 507 n. 12) be “designed, marketed, and operated for the processing and transmission of financial, banking, or economic data,” 12 C.F.R. § 225.25(b)(7)(h), and the Citicorp order makes it clear that this imposes the obligation to “take the technical steps necessary to ensure” this result. 68 Fed.Res.Bull. at 508. It is unthinkable that any more should be required.

Hardware

In both the Citicorp order and the Regulation Y order, the Board approved bank holding company provision of data processing hardware to their customers. Hardware is the equipment used in data processing systems, such as the mainframe computer, terminals, printers, memory devices, and the like. Software is the coded instructions which control the way data is processed, for example, individual programs. See Recommended Decision at 6-7, J.A. B-77 to B-78. For present purposes, data processing hardware provided by bank holding companies can be divided into two types, which were approved subject to different conditions and must be discussed separately.

Specialized hardware is specifically designed “to provide a permissible data processing or transmission service[], and is not likely to be used, to any significant extent, for nonfinancial purposes.” Citicorp Order, 68 Fed.Res.Bull. at 508 n. 14. The prime example is the automated teller machine (ATM), which is designed to execute banking transactions and has special security features appropriate to that purpose. The Board found that such hardware, when “offered only in conjunction with software designed and marketed for *317the processing and transmission of financial, banking, or economic data,” 12 C.F.R. § 225.25(b)(7)(iii), meets the third National Courier test. Citicorp Order, 68 Fed.Res. Bull, at 508 & n. 14, 509; Regulation Y Order, 47 Fed.Reg. 37,370. As noted earlier, that test reads as follows:

Banks generally provide services that are so integrally related to the proposed services as to require their provision in a specialized form.

516 F.2d at 1237. Petitioners make several attacks upon this finding, none of which seems to us well taken.

First, they assert that the third National Courier test could not conceivably apply because by its terms it pertains to “services,” which the provision of computer hardware is not. Petitioners’ Brief at 54. But surely National Courier is not to be interpreted in that fashion. Since the case dealt with services (courier services) it framed all its tests in those terms. But the object of the tests was to determine what are, within the words of the Act, “activities ... closely related to banking,” 12 U.S.C. § 1843(c)(8) (emphasis added), and they are obviously meant to apply, mutatis mutandis, to all activities, including not just services but also sale of material and equipment. It cannot seriously be thought that a bank holding company’s sale of checkbooks stands in a less favored position, under National Courier than its provision of courier services.

Next, petitioners argue that the third National Courier test is not met because it requires that the new services (or, as we have said, other activities) be integrally related to “traditional bank services” such as “check collection, credit extension and deposit gathering.” Petitioners’ Brief at 53. As far as we can determine, this limitation has been created out of whole cloth. It does not appear in National Courier, which only requires that the integrally related services be services that “banks generally provide.” If one were to hang upon this phraseology, the issue would presumably come down to whether the relevant integrally related services are data processing services in general (which banks now do generally provide) or the new types of data processing services authorized by these orders (at least the most futuristic of which banks now do not generally provide). Or it might be argued, as intervenor Citicorp does, Citicorp Brief at 61, that “services which banks generally provide” really means, as shown by language elsewhere in the National Courier opinion,11 “services which banks are authorized to provide.”

But at this point it begins to become foolish to devote one’s analytic energy to a parsing of the National Courier tests as though they were the statute itself, instead of referring to the underlying intent of the “closely related” requirement of which National Courier is, and only purports to be, a partial elaboration. That is to say, surely one of the most significant elements of the National Courier criterion is its prologue. After noting that the 1970 amendments to the Bank Holding Company Act (which slightly altered the text of the “closely related” provision and added the “public benefits” requirement) mean that the “closely related” test “no longer bears the full load, and may now be thought of as setting off as forbidden to banks those activities which are so clearly of a purely commercial nature that the predominantly adverse effects of a bank’s engaging in them may be presumed,” 516 F.2d at 1237, we continued:

Against this background, and reminding ourselves that the matter is one expressly committed by the statute to the Board, we think we owe considerable deference to the Board’s judgment that a particular activity is “closely related to banking.” Rather than define that term with any precision, therefore, we simply require that the Board go about making its “closely related” decision in a rea*318soned fashion consistent with the legislative intent.

The Board must, we think, articulate the ways in which banking activities and the proposed activities are assertedly connected, and must determine, not arbitrarily or capriciously, that the connections are close. As to what kinds of connections may qualify, at least the following seem to us within the statutory intent.'...

Id. (emphasis added).

Whether or not the provision of specialized computer hardware comes within the third National Courier test, we think the Board came to its decision in a reasoned fashion consistent with the legislative intent, and that that decision is not arbitrary or capricious. The Board noted, with adequate record support, that customers do not buy software and hardware but data processing. Citicorp Order, 68 Fed. Res.Bull, at 508-09; Regulation Y Order, 47 Fed.Reg. 37,370. Moreover, the Recommended Decision which the Board adopted discussed the impact of large scale integrated (LSI) circuits upon data processing technology, which now permit simple software functions, and in the future will permit more complex software functions, to be built into the hardware; and enable some of the computing function to be removed from the central computer and located in a customized “micro-processor” or “minicomputer” located at the point of use. Recommended Decision at 7-8, J.A. B-78 to B-79. This is the way some data processing is conducted now, and much more will be conducted in the future. Finally, the Recommended Decision also noted, with adequate record support, that a software producer marketing a “package” of software plus hardware is able to get a manufacturer's discount on the hardware, and is thus able to provide the full service to the purchaser at a more competitive price. Recommended Decision at 8, J.A. B-79. On the basis of these factors the Board concluded that “the activity of providing software without the corresponding authority to provide related hardware is of questionable economic feasibility.” Citicorp Order, 68 Fed.Res.Bull. at 508 (footnote omitted). It is not the economic infeasibility in itself that impresses us, but the underlying cause of that infeasibility, recognized by the Board—that in both market contemplation and technological reality the service in question is a unitary one. The issue boils down, as Citicorp suggests, Citicorp Brief at 58, to almost “a tautology.” In effect, to authorize the provision of banking, financial and economic data processing is to authorize the provision of banking, financial and economic hardware and software. Whether it be considered “integrally related” to the authorized service or, perhaps more realistically, simply part of it, the provision of specialized hardware is reasonably included.

General purpose hardware is, as the name would suggest, hardware that is designed to perform data processing functions in addition to banking, financial and economic. The Board approved bank holding company provision of this hardware, subject to the same condition that it be offered only in conjunction with banking, financial or economic data software, and subject to the additional condition that it “not constitute more than 30 percent of the cost of any packaged offering.” 12 C.F.R. § 225.25(b)(7)(iii).12 The Board acknowledged that the sale of general purpose hardware “is not itself an activity that is closely related to banking,” but found that with the limitations the Board imposed it would be permissible as “incidental” to the provision of permissible data processing services, Citicorp Order, 68 Fed.Res.Bull. *319at 508; see also Regulation Y Order, 47 Fed.Reg. 37,370.

The notion of permissibility of “any incidental activities that are necessary to carry on” activities closely related to banking has been embodied in Regulation Y since 1971. 12 C.F.R. § 225.21(a)(2). We approved it in National Courier, noting that “[i]n enumerating the activities that could be carried on, [Congress] certainly could not have meant to forbid engagement in such other ‘incidental’ activities as were reasonably necessary to carrying out those that were enumerated.” 516 F.2d at 1240. We disapproved its application to bank courier handling of nonfinancial materials for the following reason:

The justification for the ‘incidental’ courier services ... is not, as far as we can tell, that the carriage of non-financially related material is in any way necessary to the successful operation of the courier service affiliates. Rather, it is that the provision of such service would serve ‘the convenience of the public.’

Id. (footnote omitted). Here, by contrast, the Board’s justification was precisely that the provision of banking, financial and economic data processing services could not be successful unless they were offered in conjunction with the necessary hardware. Even where the hardware itself was not required to be specialized or to contain any software—so that the hardware could not be regarded as in itself the provision of banking, financial or economic data processing—the nature of the data processing market was such, the Board found, that hardware and software were a single package. The provision of data processing software in isolation, even where no specialized or software-inclusive hardware was required, was “of questionable economic feasibility.” Citicorp Order, 68 Fed.Res.Bull. at 508, referred to in Regulation Y Order, 47 Fed.Reg. 37,370. There was record evidence to support this conclusion, and, as the excerpt from National Courier above suggests, the conclusion is sufficient to support the Board’s action. Indeed, the Fifth Circuit Court of Appeals has permitted an even lesser connection to sustain bank holding company provision of liability insurance as “incidental” to the closely-related-to-banking activity of providing property damage insurance for the collateral in bank loans. Alabama Association of Insurance Agents, supra. While that court found evidence to sustain the proposition that, “from the consumer’s point of view, packaged property damage and liability policies are more desirable than the same policies separately sold,” 533 F.2d at 245, it did not make the further finding that the sale of the one had been shown to be “of questionable economic feasibility” without the sale of the other. In that respect, Alabama Association goes further than we were prepared to go in National Courier, though there was, it must be acknowledged, the added factor that liability insurance in itself to some degree increased the security of the banks’ loans. In any case, we think that the reasoning of both National Courier and Alabama Association supports the proposition that the economic necessity of offering a service that is not closely related to banking in order to sell another service that is, justifies the provision of the one as “incidental” to the other.

There is an obvious limitation upon this principle: At some point the tail begins to wag the dog. If it should be found, for example, that data processing services cannot be sold in an economically feasible manner without manufacturing data processing hardware; and if the banks’ profits from the latter should exceed their profits from the former; surely the provision of data processing services would be incidental to hardware manufacture rather than viceversa. But the Board has adequately taken that limitation into account, by specifying, as described above, that the cost of the hardware (including both specialized and general purpose hardware) cannot exceed 30 percent of the cost of the package. The Board derived this figure by noting from the record that hardware costs for the data processing industry as a whole represent about 25 percent of total costs, Citicorp Order, 68 Fed.Res.Bull. at 509,—so that a “package” in which hardware accounted *320for about that percentage of the cost could reasonably be considered primarily a sale of data processing rather than a sale of processing hardware. We think that a reasonable way to proceed, and we cannot say that an element of a permissible service which constitutes less than one-third the cost of that service is not “incidental.”

III. Procedural Objection

Petitioners raise a procedural objection to the Citicorp order. They claim that the Board approved Citicorp’s request without stating its holding or reasoning regarding four components of that request, in violation of the APA’s requirement that agency orders in formal proceedings address “all the material issues of fact, law, or discretion presented on the record.” 5 U.S.C. § 557(c). They complain that the Board did not address the sale of applications software products, the provision of facilities management services, the sale of turnkey systems products, and the provision of software-related services.

We do not agree. What is a “material issue” that must be separately addressed obviously depends upon one’s view of the logical or operational differences between various aspects of the problem in question. If an appellant before this court chooses to analyze a freedom-of-speech issue by discussing verbs, nouns, adjectives and pronouns separately, our decision will not necessarily disregard a “material issue” if it ignores this distinction. So also here, petitioners cannot insist upon imposing their structure of analysis upon the Board. The characteristic feature of petitioners’ approach throughout this proceeding has been the insistence that each separate element of the provision of data processing services be analyzed separately; this is simply contrary to the Board’s view, which we have found reasonable, that “the particular technology by which a data processing activity is provided is not determinative” and the “permissible data processing activities may be provided by any technologically feasible method.” Regulation Y Order, 47 Fed.Reg. 37,369. The Board’s decision in the Citicorp order clearly disposed of the various elements petitioners complain of here, whether or not it addressed each of them separately; and the reasons for its dispositions are "fully set forth. That is all that reason or even the APA requires.

Applications software, which petitioners define as software that provides instructions for the performance of specific tasks, is part of the software included in on-site data processing, electronic funds transfer, and authentication programs, all of which were approved and fully discussed. Facilities management by employees of the bank holding company and “turnkey” systems (that is, systems including both hardware and software in which control is turned over to the purchaser) were described by the AU as the two modes of providing on-site data processing, Recommended Decision at 7-8, 12-13, J.A. B-78 to B-79, B-83 to B-84; that activity, in all of its modes, was approved and fully discussed. Citicorp Order, 68 Fed.Res.Bull. at 508-09. Finally, as to software related services, such as consulting and systems analysis: Citicorp’s application did not include these services as a separate activity, and neither we nor, judging by its brief, the Board, see Board’s Brief at 63, regards the operation of a consulting service to give advice regarding data processing to be the same thing as data processing, which is all that the orders approve. Of course some consultation and advice will be necessary as an incident to the sale of data processing, but authorization for that was neither separately sought nor is separately required, since it is obviously included within Regulation Y’s existing authorization for “incidental activities,” discussed above. We therefore reject the procedural objection to the Board’s action.

* * * # # *

The challenge which the Board faced in the current proceeding was the perennial regulator’s problem of adjusting the outlines of regulation to accommodate technological change. Virtually unprecedented difficulties of adjustment are presented by *321the torrent of recent scientific innovation in the fields of electronics and telecommunications, which possess a peculiar capability to destroy the categories of enterprise upon which regulation is based. Communications merges into data processing and data processing into banking. Intervenor California Bankers Clearing House Association perceptively observes, CBCHA Brief at 29 n. 20, that this proceeding parallels the Federal Communications Commission’s Second Computer Inquiry, the order resulting from which we upheld in Computer and Communications Industry Association v. FCC, 693 F.2d 198 (D.C.Cir.1982).

The record of the present proceeding displays a careful and conscientious effort by the Board to cope with these difficulties. We are not inclined to complicate its task further by attempting to exercise close and necessarily inexpert supervision of its judgments. That would be particularly inappropriate under a governing statute such as this one, which commits it to the Board to apply a standard of such inherent imprecision (“closely related to banking”) that a discretion of almost legislative scope was necessarily contemplated. If there is a problem in such broad delegation, it would assuredly not be solved by effectively taking the delegation from the Board and placing it in our own hands. Having assured ourselves that the Board has acted reasonably, consistently and with procedural regularity in giving content to the statutory standard, our task is at an end.

In addition to the major points discussed above, we have considered the other objections urged by petitioners and find them insubstantial. For these reasons, the petitions are

Denied.

Justice KAGAN delivered the opinion of the Court.

The Social Security Administration (SSA) provides benefits to individuals who cannot obtain work because of a physical or mental disability. To determine whether *1152an applicant is entitled to benefits, the agency may hold an informal hearing examining (among other things) the kind and number of jobs available for someone with the applicant's disability and other characteristics. The agency's factual findings on that score are "conclusive" in judicial review of the benefits decision so long as they are supported by "substantial evidence." 42 U.S.C. § 405(g).

This case arises from the SSA's reliance on an expert's testimony about the availability of certain jobs in the economy. The expert largely based her opinion on private market-survey data. The question presented is whether her refusal to provide that data upon the applicant's request categorically precludes her testimony from counting as "substantial evidence." We hold it does not.

I

Petitioner Michael Biestek once worked as a carpenter and general laborer on construction sites. But he stopped working after he developed degenerative disc disease, Hepatitis C, and depression. He then applied for social security disability benefits, claiming eligibility as of October 2009.

After some preliminary proceedings, the SSA assigned an Administrative Law Judge (ALJ) to hold a hearing on Biestek's application. Those hearings, as described in the Social Security Act, 49 Stat. 620, as amended, 42 U.S.C. § 301 et seq. , are recognizably adjudicative in nature. The ALJ may "receive evidence" and "examine witnesses" about the contested issues in a case. §§ 405(b)(1), 1383(c) (1)(A). But many of the rules governing such hearings are less rigid than those a court would follow. See Richardson v. Perales , 402 U.S. 389, 400-401, 91 S.Ct. 1420, 28 L.Ed.2d 842 (1971). An ALJ is to conduct a disability hearing in "an informal, non-adversarial manner." 20 CFR § 404.900(b) (2018) ; § 416.1400(b). Most notably, an ALJ may receive evidence in a disability hearing that "would not be admissible in court." §§ 404.950(c), 416.1450(c); see 42 U.S.C. §§ 405(b) (1), 1383(c)(1)(A).

To rule on Biestek's application, the ALJ had to determine whether the former construction laborer could successfully transition to less physically demanding work. That required exploring two issues. The ALJ needed to identify the types of jobs Biestek could perform notwithstanding his disabilities. See 20 CFR §§ 404.1560(c)(1), 416.960(c)(1). And the ALJ needed to ascertain whether those kinds of jobs "exist[ed] in significant numbers in the national economy." §§ 404.1560(c)(1), 416.960(c)(1) ; see §§ 404.1566, 416.966.

For guidance on such questions, ALJs often seek the views of "vocational experts." See §§ 404.1566(e), 416.966(e); SSA, Hearings, Appeals, and Litigation Law Manual I-2-5-50 (Aug. 29, 2014). Those experts are professionals under contract with SSA to provide impartial testimony in agency proceedings. See id. , at I-2-1-31.B.1 (June 16, 2016); id. , at I-2-5-48. They must have "expertise" and "current knowledge" of "[w]orking conditions and physical demands of various" jobs; "[k]nowledge of the existence and numbers of [those jobs] in the national economy"; and "[i]nvolvement in or knowledge of placing adult workers[ ] with disabilities[ ] into jobs." Id. , at I-2-1-31.B.1. Many vocational experts simultaneously work in the private sector locating employment for persons with disabilities. See C. Kubitschek & J. Dubin, Social Security Disability Law & Procedure in Federal Court § 3:89 (2019). When offering testimony, the experts may invoke not only publicly available sources but also "information obtained directly from employers" and data *1153otherwise developed from their own "experience in job placement or career counseling." Social Security Ruling, SSR 00-4p, 65 Fed. Reg. 75760 (2000).

At Biestek's hearing, the ALJ asked a vocational expert named Erin O'Callaghan to identify a sampling of "sedentary" jobs that a person with Biestek's disabilities, education, and job history could perform. Tr. 59 (July 21, 2015); see 20 CFR §§ 404.1567(a), 416.967(a) (defining a "sedentary" job as one that "involves sitting" and requires "lifting no more than 10 pounds"). O'Callaghan had served as a vocational expert in SSA proceedings for five years; she also had more than ten years' experience counseling people with disabilities about employment opportunities. See Stachowiak v. Commissioner of Social Security , 2013 WL 593825, *1 (E.D. Mich., Jan. 11, 2013) ; Record in No. 16-10422 (ED Mich.), Doc. 17-13, p. 1274 (resume). In response to the ALJ's query, O'Callaghan listed sedentary jobs "such as a bench assembler [or] sorter" that did not require many skills. Tr. 58-59. And she further testified that 240,000 bench assembler jobs and 120,000 sorter jobs existed in the national economy. See ibid.

On cross-examination, Biestek's attorney asked O'Callaghan "where [she was] getting those [numbers] from." Id. , at 71. O'Callaghan replied that they came from the Bureau of Labor Statistics and her "own individual labor market surveys." Ibid. The lawyer then requested that O'Callaghan turn over the private surveys so he could review them. Ibid. O'Callaghan responded that she wished to keep the surveys confidential because they were "part of [her] client files." Id ., at 72. The lawyer suggested that O'Callaghan could "take the clients' names out." Ibid. But at that point the ALJ interjected that he "would not require" O'Callaghan to produce the files in any form. Ibid. Biestek's counsel asked no further questions about the basis for O'Callaghan's assembler and sorter numbers.

After the hearing concluded, the ALJ issued a decision granting Biestek's application in part and denying it in part. According to the ALJ, Biestek was entitled to benefits beginning in May 2013, when his advancing age (he turned fifty that month) adversely affected his ability to find employment. See App. to Pet. for Cert. 19a, 112a-113a. But before that time, the ALJ held, Biestek's disabilities should not have prevented a "successful adjustment to other work." Id. , at 110a-112a. The ALJ based that conclusion on O'Callaghan's testimony about the availability in the economy of "sedentary unskilled occupations such as bench assembler [or] sorter."Id. , at 111a (emphasis deleted).

Biestek sought review in federal court of the ALJ's denial of benefits for the period between October 2009 and May 2013. On judicial review, an ALJ's factual findings-such as the determination that Biestek could have found sedentary work-"shall be conclusive" if supported by "substantial evidence." 42 U.S.C. § 405(g) ; see supra , at 1151. Biestek contended that O'Callaghan's testimony could not possibly constitute such evidence because she had declined, upon request, to produce her supporting data. See Plaintiff's Motion for Summary Judgment in No. 16-10422 (ED Mich.), Doc. 22, p. 23. But the District Court rejected that argument. See 2017 WL 1173775, *2 (Mar. 30, 2017). And the Court of Appeals for the Sixth Circuit affirmed. See Biestek v. Commissioner of Social Security , 880 F.3d 778 (2017). That court recognized that the Seventh Circuit had adopted the categorical rule Biestek proposed, precluding a vocational expert's testimony from qualifying as substantial if the expert had declined an applicant's request to *1154provide supporting data. See id. , at 790 (citing McKinnie v. Barnhart , 368 F.3d 907, 910-911 (2004) ). But that rule, the Sixth Circuit observed in joining the ranks of unconvinced courts, "ha[d] not been a popular export." 880 F.3d at 790 (internal quotation marks omitted).

And no more is it so today.

II

The phrase "substantial evidence" is a "term of art" used throughout administrative law to describe how courts are to review agency factfinding. T-Mobile South , LLC v. Roswell , 574 U.S. ----, ----, 135 S.Ct. 808, 815, 190 L.Ed.2d 679 (2015). Under the substantial-evidence standard, a court looks to an existing administrative record and asks whether it contains "sufficien[t] evidence" to support the agency's factual determinations. Consolidated Edison Co. v. NLRB , 305 U.S. 197, 229, 59 S.Ct. 206, 83 L.Ed. 126 (1938) (emphasis deleted). And whatever the meaning of "substantial" in other contexts, the threshold for such evidentiary sufficiency is not high. Substantial evidence, this Court has said, is "more than a mere scintilla." Ibid. ; see, e.g. , Perales , 402 U.S. at 401, 91 S.Ct. 1420 (internal quotation marks omitted). It means-and means only-"such relevant evidence as a reasonable mind might accept as adequate to support a conclusion." Consolidated Edison , 305 U.S. at 229, 59 S.Ct. 206. See Dickinson v. Zurko , 527 U.S. 150, 153, 119 S.Ct. 1816, 144 L.Ed.2d 143 (1999) (comparing the substantial-evidence standard to the deferential clearly-erroneous standard).

Today, Biestek argues that the testimony of a vocational expert who (like O'Callaghan) refuses a request for supporting data about job availability can never clear the substantial-evidence bar. See Brief for Petitioner 21-34. As that formulation makes clear, Biestek's proposed rule is categorical, rendering expert testimony insufficient to sustain an ALJ's factfinding whenever such a refusal has occurred.1 But Biestek hastens to add two caveats. The first is to clarify what the rule is not, the second to stress where its limits lie.

Biestek initially takes pains-and understandably so-to distinguish his argument from a procedural claim. Reply Brief 12-14. At no stage in this litigation, Biestek says, has he ever espoused "a free-standing procedural rule under which a vocational expert would always have to produce [her underlying data] upon request." Id. , at 2. That kind of rule exists in federal court: There, an expert witness must produce all data she has considered in reaching her conclusions. See Fed. Rule Civ. Proc. 26(a)(2)(B). But as Biestek appreciates, no similar requirement applies in SSA hearings. As explained above, Congress intended those proceedings to be "informal" and provided that the "strict rules of evidence, applicable in the courtroom, *1155are not to" apply. Perales , 402 U.S. at 400, 91 S.Ct. 1420 ; see 42 U.S.C. § 405(b)(1) ; supra , at 1152. So Biestek does not press for a "procedural rule" governing "the means through which an evidentiary record [must be] created." Tr. of Oral Arg. 6; Reply Brief 13. Instead, he urges a "substantive rule" for "assess[ing] the quality and quantity of [record] evidence"-which would find testimony like O'Callaghan's inadequate, when taken alone, to support an ALJ's factfinding. Id. , at 12.

And Biestek also emphasizes a limitation within that proposed rule. For the rule to kick in, the applicant must make a demand for the expert's supporting data. See Brief for Petitioner i, 5, 18, 40, 55; Tr. of Oral Arg. 25-26. Consider two cases in which vocational experts rely on, but do not produce, nonpublic information. In the first, the applicant asks for the data; in the second, not. According to Biestek, the expert's testimony in the first case cannot possibly clear the substantial-evidence bar; but in the second case, it may well do so, even though the administrative record is otherwise the same. And Biestek underscores that this difference in outcome has nothing to do with waiver or forfeiture: As he acknowledges, an applicant "cannot waive the substantial evidence standard." Id. , at 27. It is just that the evidentiary problem arises from the expert's refusal of a demand, not from the data's absence alone. In his words, the testimony "can constitute substantial evidence if unchallenged, but not if challenged." Reply Brief 18.

To assess Biestek's proposal, we begin with the parties' common ground: Assuming no demand, a vocational expert's testimony may count as substantial evidence even when unaccompanied by supporting data. Take an example. Suppose an expert has top-of-the-line credentials, including professional qualifications and many years' experience; suppose, too, she has a history of giving sound testimony about job availability in similar cases (perhaps before the same ALJ). Now say that she testifies about the approximate number of various sedentary jobs an applicant for benefits could perform. She explains that she arrived at her figures by surveying a range of representative employers; amassing specific information about their labor needs and employment of people with disabilities; and extrapolating those findings to the national economy by means of a well-accepted methodology. She answers cogently and thoroughly all questions put to her by the ALJ and the applicant's lawyer. And nothing in the rest of the record conflicts with anything she says. But she never produces her survey data. Still, her testimony would be the kind of evidence-far "more than a mere scintilla"-that "a reasonable mind might accept as adequate to support" a finding about job availability. Consolidated Edison , 305 U.S. at 229, 59 S.Ct. 206. Of course, the testimony would be even better-more reliable and probative-if she had produced supporting data; that would be a best practice for the SSA and its experts.2 And of course, a different (maybe less qualified) expert failing to produce such data might offer testimony that is so feeble, or contradicted, that it would fail to clear the substantial-evidence *1156bar. The point is only-as, again, Biestek accepts-that expert testimony can sometimes surmount that bar absent underlying data.

But if that is true, why should one additional fact-a refusal to a request for that data-make a vocational expert's testimony categorically inadequate? Assume that an applicant challenges our hypothetical expert to turn over her supporting data; and assume the expert declines because the data reveals private information about her clients and making careful redactions will take a fair bit of time. Nothing in the expert's refusal changes her testimony (as described above) about job availability. Nor does it alter any other material in the record. So if our expert's opinion was sufficient-i.e. , qualified as substantial evidence-before the refusal, it is hard to see why the opinion has to be insufficient afterward.

Biestek suggests two reasons for that non-obvious result. First, he contends that the expert's rejection of a request for backup data necessarily "cast[s her testimony] into doubt." Reply Brief 16. And second, he avers that the refusal inevitably "deprives an applicant of the material necessary for an effective cross-examination." Id. , at 2. But Biestek states his arguments too broadly-and the nuggets of truth they contain cannot justify his proposed flat rule.

Consider Biestek's claim about how an expert's refusal undercuts her credibility. Biestek here invokes the established idea of an "adverse inference": If an expert declines to back up her testimony with information in her control, then the factfinder has a reason to think she is hiding something. See id. , at 16 (citing cases). We do not dispute that possibility-but the inference is far from always required. If an ALJ has no other reason to trust the expert, or finds her testimony iffy on its face, her refusal of the applicant's demand for supporting data may properly tip the scales against her opinion. (Indeed, more can be said: Even if the applicant makes no demand, such an expert's withholding of data may count against her.) But if (as in our prior hypothetical example, see supra , at 1154 - 1156) the ALJ views the expert and her testimony as otherwise trustworthy, and thinks she has good reason to keep her data private, her rejection of an applicant's demand need not make a difference. So too when a court reviews the ALJ's decision under the deferential substantial-evidence standard. In some cases, the refusal to disclose data, considered along with other shortcomings, will prevent a court from finding that "a reasonable mind" could accept the expert's testimony. Consolidated Edison , 305 U.S. at 229, 59 S.Ct. 206. But in other cases, that refusal will have no such consequence. Even taking it into account, the expert's opinion will qualify as "more than a mere scintilla" of evidence supporting the ALJ's conclusion. Which is to say it will count, contra Biestek, as substantial.

And much the same is true of Biestek's claim that an expert's refusal precludes meaningful cross-examination. We agree with Biestek that an ALJ and reviewing court may properly consider obstacles to such questioning when deciding how much to credit an expert's opinion. See Perales , 402 U.S. at 402-406, 91 S.Ct. 1420. But Biestek goes too far in suggesting that the refusal to provide supporting data always interferes with effective cross-examination, or that the absence of such testing always requires treating an opinion as unreliable. Even without specific data, an applicant may probe the strength of testimony by asking an expert about (for example) her sources and methods-where she got the information at issue and how she analyzed it and derived her conclusions.

*1157See, e.g. , Chavez v. Berryhill , 895 F.3d 962, 969-970 (CA7 2018). And even without significant testing, a factfinder may conclude that testimony has sufficient indicia of reliability to support a conclusion about whether an applicant could find work. Indeed, Biestek effectively concedes both those points in cases where supporting data is missing, so long as an expert has not refused an applicant's demand. See supra , at 1154 - 1155. But once that much is acknowledged, Biestek's argument cannot hold. For with or without an express refusal, the absence of data places the selfsame limits on cross-examination.

Where Biestek goes wrong, at bottom, is in pressing for a categorical rule, applying to every case in which a vocational expert refuses a request for underlying data. Sometimes an expert's withholding of such data, when combined with other aspects of the record, will prevent her testimony from qualifying as substantial evidence. That would be so, for example, if the expert has no good reason to keep the data private and her testimony lacks other markers of reliability. But sometimes the reservation of data will have no such effect. Even though the applicant might wish for the data, the expert's testimony still will clear (even handily so) the more-than-a-mere-scintilla threshold. The inquiry, as is usually true in determining the substantiality of evidence, is case-by-case. See, e.g. , Perales , 402 U.S. at 399, 410, 91 S.Ct. 1420 (rejecting a categorical rule pertaining to the substantiality of medical reports in a disability hearing). It takes into account all features of the vocational expert's testimony, as well as the rest of the administrative record. And in so doing, it defers to the presiding ALJ, who has seen the hearing up close.

That much is sufficient to decide this case. Biestek petitioned us only to adopt the categorical rule we have now rejected. He did not ask us to decide whether, in the absence of that rule, substantial evidence supported the ALJ in denying him benefits. Accordingly, we affirm the Court of Appeals' judgment.

It is so ordered.

Justice SOTOMAYOR, dissenting.

The Court focuses on the propriety of a categorical rule that precludes private data that a vocational expert refuses to provide upon request from qualifying as " 'substantial evidence.' " See ante , at 1152. I agree with Justice GORSUCH that the question presented by this case encompasses an inquiry not just into the propriety of a categorical rule in such circumstances but also into whether the substantial-evidence standard was met in the narrower circumstances of Michael Biestek's case. See post , at 1161 - 1162 (dissenting opinion). For the reasons that Justice GORSUCH sets out, the vocational expert's conclusory testimony in this case, offered without even a hint of support, did not constitute substantial evidence.

Once Biestek established that he had impairments, the agency bore the burden of proving that work opportunities were available to someone with his disabilities and individual characteristics. 20 CFR § 416.912(b)(3) (2018). To meet that burden, the agency relied on a vocational expert's testimony that Biestek could qualify for one of 240,000 "bench assembler" jobs or 120,000 "sorter" jobs nationwide. Tr. 59 (July 21, 2015). The expert said that those numbers were based in part on her "professional experience." Id. , at 61. When Biestek's counsel understandably asked for more details, the expert said only that she got the numbers from a publicly available source as well as from her "own individual labor market surveys" that were part of confidential client files. Id. , at 71; see *1158id. , at 67, 71-72. Biestek's counsel asked if the names in the files could be redacted, but the administrative law judge (ALJ) interrupted and ruled that she would not require the surveys to be produced in redacted form. Id. , at 72; see also id ., at 67.

Perhaps the ALJ would have allowed Biestek's counsel to ask followup questions about the basis for the testimony at that point, and perhaps Biestek's counsel should have tried to do so. But a Social Security proceeding is "inquisitorial rather than adversarial." Sims v. Apfel , 530 U.S. 103, 110-111, 120 S.Ct. 2080, 147 L.Ed.2d 80 (2000) ; see 20 CFR §§ 404.900(b), 416.1400(b). The ALJ acts as "an examiner charged with developing the facts," Richardson v. Perales , 402 U.S. 389, 410, 91 S.Ct. 1420, 28 L.Ed.2d 842 (1971), and has a duty to "develop the arguments both for and against granting benefits," Sims , 530 U.S. at 111, 120 S.Ct. 2080 ; see also Social Security Ruling, SSR 00-4P, 65 Fed. Reg. 75760 (2000) (noting "the adjudicator's duty to fully develop the record"). Here, instead of taking steps to ensure that the claimant had a basis from which effective cross-examination could be made and thus the record could be developed, the ALJ cut off that process by intervening when Biestek's counsel asked about the possibility of redaction.

The result was that the expert offered no detail whatsoever on the basis for her testimony. She did not say whom she had surveyed, how many surveys she had conducted, or what information she had gathered, nor did she offer any other explanation of the data on which she relied. In conjunction with the failure to proffer the surveys themselves, the expert's conclusory testimony alone could not constitute substantial evidence to support the ALJ's factfinding.1

I agree with much of Justice GORSUCH's reasoning. I emphasize that I do not foreclose the possibility that a more developed record could justify an ALJ's reliance on vocational-expert testimony in some circumstances even if the expert does not produce records underlying that testimony on request. An expert may have legitimate reasons for not turning over data, such as the burden of gathering records or confidentiality concerns that redaction cannot address. In those circumstances, as the majority suggests, the agency may be able to support an expert's testimony in ways other than by providing underlying data, such as by offering a fulsome description of the data and methodology on which the expert relies. See ante , at 1162. The agency simply did not do so here.

Justice GORSUCH, with whom Justice GINSBURG joins, dissenting.

Walk for a moment in Michael Biestek's shoes. As part of your application for disability benefits, you've proven that you suffer from serious health problems and can't return to your old construction job. Like many cases, yours turns on whether a significant number of other jobs remain that someone of your age, education, and experience, and with your physical limitations, could perform. When it comes to that question, the Social Security Administration bears the burden of proof. To meet its burden in your case, the agency chooses to rest on the testimony of a vocational *1159expert the agency hired as an independent contractor. The expert asserts there are 120,000 "sorter" and 240,000 "bench assembler" jobs nationwide that you could perform even with your disabilities.

Where did these numbers come from? The expert says she relied on data from the Bureau of Labor Statistics and her own private surveys. But it turns out the Bureau can't be the source; its numbers aren't that specific. The source-if there is a source-must be the expert's private surveys. So you ask to see them. The expert refuses-she says they're part of confidential client files. You reply by pointing out that any confidential client information can be redacted. But rather than ordering the data produced, the hearing examiner, herself a Social Security Administration employee, jumps in to say that won't be necessary. Even without the data, the examiner states in her decision on your disability claim, the expert's say-so warrants "great weight" and is more than enough evidence to deny your application. Case closed. App. to Pet. for Cert. 111a-112a, 118a-119a.

Would you say this decision was based on "substantial evidence"? Count me with Judge Easterbrook and the Seventh Circuit in thinking that an agency expert's bottom-line conclusion, supported only by a claim of readily available evidence that she refuses to produce on request, fails to satisfy the government's statutory burden of producing substantial evidence of available other work. See Donahue v. Barnhart , 279 F.3d 441, 446 (CA7 2002) ; McKinnie v. Barnhart , 368 F.3d 907, 910-911 (CA7 2004) (per curiam ).

Start with the legal standard. The Social Security Act of 1935 requires the agency to support its conclusions about the number of available jobs with "substantial evidence." 42 U.S.C. § 405(g). Congress borrowed that standard from civil litigation practice, where reviewing courts may overturn a jury verdict when the record lacks "substantial evidence"-that is, evidence sufficient to permit a reasonable jury to reach the verdict it did. Much the same standard governs summary judgment and directed verdict practice today. See 2 K. Hickman & R. Pierce, Administrative Law § 10.2.1, pp. 1082-1085 (6th ed. 2019); Anderson v. Liberty Lobby, Inc. , 477 U.S. 242, 252, 106 S.Ct. 2505, 91 L.Ed.2d 202 (1986) ; NLRB v. Columbian Enameling & Stamping Co. , 306 U.S. 292, 300, 59 S.Ct. 501, 83 L.Ed. 660 (1939).

Next, consider what we know about this standard. Witness testimony that's clearly wrong as a matter of fact cannot be substantial evidence. See Scott v. Harris , 550 U.S. 372, 380, 127 S.Ct. 1769, 167 L.Ed.2d 686 (2007). Falsified evidence isn't substantial evidence. See, e.g ., Firemen's and Policemen's Civil Serv. Comm'n v. Brinkmeyer , 662 S.W.2d 953, 956 (Tex. 1984). Speculation isn't substantial evidence. See, e.g. , Cao He Lin v. Department of Justice , 428 F.3d 391, 400 (CA2 2005) ; Alpo Petfoods, Inc. v. NLRB , 126 F.3d 246, 250 (CA4 1997). And, maybe most pointedly for our purposes, courts have held that a party or expert who supplies only conclusory assertions fails this standard too. See, e.g. , Lujan v. National Wildlife Federation , 497 U.S. 871, 888, 110 S.Ct. 3177, 111 L.Ed.2d 695 (1990) ("The object of [summary-judgment practice] is not to replace conclusory allegations of the complaint or answer with conclusory allegations of an affidavit"); Regents of Univ. of Minn. v. AGA Medical Corp. , 717 F.3d 929, 941 (CA Fed. 2013) ("conclusory expert assertions cannot raise triable issues of material fact") (collecting cases); Mid-State Fertilizer Co. v. Exchange Nat. Bank of Chicago , 877 F.2d 1333, 1339 (CA7 1989) ("An expert who supplies nothing but a bottom line supplies nothing of value to the judicial *1160process"); Sea Robin Pipeline Co. v. FERC , 795 F.2d 182, 188 (CADC 1986) ("[I]nordinate faith in the conclusory assertions of an expert ... cannot satisfy the requirement [of] substantial evidence").

If clearly mistaken evidence, fake evidence, speculative evidence, and conclusory evidence aren't substantial evidence, the evidence here shouldn't be either. The case hinges on an expert who (a) claims to possess evidence on the dispositive legal question that can be found nowhere else in the record, but (b) offers only a conclusion about its contents, and (c) refuses to supply the evidence when requested without showing that it can't readily be made available. What reasonable factfinder would rely on evidence like that? It seems just the sort of conclusory evidence courts have long held insufficient to meet the substantial evidence standard. And thanks to its conclusory nature, for all anyone can tell it may have come out of a hat-and, thus, may wind up being clearly mistaken, fake, or speculative evidence too. Unsurprisingly given all this, the government fails to cite even a single authority blessing the sort of evidence here as substantial evidence, despite the standard's long history and widespread use.

Veteran Social Security practitioners must be feeling a sense of déjà vu . Half a century ago, Judge Henry Friendly encountered Kerner v. Flemming , 283 F.2d 916 (CA2 1960). There, the agency's hearing examiner offered "nothing save [his own] speculation" to support his holding that the claimant "could in fact obtain substantial gainful employment." Id. , at 921. The Second Circuit firmly explained that this kind of conclusory claim is insufficient to meet the substantial evidence standard. In response, the Social Security Administration began hiring vocational experts, like the one in this case, to document the number of jobs available to a given claimant. But if the government can do what it did in this case, it's hard to see what all the trouble was for. The agency might still rest decisions on a hunch-just so long as the hunch comes from an agency contractor rather than an agency examiner.

Instead of addressing the realities of this case, the government asks us to imagine a hypothetical one. Assume, it says, that no one had requested the underlying data. In those circumstances, the government points out, even Mr. Biestek appears to accept that the agency's decision could have stood. And if that's true, the government asks, why should it make a difference if we add only one additional fact-the expert's refusal to produce the data? See ante , at 1154 - 1156 (presenting the same argument).

The answer is an old and familiar one. The refusal to supply readily available evidentiary support for a conclusion strongly suggests that the conclusion is, well, unsupported. See, e.g. , Interstate Circuit, Inc. v. United States , 306 U.S. 208, 226, 59 S.Ct. 467, 83 L.Ed. 610 (1939) ("The production of weak evidence when strong is available can lead only to the conclusion that the strong would have been adverse"); Clifton v. United States , 4 How. 242, 248, 11 L.Ed. 957 (1846) (the withholding of "more direct" proof suggests that "if the more perfect exposition had been given it would have laid open deficiencies and objections which the more obscure and uncertain testimony was intended to conceal"); 31A C.J. S., Evidence § 156(2), p. 402 (1964) ("The unfavorable inference ... is especially applicable where the party withholding the evidence has had notice or has been ordered to produce it"). Meanwhile, a similar inference may not arise if no one's bothered to ask for the evidence, or if the evidence is shown to be unavailable for a good reason. In cases like those, there may be just too many other plausible *1161and innocent excuses for the evidence's absence. Maybe, for example, nobody bothered to seek the underlying data because everyone knew what it would show.

Fine, the Court responds, all that's true enough. But even if we accept that an expert's failure to produce the evidence underlying her conclusion may support an inference that her conclusion is unsupported, that doesn't mean such an inference must follow. Whether an inference is appropriate depends on the facts of the particular case. See ante , at 1156 - 1157.

But what more do we need to know about the facts of this case? All of the relevant facts are undisputed, and it remains only to decide the legal question whether they meet the substantial evidence standard. We know that the expert offered a firm and exact conclusion about the number of available jobs. We know that the expert claimed to have private information to support her conclusion. We know Mr. Biestek requested that information and we have no reason to think any confidentiality concerns could not have been addressed. We know, too, that the hearing examiner had "no other reason to trust the expert['s]" numbers beyond her say-so. Ibid. Finally and looking to the law, we know that a witness's bare conclusion is regularly held insufficient to meet the substantial evidence threshold-and we know that the government hasn't cited a single case finding substantial evidence on so little. This is exactly the sort of case where an adverse inference should "tip the scales." Ibid.

With so much now weighing against the government, everything seems to turn on a final hypothetical. Now we are asked to imagine that the expert had offered detailed oral testimony about the withheld data. Her testimony was so detailed, we are asked to suppose, that Mr. Biestek could have thoroughly tested the data's reliability through cross-examination. (You might wonder just how effective this cross-examination could be if Mr. Biestek didn't have access to the data. But overlook that.) Surely in those circumstances it wouldn't matter whether the expert failed to produce the data even in bad faith. Any failure to produce would be harmless as a matter of law because the expert's testimony, all by itself, would amount to substantial evidence on which a rational factfinder might rely. Ante , at 1156 - 1157.

The problem is that this imaginary case has nothing to teach us about our real one. In Mr. Biestek's case, it is undisputed that the expert offered only a bare conclusion about the number of available jobs. No other relevant testimony was offered or received: no testimony about the underlying data, no testimony about its specific sources, no testimony about its reliability. In our real case, there is simply no way to shrug off the failure to produce the data as harmless error. To the contrary, and as we have seen, cases like this routinely fail to satisfy the substantial evidence standard. And if the government has a "duty to fully develop the record," ante , at 1158 (SOTOMAYOR, J., dissenting), that conclusion should follow all the more strongly.

What leads the Court to a different conclusion? It says that it views Mr. Biestek's petition as raising only the "categorical" question whether an expert's failure to produce underlying data always and in "every case" precludes her testimony from qualifying as substantial evidence. Ante , at 1152, 1156 - 1157. And once the question is ratcheted up to that level of abstraction, of course it is easy enough to shoot it down: just point to a series of hypothetical cases where the record contains additional justification for the expert's failure to produce or additional evidence to support her opinion. In such counterfactual cases, the failure to produce either would not be enough *1162to give rise to an adverse inference under traditional legal principles or could be held harmless as a matter of law. See ante , at 1154 - 1157.

But as I understand Mr. Biestek's submission, it does not require an all-or-nothing approach that would cover "every case." As the Court acknowledges, Mr. Biestek has focused us "on the Seventh Circuit's categorical rule." Ante , at 1154, n. 1. And that "rule" targets the narrower "category" of circumstances we have here-where an expert " 'give[s] a bottom line,' " fails to provide evidence "underlying that bottom line" when challenged, and fails to show the evidence is unavailable. McKinnie , 368 F.3d at 911 (quoting Donahue , 279 F.3d at 446 ). What to do about that category falls well within the question presented: "[w]hether a vocational expert's testimony can constitute substantial evidence of 'other work' ... when the expert fails upon the applicant's request to provide the underlying data on which that testimony is premised." Pet. for Cert. i. The answer to that question may be "always," "never," or-as the Court itself seems to acknowledge-"[s]ometimes." Ante , at 1157. And if the answer is "sometimes," the critical question becomes "in what circumstances"?

I suppose we could stop short and leave everyone guessing. But another option is to follow the Seventh Circuit's lead, resolve the smaller yet still significant "category" of cases like the one before us, and in that way begin to offer lower courts meaningful guidance in this important area. While I would not hesitate to take this course and make plain that cases like Mr. Biestek's fail the substantial evidence standard, I understand the Court today to choose the first option and leave these matters for another day.

There is good news and bad news in this. If my understanding of the Court's opinion is correct, the good news is that the Court remains open to the possibility that in real-world cases like Mr. Biestek's, lower courts may-and even should-find the substantial evidence test unmet. The bad news is that we must wait to find out, leaving many people and courts in limbo in the meantime. Cases with facts like Mr. Biestek's appear to be all too common. See, e.g. , Dubin, Overcoming Gridlock: Campbell After a Quarter-Century and Bureaucratically Rational Gap-Filling in Mass Justice Adjudication in the Social Security Administration's Disability Programs, 62 Admin. L. Rev. 937, 966 (2010). And many courts have erred in them by finding the substantial evidence test met, as the Sixth Circuit did in the case now before us. Some courts have even conflated the substantial evidence standard-a substantive standard governing what's needed to sustain a judgment as a matter of law-with procedural rules governing the admission of evidence. These courts have mistakenly suggested that, because the Federal Rules of Evidence don't apply in Social Security proceedings, anything an expert says will suffice to meet the agency's burden of proof. See, e.g ., Welsh v. Commissioner of Social Security , 662 Fed. Appx. 105, 109-110 (CA3 2016) ; Bayliss v. Barnhart , 427 F.3d 1211, 1218, and n. 4 (CA9 2005). Definitively resolving this case would have provided more useful guidance for practitioners and lower courts that have struggled with a significant category of cases like Mr. Biestek's, all while affording him the relief the law promises in disputes like his.

The principle that the government must support its allegations with substantial evidence, not conclusions and secret evidence, guards against arbitrary executive decisionmaking. See Friendly, " Some Kind of Hearing," 123 U. Pa. L. Rev. 1267, 1313-1314 (1975). Without it, people like Mr. *1163Biestek are left to the mercy of a bureaucrat's caprice. Over 100 years ago, in ICC v. Louisville & Nashville R. Co. , 227 U.S. 88, 33 S.Ct. 185, 57 L.Ed. 431 (1913), the government sought to justify an agency order binding private parties without producing the information on which the agency had relied. The government argued that its findings should be "presumed to have been supported." Id. , at 93, 33 S.Ct. 185. In essence, the government sought the right to "act upon any sort of secret evidence." Gellhorn, Official Notice in Administrative Adjudication, 20 Texas L. Rev. 131, 145 (1941). This Court did not approve of that practice then, and I would not have hesitated to make clear that we do not approve of it today.

I respectfully dissent.

Justice O’Connor

delivered the opinion of the Court.

This case presents the question whether ultimate consumers of dairy products may obtain judicial review of milk market orders issued by the Secretary of Agriculture (Secretary) under the authority of the Agricultural Marketing Agreement Act of 1937 (Act), ch. 296, 50 Stat. 246, as amended, 7 U. S. C. § 601 et seq. We conclude that consumers may not obtain judicial review of such orders.

I-H

A

In the early 1900’s, dairy farmers engaged in intense competition in the production of fluid milk products. See Zuber v. Allen, 396 U. S. 168, 172-176 (1969). To bring this destabilizing competition under control, the 1937 Act authorizes the Secretary to issue milk market orders setting the minimum prices that handlers (those who process dairy products) *342must pay to producers (dairy farmers) for their milk products. 7 U. S. C. §608c. The “essential purpose [of this milk market order scheme is] to raise producer prices,” S. Rep. No. 1011, 74th Cong., 1st Sess., 3 (1935), and thereby to ensure that the benefits and burdens of the milk market are fairly and proportionately shared by all dairy farmers. See Nebbia v. New York, 291 U. S. 502, 517-518 (1934).

Under the scheme established by Congress, the Secretary must conduct an appropriate rulemaking proceeding before issuing a milk market order. The public must be notified of these proceedings and provided an opportunity for public hearing and comment. See 7 U. S. C. § 608c(3). An order may be issued only if the evidence adduced at the hearing shows “that [it] will tend to effectuate the declared policy of this chapter with respect to such commodity.” 7 U. S. C. § 608c(4). Moreover, before any market order may become effective, it must be approved by the handlers of at least 50% of the volume of milk covered by the proposed order and at least two-thirds of the affected dairy producers in the region. 7 U. S. C. §§ 608c(8), 608c(5)(B)(i). If the handlers withhold their consent, the Secretary may nevertheless impose the order. But the Secretary’s power to do so is conditioned upon at least two-thirds of the producers consenting to its promulgation and upon his making an administrative determination that the order is “the only practical means of advancing the interests of the producers.” 7 U. S. C. § 608c(9)(B).

The Secretary currently has some 45 milk market orders in effect. See 7 CFR pts. 1001-1139 (1984). Each order covers a different region of the country, and collectively they cover most, though not all, of the United States. The orders divide dairy products into separately priced classes based on the uses to which raw milk is put. See 44 Fed. Reg. 65990 (1979). Raw milk that is processed and bottled for fluid consumption is termed “Class I” milk. Raw milk that is used to *343produce milk products such as butter, cheese, or dry milk powder is termed “Class II” milk.1

For a variety of economic reasons, fluid milk products would command a higher price than surplus milk products in a perfectly functioning market. Accordingly, the Secretary’s milk market orders require handlers to pay a higher order price for Class I products than for Class II products. To discourage destabilizing competition among producers for the more desirable fluid milk sales, the orders also require handlers to submit their payments for either class of milk to a regional pool. Administrators of these regional pools are then charged with distributing to dairy farmers a weighted average price for each milk product they have produced, irrespective of its use. See 7 U. S. C. §608c(5)(B)(ii).

In particular, the Secretary has regulated the price of “reconstituted milk” — that is, milk manufactured by mixing milk powder with water — since 1964. See 29 Fed. Reg. 9002, 9010 (1964); see also 34 Fed. Reg. 16548, 16551 (1969). The Secretary’s orders assume that handlers will use reconstituted milk to manufacture surplus milk products. Handlers are therefore required to pay only the lower Class II minimum price. See 44 Fed. Reg. 65989, 65990 (1979). However, handlers are required to make a “compensatory payment” on any portion of the reconstituted milk that their records show has not been used to manufacture surplus milk products. 7 CFR §§ 1012.44(a)(5)(i), 1012.60(e) (1984). The compensatory payment is equal to the difference between the Class I and Class II milk product prices. Handlers make these payments to the regional pool, from which moneys are then distributed to producers of fresh fluid milk in the region where the reconstituted milk was manufactured and sold. § 1012.71(a)(1).

*344B

In December 1980, respondents brought suit in District Court, contending that the compensatory payment requirement makes reconstituted milk uneconomical for handlers to process.2 Respondents, as plaintiffs in the District Court, included three individual consumers of fluid dairy products, a handler regulated by the market orders, and a nonprofit organization. The District Court concluded that the consumers and the nonprofit organization did not have standing to challenge the market orders. In addition, it found that Congress had intended by the Act to preclude such persons from obtaining judicial review. The District Court dismissed the milk handler’s complaint because he had failed to exhaust his administrative remedies.

The Court of Appeals affirmed in part and reversed in part, and remanded the case for a decision on the merits. 225 U. S. App. D. C. 387, 698 F. 2d 1239 (1983). The Court of Appeals agreed that the milk handler and the nonprofit organization had been properly dismissed by the District Court. But the court concluded that the individual consumers had standing: they had suffered an injury-in-fact, *345their injuries were redressable, and they were within the zone of interests arguably protected by the Act. The Court also concluded that the statutory structure and purposes of the Act did not reveal “the type of clear and convincing evidence of congressional intent needed to overcome the presumption in favor of judicial review.” Id., at 400, and n. 75, 698 F. 2d, at 1252, and n. 75. The Court of Appeals expressly refused to follow the decision of the Ninth Circuit in Rasmussen v. Hardin, 461 F. 2d 595, cert. denied sub nom. Rasmussen v. Butz, 409 U. S. 933 (1972), which had held consumers precluded by statute from seeking judicial review.

We granted certiorari to resolve the conflict in the Circuits. 464 U. S. 991 (1983). We now reverse the judgment of the Court of Appeals in this case.

f-H I — (

Respondents filed this suit under the Administrative Procedure Act (APA), 5 U. S. C. § 701 et seq. The APA confers a general cause of action upon persons “adversely affected or aggrieved by agency action within the meaning of a relevant statute,” 5 U. S. C. § 702, but withdraws that cause of action to the extent the relevant statute “precluded] judicial review,” 5 U. S. C. § 701(a)(1). Whether and to what extent a particular statute precludes judicial review is determined not only from its express language, but also from the structure of the statutory scheme, its objectives, its legislative history, and the nature of the administrative action involved. See Southern R. Co. v. Seaboard Allied Mining Corp., 442 U. S. 444, 454-463 (1979); Morris v. Gressette, 432 U. S. 491, 499-507 (1977); see generally Note, Statutory Preclusion of Judicial Review Under the Administrative Procedure Act, 1976 Duke L. J. 431, 442-449. Therefore, we must examine this statutory scheme “to determine whether Congress precluded all judicial review, and, if not, whether Congress nevertheless foreclosed review to the class to which the [re*346spondents] belon[g].” Barlow v. Collins, 397 U. S. 159, 173 (1970) (opinion of Brennan, J.,); see also Data Processing Service v. Camp, 397 U. S. 150, 156 (1970).

It is clear that Congress did not intend to strip the judiciary of all authority to review the Secretary’s milk market orders. The Act’s predecessor, the Agricultural Adjustment Act of 1933, 48 Stat. 31, contained no provision relating to administrative or judicial review. In 1935, however, Congress added a mechanism by which dairy handlers could obtain review of the Secretary’s market orders. 49 Stat. 760. That mechanism was retained in the 1937 legislation and remains in the Act as §608c(15) today. Section 608c(15) requires handlers first to exhaust the administrative remedies made available by the Secretary. 7 U. S. C. § 608c(15)(A); see 7 CFR §§900.50-900.71 (1984). After these formal administrative remedies have been exhausted, handlers may obtain judicial review of the Secretary’s ruling in the federal district court in any district “in which [they are] inhabitants], or ha[ve their] principal place[s] of business.” 7 U. S. C. § 608c(15)(B). These provisions for handler-initiated review make evident Congress’ desire that some persons be able to obtain judicial review of the Secretary’s market orders.

The remainder of the statutory scheme, however, makes equally clear Congress’ intention to limit the classes entitled to participate in the development of market orders. The Act contemplates a cooperative venture among the Secretary, handlers, and producers the principal purposes of which are to raise the price of agricultural products and to establish an orderly system for marketing them. Handlers and producers — but not consumers — are entitled to participate in the adoption and retention of market orders. 7 U. S. C. §§608c(8), (9), (16)(B). The Act provides for agreements among the Secretary, producers, and handlers, 7 U. S. C. §608(2), for hearings among them, §§608(5), 608c(3), and for votes by producers and handlers, §§608c(8)(A), (9)(B), (12), *347608c(19). Nowhere in the Act, however, is there an express provision for participation by consumers in any proceeding. In a complex scheme of this type, the omission of such a provision is sufficient reason to believe that Congress intended to foreclose consumer participation in the regulatory process. See Switchmen v. National Mediation Board, 320 U. S. 297, 305-306 (1943); cf. United States v. Erika, Inc., 456 U. S. 201, 208 (1982).

To be sure, the general purpose sections of the Act allude to general consumer interests. See 7 U. S. C. §§ 602(2), (4). But the preclusion issue does not only turn on whether the interests of a particular class like consumers are implicated. Rather, the preclusion issue turns ultimately on whether Congress intended for that class to be relied upon to challenge agency disregard of the law. See Barlow v. Collins, supra, at 167. The structure of this Act indicates that Congress intended only producers and handlers, and not consumers, to ensure that the statutory objectives would be realized.

Respondents would have us believe that, while Congress unequivocally directed handlers first to complain to the Secretary that the prices set by milk market orders are too high, it was nevertheless the legislative judgment that the same challenge, if advanced by consumers, does not require initial administrative scrutiny. There is no basis for attributing to Congress the intent to draw such a distinction. The regulation of agricultural products is a complex, technical undertaking. Congress channelled disputes concerning marketing orders to the Secretary in the first instance because it believed that only he has the expertise necessary to illuminate and resolve questions about them. Had Congress intended to allow consumers to attack provisions of marketing orders, it surely would have required them to pursue the administrative remedies provided in § 608c(15)(A) as well. The restriction of the administrative remedy to handlers strongly suggests that Congress intended a similar restriction of judicial review of market orders.

*348Allowing consumers to sue the Secretary would severely disrupt this complex and delicate administrative scheme. It would provide handlers with a convenient device for evading the statutory requirement that they first exhaust their administrative remedies. A handler may also be a consumer and, as such, could sue in that capacity. Alternatively, a handler would need only to find a consumer who is willing to join in or initiate an action in the district court. The consumer or consumer-handler could then raise precisely the same exceptions that the handler must raise administratively. Consumers or consumer-handlers could seek injunctions against the operation of market orders that “impede, hinder, or delay” enforcement actions, even though such injunctions are expressly prohibited in proceedings properly instituted under 7 U. S. C. §608c(15). Suits of this type would effectively nullify Congress’ intent to establish an “equitable and expeditious procedure for testing the validity of orders, without hampering the Government’s power to enforce compliance with their terms.” S. Rep. No. 1011, 74th Cong., 1st Sess., 14 (1935); see also United States v. Ruzicka, 329 U. S. 287, 293-294, and n. 3 (1946). For these reasons, we think it clear that Congress intended that judicial review of market orders issued under the Act ordinarily be confined to suits brought by handlers in accordance with 7 U. S. C. § 608c(15).

Ill

The Court of Appeals viewed the preclusion issue from a somewhat different perspective. First, it recited the presumption in favor of judicial review of administrative action that this Court usually employs. It then noted that the Act has been interpreted to authorize producer challenges to the administration of market order settlement funds, see Stark v. Wickard, 321 U. S. 288 (1944), and that no legislative history or statutory language directly and specifically supported the preclusion of consumer suits. In these circumstances, the Court of Appeals reasoned that the Act could not fairly be *349interpreted to overcome the presumption favoring judicial review and to leave consumers without a judicial remedy. See 225 U. S. App. D. C., at 400, and n. 75, 698 F. 2d, at 1252, and n. 75. We disagree with the Court of Appeals’ analysis.

The presumption favoring judicial review of administrative action is just that — a presumption. This presumption, like all presumptions used in interpreting statutes, may be overcome by specific language or specific legislative history that is a reliable indicator of congressional intent. See, e. g., Southern R. Co. v. Seaboard Allied Milling Corp., 442 U. S., at 454-463; Schilling v. Rogers, 363 U. S. 666, 670-677 (1960). The congressional intent necessary to overcome the presumption may also be inferred from contemporaneous judicial construction barring review and the congressional acquiescence in it, see, e. g., Ludecke v. Watkins, 335 U. S. 160 (1948), or from the collective import of legislative and judicial history behind a particular statute, see, e. g., Heikkila v. Barber, 345 U. S. 229 (1953). More important for purposes of this case, the presumption favoring judicial review of administrative action may be overcome by inferences of intent drawn from the statutory scheme as a whole. See, e. g., Morris v. Gressette, 432 U. S. 491 (1977); Switchmen v. National Mediation Board, 320 U. S. 297 (1943). In particular, at least when a statute provides a detailed mechanism for judicial consideration of particular issues at the behest of particular persons, judicial review of those issues at the behest of other persons may be found to be impliedly precluded. See Barlow v. Collins, 397 U. S., at 168, and n. 2, 175, and n. 9 (opinion of Brennan, J.); Switchmen v. National Mediation Board, supra, at 300-301; cf. Associated General Contractors of California, Inc. v. Carpenters, 459 U. S. 519, 542 (1983).

A case that best illustrates the relevance of a statute’s structure to the Court’s preclusion analysis is Morris v. Gressette, supra. In that case, the Court held that the Attorney General’s failure to object to a change in voting *350procedures was an unreviewable administrative determination under the Voting Rights Act of 1965. Neither the Voting Rights Act nor its legislative history said anything about judicial review. Nevertheless, the Morris Court concluded that the “nature of the [statutory] remedy . . . strongly suggests that Congress did not intend the Attorney General’s actions under that provision to be subject to judicial review.” Id., at 501. The Court reasoned that Congress had intended the approval procedure to be expeditious and that review-ability would unnecessarily extend the period the State must wait for effecting its change. Id., at 504-505. The Court also found relevant the existence of other remedies to ensure the realization of the Voting Rights Act’s objectives. Id., at 505-507. In these circumstances, even though proof of specific congressional intent was not “clear and convincing” in the traditional evidentiary sense, the Court unremarkably found the intent to preclude judicial review implicit in the statutory scheme.

In this case, the Court of Appeals did not take the balanced approach to statutory construction reflected in the Morris opinion. Rather, it recited this Court’s oft-quoted statement that “only upon a showing of ‘clear and convincing evidence’ of a contrary legislative intent should the courts restrict access to judicial review.” Abbott Laboratories v. Gardner, 387 U. S. 136, 141 (1967). See also Southern R. Co. v. Seaboard Allied Milling Corp., supra, at 462; Dunlop v. Bachowski, 421 U. S. 560, 568 (1975). According to the Court of Appeals, the “clear and convincing evidence” standard required it to find unambiguous proof, in the traditional evidentiary sense, of a congressional intent to preclude judicial review at the consumers’ behest. Since direct statutory language or legislative history on this issue could not be found, the Court of Appeals found the presumption favoring judicial review to be controlling.

This Court has, however, never applied the “clear and convincing evidence” standard in the strict evidentiary sense the *351Court of Appeals thought necessary in this case. Rather, the Court has found the standard met, and the presumption favoring judicial review overcome, whenever the congressional intent to preclude judicial review is “fairly discernible in the statutory scheme.” Data Processing Service v. Camp, 397 U. S., at 157. In the context of preclusion analysis, the “clear and convincing evidence” standard is not a rigid evidentiary test but a useful reminder to courts that, where substantial doubt about the congressional intent exists, the general presumption favoring judicial review of administrative action is controlling. That presumption does not control in cases such as this one, however, since the congressional intent to preclude judicial review is “fairly discernible” in the detail of the legislative scheme. Congress simply did not intend for consumers to be relied upon to challenge agency disregard of the law.

It is true, as the Court of Appeals also noted, that this Court determined, in Stark v. Wickard, 321 U. S. 288 (1944), that dairy producers could challenge certain administrative actions even though the Act did not expressly provide them a right to judicial review. The producers challenged certain deductions the Secretary had made from the “producer settlement fund” established in connection with the milk market order in effect at the time. “[T]he challenged deduction[s] reduce[d] pro tanto the amount actually received by the producers for their milk.” Id., at 302. These deductions injured what the producers alleged were “definite personal rights” that were “not possessed by the people generally,” id., at 304, 309, and gave the producers standing to object to the administration of the settlement fund. See id., at 306. Though the producers’ standing could not by itself ensure judicial review of the Secretary’s action at their behest, see ibid., the statutory scheme as a whole, the Court concluded, implicitly authorized producers’ suits concerning settlement fund administration. See id., at 309-310. “[Hjandlers [could not] question the use of the fund, because handlers had *352no financial interest in the fund or its use.” Id., at 308. Thus, there was “no forum” in which this aspect of the Secretary’s actions could or would be challenged. Judicial review of the producers’ complaint was therefore necessary to ensure achievement of the Act’s most fundamental objectives— to wit, the protection of the producers of milk and milk products.

By contrast, preclusion of consumer suits will not threaten realization of the fundamental objectives of the statute. Handlers have interests similar to those of consumers. Handlers, like consumers, are interested in obtaining reliable supplies of milk at the cheapest possible prices. See Zuber v. Allen, 396 U. S., at 190. Handlers can therefore be expected to challenge unlawful agency action and to ensure that the statute’s objectives will not be frustrated.3 Indeed, as noted above, consumer suits might themselves frustrate achievement of the statutory purposes. The Act contemplates a cooperative venture among the Secretary, producers, and handlers; consumer participation is not provided for or desired under the complex scheme enacted by Congress. Consumer suits would undermine the congressional preference for administrative remedies and provide a mechanism for disrupting administration of the congressional scheme. Thus, preclusion of consumer suits is perfectly consistent with the Court’s contrary conclusion concerning producer challenges in Stark v. Wickard and its analogous conclusion concerning voter challenges in Morris v. Gressette.

<1

The structure of this Act implies that Congress intended to preclude consumer challenges to the Secretary’s market orders. Preclusion of such suits does not pose any threat to *353realization of the statutory objectives; it means only that those objectives must be realized through the specific remedies provided by Congress and at the behest of the parties directly affected by the statutory scheme.4 Accordingly, the judgment of the Court of Appeals is reversed.

It is so ordered.

Justice Stevens took no part in the decision of this case.

Justice BREYER delivered the opinion of the Court.

The Leahy-Smith America Invents Act, 35 U.S.C. § 100 et seq., creates a process called "inter partes review." That review process allows a third party to ask the U.S. Patent and Trademark Office to reexamine the claims in an already-issued patent and to cancel any claim that the agency finds to be unpatentable in light of prior art. See § 102 (requiring "novel[ty]"); § 103 (disqualifying claims that are "obvious").

We consider two provisions of the Act. The first says:

"No Appeal.-The determination by the Director [of the Patent Office] whether to institute an inter partes review under this section shall be final and non-appealable." § 314(d).

Does this provision bar a court from considering whether the Patent Office wrongly "determin[ed] ... to institute an inter partes review," ibid., when it did so on grounds not specifically mentioned in a third party's review request?

The second provision grants the Patent Office the authority to issue

"regulations ... establishing and governing inter partes review under this chapter." § 316(a)(4).

Does this provision authorize the Patent Office to issue a regulation stating that the agency, in inter partes review,

"shall [construe a patent claim according to] its broadest reasonable construction in light of the specification of the patent in which it appears"? 37 CFR § 42.100(b) (2015).

We conclude that the first provision, though it may not bar consideration of a constitutional question, for example, does bar judicial review of the kind of mine-run claim at issue here, involving the Patent Office's decision to institute inter partes review. We also conclude that the second provision authorizes the Patent Office to issue the regulation before us. See, e.g., United States v. Mead Corp., 533 U.S. 218, 229, 121 S.Ct. 2164, 150 L.Ed.2d 292 (2001) ; Chevron, U.S.A. Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 842, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984).

I

A

An inventor obtains a patent by applying to the Patent Office. A patent examiner *2137with expertise in the relevant field reviews an applicant's patent claims, considers the prior art, and determines whether each claim meets the applicable patent law requirements. See, e.g., 35 U.S.C. §§ 101, 102, 103, 112. Then, the examiner accepts a claim, or rejects it and explains why. See § 132(a).

If the examiner rejects a claim, the applicant can resubmit a narrowed (or otherwise modified) claim, which the examiner will consider anew, measuring the new claim against the same patent law requirements. If the examiner rejects the new claim, the inventor typically has yet another chance to respond with yet another amended claim. Ultimately, the Patent Office makes a final decision allowing or rejecting the application. The applicant may seek judicial review of any final rejection. See §§ 141(a), 145.

For several decades, the Patent Office has also possessed the authority to reexamine-and perhaps cancel-a patent claim that it had previously allowed. In 1980, for example, Congress enacted a statute providing for "ex parte reexamination." Act to Amend the Patent and Trademark Laws, 35 U.S.C. § 301 et seq. That statute (which remains in effect) gives "[a]ny person at any time" the right to "file a request for reexamination" on the basis of certain prior art "bearing on the patentability" of an already-issued patent. §§ 301(a)(1), 302. If the Patent Office concludes that the cited prior art raises "a substantial new question of patentability," the agency can reexamine the patent. § 303(a). And that reexamination can lead the Patent Office to cancel the patent (or some of its claims). Alternatively, the Director of the Patent Office can, on her "own initiative," trigger such a proceeding. Ibid . And, as with examination, the patent holder can seek judicial review of an adverse final decision. § 306.

In 1999 and 2002, Congress enacted statutes that established another, similar procedure, known as "inter partes reexamination ." Those statutes granted third parties greater opportunities to participate in the Patent Office's reexamination proceedings as well as in any appeal of a Patent Office decision. See, e.g., American Inventors Protection Act of 1999, § 297 et seq. (2006 ed.) (superseded).

In 2011, Congress enacted the statute before us. That statute modifies "inter partes reexamination, " which it now calls "inter partes review ." See H.R.Rep. No. 112-98, pt. 1, pp. 46-47 (2011) (H.R. Rep.). Like inter partes reexamination, any third party can ask the agency to initiate inter partes review of a patent claim. But the new statute has changed the standard that governs the Patent Office's institution of the agency's process. Instead of requiring that a request for reexamination raise a "substantial new question of patentability," it now requires that a petition show "a reasonable likelihood that" the challenger "would prevail." Compare § 312(a) (2006 ed.) (repealed) with § 314(a) (2012 ed.).

The new statute provides a challenger with broader participation rights. It creates within the Patent Office a Patent Trial and Appeal Board (Board) composed of administrative patent judges, who are patent lawyers and former patent examiners, among others. § 6. That Board conducts the proceedings, reaches a conclusion, and sets forth its reasons. See ibid .

The statute sets forth time limits for completing this review. § 316(a)(11). It grants the Patent Office the authority to issue rules. § 316(a)(4). Like its predecessors, the statute authorizes judicial review of a "final written decision" canceling a patent claim. § 319. And, the statute says that the agency's initial decision "whether to institute an inter partes review" is "final and nonappealable." § 314(d) ; compare *2138ibid. with §§ 312(a), (c) (2006 ed.) (repealed) (the "determination" that a petition for inter partes reexamination "raise[s]" "a substantial new question of patentability" is "final and non-appealable"), and § 303(c) (2012 ed.) (similar in respect to ex parte reexamination).

B

In 2002, Giuseppe A. Cuozzo applied for a patent covering a speedometer that will show a driver when he is driving above the speed limit. To understand the basic idea, think of the fact that a white speedometer needle will look red when it passes under a translucent piece of red glass or the equivalent (say, red cellophane). If you attach a piece of red glass or red cellophane to a speedometer beginning at 65 miles per hour, then, when the white needle passes that point, it will look red. If we attach the red glass to a plate that can itself rotate, if we attach the plate to the speedometer, if we connect the plate to a Global Positioning System (GPS) receiver, and if we enter onto a chip or a disk all the speed limits on all the Nation's roads, then the GPS can signal where the car is, the chip or disk can signal the speed limit at that place, and the plate can rotate to the right number on the speedometer. Thus, if the speed limit is 35 miles per hour, then the white speedometer needle will pass under the red plate at 35, not 65, and the driver will know if he is driving too fast.

In 2004, the Patent Office granted the patent. See U.S. Patent No. 6,778,074 (Cuozzo Patent). The Appendix contains excerpts from this patent, offering a less simplified (and more technical) description.

C

Petitioner Cuozzo Speed Technologies, LLC (Cuozzo), now holds the rights to the Cuozzo Patent. In 2012, Garmin International, Inc., and Garmin USA, Inc., filed a petition seeking inter partes review of the Cuozzo Patent's 20 claims. Garmin backed up its request by stating, for example, that the invention described in claim 17 was obvious in light of three prior patents, the Aumayer, Evans, and Wendt patents. U.S. Patent No. 6,633,811 ; U.S. Patent No. 3,980,041 ; and U.S. Patent No. 2,711,153. Cf. Goodyear Tire & Rubber Co. v. Ray-O-Vac Co., 321 U.S. 275, 280, 64 S.Ct. 593, 88 L.Ed. 721 (1944) (Black, J., dissenting) ("[S]omeone, somewhere, sometime, made th [is] discovery [but] I cannot agree that this patentee is that discoverer").

The Board agreed to reexamine claim 17, as well as claims 10 and 14. The Board recognized that Garmin had not expressly challenged claim 10 and claim 14 on the same obviousness ground. But, believing that "claim 17 depends on claim 14 which depends on claim 10," the Board reasoned that Garmin had "implicitly" challenged claims 10 and 14 on the basis of the same prior inventions, and it consequently decided to review all three claims together. App. to Pet. for Cert. 188a.

After proceedings before the Board, it concluded that claims 10, 14, and 17 of the Cuozzo Patent were obvious in light of the earlier patents to which Garmin had referred. The Board explained that the Aumayer patent "makes use of a GPS receiver to determine ... the applicable speed limit at that location for display," the Evans patent "describes a colored plate for indicating the speed limit," and the Wendt patent "describes us[ing] a rotatable pointer for indicating the applicable speed limit." Id., at 146a-147a. Anyone, the Board reasoned, who is "not an automaton"-anyone with "ordinary skill" and "ordinary creativity"-could have taken the automated approach suggested by the Aumayer patent and applied it to the manually adjustable signals described in the Evans *2139and Wendt patents. Id., at 147a. The Board also concluded that Cuozzo's proposed amendments would not cure this defect, id., at 164a-166a, and it consequently denied Cuozzo's motion to amend its claims. Ultimately, it ordered claims 10, 14, and 17 of the Cuozzo Patent canceled, id., at 166a.

Cuozzo appealed to the United States Court of Appeals for the Federal Circuit. Cuozzo argued that the Patent Office improperly instituted inter partes review, at least in respect to claims 10 and 14, because the agency found that Garmin had only implicitly challenged those two claims on the basis of the Aumayer, Evans, and Wendt patents, while the statute required petitions to set forth the grounds for challenge "with particularity." § 312(a)(3). Cuozzo also argued that the Board, when construing the claims, improperly used the interpretive standard set forth in the Patent Office's regulation (i.e., it gave those claims their "broadest reasonable construction," 37 CFR § 42.100(b) ), when it should have applied the standard that courts normally use when judging a patent's validity (i.e., it should have given those claims their "ordinary meaning ... as understood by a person of skill in the art," Phillips v. AWH Corp., 415 F.3d 1303, 1314 (C.A.Fed.2005) (en banc)).

A divided panel of the Court of Appeals rejected both arguments. First, the panel majority pointed out that 35 U.S.C. § 314(d) made the decision to institute inter partes review "nonappealable." In re Cuozzo Speed Technologies, LLC, 793 F.3d 1268, 1273 (C.A.Fed.2015) (internal quotation marks omitted). Second, the panel majority affirmed the application of the broadest reasonable construction standard on the ground (among others) that the regulation was a reasonable, and hence lawful, exercise of the Patent Office's statutorily granted rulemaking authority. Id., at 1278-1279 ; see § 314(a)(4). By a vote of 6 to 5, the Court of Appeals denied Cuozzo's petition for rehearing en banc. In re Cuozzo Speed Technologies, LLC, 793 F.3d 1297, 1298 (C.A.Fed.2015).

We granted Cuozzo's petition for certiorari to review these two questions.

II

Like the Court of Appeals, we believe that Cuozzo's contention that the Patent Office unlawfully initiated its agency review is not appealable. For one thing, that is what § 314(d) says. It states that the "determination by the [Patent Office] whether to institute an inter partes review under this section shall be final and nonappealable. " (Emphasis added.)

For another, the legal dispute at issue is an ordinary dispute about the application of certain relevant patent statutes concerning the Patent Office's decision to institute inter partes review. Cuozzo points to a related statutory section, § 312, which says that petitions must be pleaded "with particularity." Those words, in its view, mean that the petition should have specifically said that claims 10 and 14 are also obvious in light of this same prior art. Garmin's petition, the Government replies, need not have mentioned claims 10 and 14 separately, for claims 10, 14, and 17 are all logically linked; the claims "rise and fall together," and a petition need not simply repeat the same argument expressly when it is so obviously implied. See 793 F.3d, at 1281. In our view, the "No Appeal" provision's language must, at the least, forbid an appeal that attacks a "determination ... whether to institute" review by raising this kind of legal question and little more. § 314(d).

Moreover, a contrary holding would undercut one important congressional objective, namely, giving the Patent Office significant *2140power to revisit and revise earlier patent grants. See H.R. Rep., at 45, 48 (explaining that the statute seeks to "improve patent quality and restore confidence in the presumption of validity that comes with issued patents"); 157 Cong. Rec. 9778 (2011) (remarks of Rep. Goodlatte) (noting that inter partes review "screen[s] out bad patents while bolstering valid ones"). We doubt that Congress would have granted the Patent Office this authority, including, for example, the ability to continue proceedings even after the original petitioner settles and drops out, § 317(a), if it had thought that the agency's final decision could be unwound under some minor statutory technicality related to its preliminary decision to institute inter partes review.

Further, the existence of similar provisions in this, and related, patent statutes reinforces our conclusion. See § 319 (limiting appellate review to the "final written decision"); § 312(c) (2006 ed.) (repealed) (the "determination" that a petition for inter partes reexamination "raise[s]" a "substantial new question of patentability" is "final and non-appealable"); see also § 303(c) (2012 ed.); In re Hiniker Co., 150 F.3d 1362, 1367 (C.A.Fed.1998) ("Section 303 ... is directed toward the [Patent Office's] authority to institute a reexamination, and there is no provision granting us direct review of that decision").

The dissent, like the panel dissent in the Court of Appeals, would limit the scope of the "No Appeal" provision to interlocutory appeals, leaving a court free to review the initial decision to institute review in the context of the agency's final decision. Post, at 2148 - 2149, 2150 - 2151 (ALITO, J., concurring in part and dissenting in part); 793 F.3d, at 1291 (Newman, J., dissenting). We cannot accept this interpretation. It reads into the provision a limitation (to interlocutory decisions) that the language nowhere mentions and that is unnecessary. The Administrative Procedure Act already limits review to final agency decisions. 5 U.S.C. § 704. The Patent Office's decision to initiate inter partes review is "preliminary," not "final." Ibid . And the agency's decision to deny a petition is a matter committed to the Patent Office's discretion. See § 701(a)(2); 35 U.S.C. § 314(a) (no mandate to institute review); see also post, at 2153, and n. 6. So, read as limited to such preliminary and discretionary decisions, the "No Appeal" provision would seem superfluous. The dissent also suggests that its approach is a "familiar practice," consistent with other areas of law. Post, at 2152. But the kind of initial determination at issue here-that there is a "reasonable likelihood" that the claims are unpatentable on the grounds asserted-is akin to decisions which, in other contexts, we have held to be unreviewable. See Kaley v. United States, 571 U.S. ----, ----, 134 S.Ct. 1090, 1097-1098, 188 L.Ed.2d 46 (2014) ("The grand jury gets to say-without any review, oversight, or second-guessing-whether probable cause exists to think that a person committed a crime").

We recognize the "strong presumption" in favor of judicial review that we apply when we interpret statutes, including statutes that may limit or preclude review. Mach Mining, LLC v. EEOC, 575 U.S. ----, ----, 135 S.Ct. 1645, 1650-1651, 191 L.Ed.2d 607 (2015) (internal quotation marks omitted). This presumption, however, may be overcome by " 'clear and convincing' " indications, drawn from "specific language," "specific legislative history," and "inferences of intent drawn from the statutory scheme as a whole," that Congress intended to bar review. Block v. Community Nutrition Institute, 467 U.S. 340, 349-350, 104 S.Ct. 2450, 81 L.Ed.2d 270 (1984). That standard *2141is met here. The dissent disagrees, and it points to Lindahl v. Office of Personnel Management, 470 U.S. 768, 105 S.Ct. 1620, 84 L.Ed.2d 674 (1985), to support its view that, in light of this presumption, § 314(d) should be read to permit judicial review of any issue bearing on the Patent Office's preliminary decision to institute inter partes review. See post, at 2150 - 2151. Lindahl is a case about the judicial review of disability determinations for federal employees. We explained that a statute directing the Office of Personnel Management to " 'determine questions of disability,' " and making those decisions " 'final,' " " 'conclusive,' " and " 'not subject to review,' " barred a court from revisiting the "factual underpinnings of ... disability determinations"-though it permitted courts to consider claims alleging, for example, that the Office of Personnel Management " 'substantial[ly] depart[ed] from important procedural rights.' " 470 U.S., at 771, 791, 105 S.Ct. 1620. Thus, Lindahl 's interpretation of that statute preserved the agency's primacy over its core statutory function in accord with Congress' intent. Our interpretation of the "No Appeal" provision here has the same effect. Congress has told the Patent Office to determine whether inter partes review should proceed, and it has made the agency's decision "final" and "nonappealable." § 314(d). Our conclusion that courts may not revisit this initial determination gives effect to this statutory command. Moreover, Lindahl 's conclusion was consistent with prior judicial practice in respect to those factual agency determinations, and legislative history "strongly suggest[ed]" that Congress intended to preserve this prior practice. Id., at 780, 105 S.Ct. 1620. These features, as explained above, also support our interpretation: The text of the "No Appeal" provision, along with its place in the overall statutory scheme, its role alongside the Administrative Procedure Act, the prior interpretation of similar patent statutes, and Congress' purpose in crafting inter partes review, all point in favor of precluding review of the Patent Office's institution decisions.

Nevertheless, in light of § 314(d)'s own text and the presumption favoring review, we emphasize that our interpretation applies where the grounds for attacking the decision to institute inter partes review consist of questions that are closely tied to the application and interpretation of statutes related to the Patent Office's decision to initiate inter partes review. See § 314(d) (barring appeals of "determinations ... to initiate an inter partes review under this section " (emphasis added)). This means that we need not, and do not, decide the precise effect of § 314(d) on appeals that implicate constitutional questions, that depend on other less closely related statutes, or that present other questions of interpretation that reach, in terms of scope and impact, well beyond "this section." Cf. Johnson v. Robison, 415 U.S. 361, 367, 94 S.Ct. 1160, 39 L.Ed.2d 389 (1974) (statute precluding review of "any question of law or fact under any law administered by the Veterans' Administration" does not bar review of constitutional challenges (emphasis deleted and internal quotation marks omitted)); Traynor v. Turnage, 485 U.S. 535, 544-545, 108 S.Ct. 1372, 99 L.Ed.2d 618 (1988) (that same statute does not bar review of decisions made under different statutes enacted at other times). Thus, contrary to the dissent's suggestion, we do not categorically preclude review of a final decision where a petition fails to give "sufficient notice" such that there is a due process problem with the entire proceeding, nor does our interpretation enable the agency to act outside its statutory limits by, for example, canceling a patent claim for "indefiniteness *2142under § 112" in inter partes review. Post, at 2153 - 2155. Such "shenanigans" may be properly reviewable in the context of § 319 and under the Administrative Procedure Act, which enables reviewing courts to "set aside agency action" that is "contrary to constitutional right," "in excess of statutory jurisdiction," or "arbitrary [and] capricious." Compare post, at 2155, with 5 U.S.C. §§ 706(2)(A)-(D).

By contrast, where a patent holder merely challenges the Patent Office's "determin[ation] that the information presented in the petition ... shows that there is a reasonable likelihood" of success "with respect to at least 1 of the claims challenged," § 314(a), or where a patent holder grounds its claim in a statute closely related to that decision to institute inter partes review, § 314(d) bars judicial review. In this case, Cuozzo's claim that Garmin's petition was not pleaded "with particularity" under § 312 is little more than a challenge to the Patent Office's conclusion, under § 314(a), that the "information presented in the petition" warranted review. Cf. United States v. Williams, 504 U.S. 36, 54, 112 S.Ct. 1735, 118 L.Ed.2d 352 (1992) ("A complaint about the quality or adequacy of the evidence can always be recast as a complaint that the ... presentation was 'incomplete' or 'misleading' "). We therefore conclude that § 314(d) bars Cuozzo's efforts to attack the Patent Office's determination to institute inter partes review in this case.

III

Cuozzo further argues that the Patent Office lacked the legal authority to issue its regulation requiring the agency, when conducting an inter partes review, to give a patent claim "its broadest reasonable construction in light of the specification of the patent in which it appears." 37 CFR § 42.100(b). Instead, Cuozzo contends that the Patent Office should, like the courts, give claims their "ordinary meaning ... as understood by a person of skill in the art." Phillips, 415 F.3d, at 1314.

The statute, however, contains a provision that grants the Patent Office authority to issue "regulations ... establishing and governing inter partes review under this chapter." 35 U.S.C. § 316(a)(4). The Court of Appeals held that this statute gives the Patent Office the legal authority to issue its broadest reasonable construction regulation. We agree.

A

We interpret Congress' grant of rulemaking authority in light of our decision in Chevron, U.S.A. Inc., 467 U.S. 837, 104 S.Ct. 2778, 81 L.Ed.2d 694. Where a statute is clear, the agency must follow the statute. Id., at 842-843, 104 S.Ct. 2778. But where a statute leaves a "gap" or is "ambigu [ous]," we typically interpret it as granting the agency leeway to enact rules that are reasonable in light of the text, nature, and purpose of the statute. Mead Corp., 533 U.S., at 229, 121 S.Ct. 2164 ; Chevron, U.S.A. Inc., supra, at 843, 104 S.Ct. 2778. The statute contains such a gap: No statutory provision unambiguously directs the agency to use one standard or the other. And the statute "express[ly] ... authoriz[es] [the Patent Office] to engage in the process of rulemaking" to address that gap. Mead Corp., supra, at 229, 121 S.Ct. 2164. Indeed, the statute allows the Patent Office to issue rules "governing inter partes review," § 316(a)(4), and the broadest reasonable construction regulation is a rule that governs inter partes review.

Both the dissenting judges in the Court of Appeals and Cuozzo believe that other ordinary tools of statutory interpretation, *2143INS v. Cardoza-Fonseca, 480 U.S. 421, 432, and n. 12, 107 S.Ct. 1207, 94 L.Ed.2d 434 (1987), lead to a different conclusion. The dissenters, for example, point to cases in which the Circuit interpreted a grant of rulemaking authority in a different statute, § 2(b)(2)(A), as limited to procedural rules. See, e.g., Cooper Technologies Co. v. Dudas, 536 F.3d 1330, 1335 (C.A.Fed.2008). These cases, however, as we just said, interpret a different statute. That statute does not clearly contain the Circuit's claimed limitation, nor is its language the same as that of § 316(a)(4). Section 2(b)(2)(A) grants the Patent Office authority to issue "regulations" "which ... shall govern ... proceedings in the Office " (emphasis added), but the statute before us, § 316(a)(4), does not refer to "proceedings"-it refers more broadly to regulations "establishing and governing inter partes review." The Circuit's prior interpretation of § 2(b)(2)(A) cannot magically render unambiguous the different language in the different statute before us.

Cuozzo and its supporting amici believe we will reach a different conclusion if we carefully examine the purpose of inter partes review. That purpose, in their view, is to modify the previous reexamination procedures and to replace them with a " 'trial, adjudicatory in nature.' " Brief for Petitioner 26 (quoting Google Inc. v. Jongerius Panoramic Techs., LLC, IPR 2013-00191, Paper No. 50, p. 4 (PTAB, Feb. 13, 2014)). They point out that, under the statute, an opposing party can trigger inter partes review. Parties can engage in "discovery of relevant evidence," including "deposition[s], ... affidavits or declarations" as well as anything "otherwise necessary in the interest of justice." § 316(a)(5). Parties may present "factual evidence and expert opinions" to support their arguments. § 316(a)(8). The challenger bears the burden of proving unpatentability. § 318(e). And, after oral argument before a panel of three of the Board's administrative patent judges, it issues a final written decision. §§ 6, 316(a)(10), 318. Perhaps most importantly, a decision to cancel a patent normally has the same effect as a district court's determination of a patent's invalidity.

In light of these adjudicatory characteristics, which make these agency proceedings similar to court proceedings, Congress, in Cuozzo's view, must have designed inter partes review as a "surrogate for court proceedings." Brief for Petitioner 28. Cuozzo points to various sources of legislative history in support of its argument. See H.R. Rep., at 48 (Inter partes review is a "quick and cost effective alternativ[e] to litigation"); id., at 46-47 ("The Act converts inter partes reexamination from an examinational to an adjudicative proceeding"); see also S.Rep. No. 110-259, p. 20 (2008) (Inter partes review is "a quick, inexpensive, and reliable alternative to district court litigation"); 157 Cong. Rec. 3429-3430 (2011) (remarks of Sen. Kyl) ("Among the reforms that are expected to expedite these proceedings [is] the shift from an examinational to an adjudicative model"). And, if Congress intended to create a "surrogate" for court proceedings, why would Congress not also have intended the agency to use the claim construction standard that district courts apply (namely, the ordinary meaning standard), rather than the claim construction standard that patent examiners apply (namely, the broadest reasonable construction standard)?

The problem with Cuozzo's argument, however, is that, in other significant respects, inter partes review is less like a judicial proceeding and more like a specialized agency proceeding. Parties that initiate the proceeding need not have a concrete stake in the outcome; indeed, they *2144may lack constitutional standing. See § 311(a) ; cf. Consumer Watchdog v. Wisconsin Alumni Research Foundation, 753 F.3d 1258, 1261-1262 (C.A.Fed.2014). As explained above, challengers need not remain in the proceeding; rather, the Patent Office may continue to conduct an inter partes review even after the adverse party has settled. § 317(a). Moreover, as is the case here, the Patent Office may intervene in a later judicial proceeding to defend its decision-even if the private challengers drop out. And the burden of proof in inter partes review is different than in the district courts: In inter partes review, the challenger (or the Patent Office) must establish unpatentability "by a preponderance of the evidence"; in district court, a challenger must prove invalidity by "clear and convincing evidence." Compare § 316(e) with Microsoft Corp. v. i4i Ltd. Partnership, 564 U.S. 91, 95, 131 S.Ct. 2238, 180 L.Ed.2d 131 (2011).

Most importantly, these features, as well as inter partes review's predecessors, indicate that the purpose of the proceeding is not quite the same as the purpose of district court litigation. The proceeding involves what used to be called a reexamination (and, as noted above, a cousin of inter partes review, ex parte reexamination, 35 U.S.C. § 302 et seq., still bears that name). The name and accompanying procedures suggest that the proceeding offers a second look at an earlier administrative grant of a patent. Although Congress changed the name from "reexamination" to "review," nothing convinces us that, in doing so, Congress wanted to change its basic purposes, namely, to reexamine an earlier agency decision. Thus, in addition to helping resolve concrete patent-related disputes among parties, inter partes review helps protect the public's "paramount interest in seeing that patent monopolies ... are kept within their legitimate scope." Precision Instrument Mfg. Co. v. Automotive Maintenance Machinery Co., 324 U.S. 806, 816, 65 S.Ct. 993, 89 L.Ed. 1381 (1945) ; see H.R. Rep., at 39-40 (Inter partes review is an "efficient system for challenging patents that should not have issued").

Finally, neither the statutory language, its purpose, or its history suggest that Congress considered what standard the agency should apply when reviewing a patent claim in inter partes review. Cuozzo contends that § 301(d), explaining that the Patent Office should "determine the proper meaning of a patent claim," reinforces its conclusion that the ordinary meaning standard should apply. But viewed against a background of language and practices indicating that Congress designed a hybrid proceeding, § 301(d)'s reference to the "proper meaning" of a claim is ambiguous. It leaves open the question of which claim construction standard is "proper."

The upshot is, whether we look at statutory language alone, or that language in context of the statute's purpose, we find an express delegation of rulemaking authority, a "gap" that rules might fill, and "ambiguity" in respect to the boundaries of that gap. Mead Corp., 533 U.S., at 229, 121 S.Ct. 2164 ; see Chevron U.S.A. Inc., 467 U.S., at 843, 104 S.Ct. 2778. We consequently turn to the question whether the Patent Office's regulation is a reasonable exercise of its rulemaking authority.

B

We conclude that the regulation represents a reasonable exercise of the rulemaking authority that Congress delegated to the Patent Office. For one thing, construing a patent claim according to its broadest reasonable construction helps to protect the public. A reasonable, yet unlawfully *2145broad claim might discourage the use of the invention by a member of the public. Because an examiner's (or reexaminer's) use of the broadest reasonable construction standard increases the possibility that the examiner will find the claim too broad (and deny it), use of that standard encourages the applicant to draft narrowly. This helps ensure precision while avoiding overly broad claims, and thereby helps prevent a patent from tying up too much knowledge, while helping members of the public draw useful information from the disclosed invention and better understand the lawful limits of the claim. See § 112(a) ; Nautilus, Inc. v. Biosig Instruments, Inc., 572 U.S. ----, ----, 134 S.Ct. 2120, 2129, 189 L.Ed.2d 37 (2014) ; see also In re Yamamoto, 740 F.2d 1569, 1571 (C.A.Fed.1984).

For another, past practice supports the Patent Office's regulation. See 77 Fed.Reg. 48697 (2012). The Patent Office has used this standard for more than 100 years. 793 F.3d, at 1276. It has applied that standard in proceedings, which, as here, resemble district court litigation. See Bamberger v. Cheruvu, 55 U.S.P.Q.2d 1523, 1527 (BPAI 1998) (broadest reasonable construction standard applies in interference proceedings); Brief for Generic Pharmaceutical Association et al. as Amici Curiae 7-16 (describing similarities between interference proceedings and adjudicatory aspects of inter partes review); see also In re Yamamoto, supra, at 1571 (broadest reasonable construction standard applies in reexamination). It also applies that standard in proceedings that may be consolidated with a concurrent inter partes review. See 77 Fed.Reg. 48697-48698.

Cuozzo makes two arguments in response. First, Cuozzo says that there is a critical difference between the Patent Office's initial examination of an application to determine if a patent should issue, and this proceeding, in which the agency reviews an already-issued patent. In an initial examination of an application for a patent the examiner gives the claim its broadest reasonable construction. But if the patent examiner rejects the claim, then, as described above, Part I-A, supra, the applicant has a right to amend and resubmit the claim. And the examiner and applicant may repeat this process at least once more. This system-broad construction with a chance to amend-both protects the public from overly broad claims and gives the applicant a fair chance to draft a precise claim that will qualify for patent protection. In inter partes review, however, the broadest reasonable construction standard may help protect certain public interests, but there is no absolute right to amend any challenged patent claims. This, Cuozzo says, is unfair to the patent holder.

The process however, is not as unfair as Cuozzo suggests. The patent holder may, at least once in the process, make a motion to do just what he would do in the examination process, namely, amend or narrow the claim. § 316(d) (2012 ed.). This opportunity to amend, together with the fact that the original application process may have presented several additional opportunities to amend the patent, means that use of the broadest reasonable construction standard is, as a general matter, not unfair to the patent holder in any obvious way.

Cuozzo adds that, as of June 30, 2015, only 5 out of 86 motions to amend have been granted. Brief for Petitioner 30; see Tr. of Oral Arg. 30 (noting that a sixth motion had been granted by the time of oral argument in this case). But these numbers may reflect the fact that no amendment could save the inventions at issue, i.e., that the patent should have never issued at all.

*2146To the extent Cuozzo's statistical argument takes aim at the manner in which the Patent Office has exercised its authority, that question is not before us. Indeed, in this particular case, the agency determined that Cuozzo's proposed amendment "enlarge[d]," rather than narrowed, the challenged claims. App. to Pet. for Cert. 165a-166a; see § 316(d)(3). Cuozzo does not contend that the decision not to allow its amendment is "arbitrary" or "capricious," or "otherwise [un]lawful." 5 U.S.C. § 706(2)(a).

Second, Cuozzo says that the use of the broadest reasonable construction standard in inter partes review, together with use of an ordinary meaning standard in district court, may produce inconsistent results and cause added confusion. A district court may find a patent claim to be valid, and the agency may later cancel that claim in its own review. We recognize that that is so. This possibility, however, has long been present in our patent system, which provides different tracks-one in the Patent Office and one in the courts-for the review and adjudication of patent claims. As we have explained above, inter partes review imposes a different burden of proof on the challenger. These different evidentiary burdens mean that the possibility of inconsistent results is inherent to Congress' regulatory design. Cf. One Lot Emerald Cut Stones v. United States, 409 U.S. 232, 235-238, 93 S.Ct. 489, 34 L.Ed.2d 438 (1972) (per curiam ).

Moreover, the Patent Office uses the broadest reasonable construction standard in other proceedings, including interference proceedings (described above), which may implicate patents that are later reviewed in district court. The statute gives the Patent Office the power to consolidate these other proceedings with inter partes review. To try to create uniformity of standards would consequently prove difficult. And we cannot find unreasonable the Patent Office's decision to prefer a degree of inconsistency in the standards used between the courts and the agency, rather than among agency proceedings. See 77 Fed.Reg. 48697-48698.

Finally, Cuozzo and its supporting amici offer various policy arguments in favor of the ordinary meaning standard. The Patent Office is legally free to accept or reject such policy arguments on the basis of its own reasoned analysis. Having concluded that the Patent Office's regulation, selecting the broadest reasonable construction standard, is reasonable in light of the rationales described above, we do not decide whether there is a better alternative as a policy matter. That is a question that Congress left to the particular expertise of the Patent Office.

* * *

For the reasons set forth above, we affirm the judgment of the Court of Appeals for the Federal Circuit.

It is so ordered.

APPENDIX

SPEED LIMIT INDICATOR AND METHOD FOR DISPLAYING SPEED AND THE RELEVANT SPEED LIMIT

*2147* * *

DESCRIPTION OF THE CURRENT EMBODIMENT

"In FIG. 1 , a new and improved speed limit indicator and method for displaying speed and the relevant speed limit 10 ... is illustrated.... More particularly, the speed limit indicator and method for displaying speed and the relevant speed limit 10 has a speedometer 12 mounted on a dashboard 26 . [The] [s]peedometer 12 has a backplate 14 made of plastic, speed denoting markings 16 painted on [that] backplate 14 , a colored display 18 made of a red plastic filter, and a plastic needle 20 rotably mounted in the center of [the] backplate 14 . A [GPS] receiver 22 is positioned adjacent to the speedometer 12 . Other gauges 24 typically present on a dashboard 26 are shown.

.....

"[I]n FIG. 4 , a new and improved speed limit indicator and method for displaying speed and the relevant speed limit 10 ...

*2148is illustrated.... More particularly, the speed limit indicator and method for displaying speed and the relevant speed limit 10 has a backplate 14 , colored display 18 , housing 28 , and axle 30 .

.....

"I claim:

.....

"10 . A speed limit indicator comprising:

"a [GPS] receiver;

"a display controller connected to said [GPS] receiver, wherein said display controller adjusts a colored display in response to signals from said [GPS] receiver to continuously update the delineation of which speed readings are in violation of the speed limit at a vehicle's present location; and

"a speedometer integrally attached to said colored display.

.....

"14 . The speed limit indicator as defined in claim 10 , wherein said colored display is a colored filter.

.....

"17 . The speed limit indicator as defined in claim 14 , wherein said display controller rotates said colored filter independently of said speedometer to continuously update the delineation of which speed readings are in violation of the speed limit at a vehicles present location." Cuozzo Patent.

Justice THOMAS, concurring.

The Court invokes Chevron U.S.A. Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984), and United States v. Mead Corp., 533 U.S. 218, 121 S.Ct. 2164, 150 L.Ed.2d 292 (2001), to resolve one of the questions presented in this case. See ante, at 2133 - 2137, 2142 - 2146. But today's decision does not rest on Chevron 's fiction that ambiguity in a statutory term is best construed as an implicit delegation of power to an administrative agency to determine the bounds of the law. In an appropriate case, this Court should reconsider that fiction of Chevron and its progeny. See Michigan v. EPA, 576 U.S. ----, ----, 135 S.Ct. 2699, 2712, 192 L.Ed.2d 674 (2015) (THOMAS, J., concurring) ("Chevron deference raises serious separation-of-powers questions"); see also Department of Transportation v. Association of American Railroads, 575 U.S. ----, ----, 135 S.Ct. 1225, 1242, 191 L.Ed.2d 153 (2015) (THOMAS, J., concurring in judgment) ("[T]he discretion inherent in executive power does not comprehend the discretion to formulate generally applicable rules of private conduct"); Perez v. Mortgage Bankers Assn., 575 U.S. ----, ---- - ----, 135 S.Ct. 1199, 1217, 191 L.Ed.2d 186 (2015) (THOMAS, J., concurring in judgment) ("Those who ratified the Constitution knew that legal texts would often contain ambiguities.... The judicial power was understood to include the power to resolve these ambiguities over time"); Cass, Is Chevron 's Game Worth the Candle? Burning Interpretation at Both Ends, in Liberty's Nemesis 57-69 (D. Reuter & J. Yoo eds. 2016).

The Court avoids those constitutional concerns today because the provision of the America Invents Act at issue contains an express and clear conferral of authority to the Patent Office to promulgate rules governing its own proceedings. See 35 U.S.C. § 316(a)(4) ; ante, at 2142 -2143. And by asking whether the Patent Office's preferred rule is reasonable, ante, at 2144 - 2146, the Court effectively asks whether the rulemaking was "arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law," in conformity with the Administrative Procedure Act, 5 U.S.C. § 706(2)(A). I therefore join the Court's opinion in full.

Justice ALITO, with whom Justice SOTOMAYOR joins, concurring in part and dissenting in part.

Congress has given the Patent and Trademark Office considerable authority *2149to review and cancel issued patent claims. At the same time, Congress has cabined that power by imposing significant conditions on the Patent Office's institution of patent review proceedings. Unlike the Court, I do not think that Congress intended to shield the Patent Office's compliance-or noncompliance-with these limits from all judicial scrutiny. Rather, consistent with the strong presumption favoring judicial review, Congress required only that judicial review, including of issues bearing on the institution of patent review proceedings, be channeled through an appeal from the agency's final decision. I respectfully dissent from the Court's contrary holding.1

I

In the Leahy-Smith America Invents Act (AIA), 35 U.S.C. § 100 et seq., Congress created three new mechanisms for Patent Office review of issued patent claims-inter partes review, post-grant review, and covered business method patent review (CBM review). This case involves the first of these proceedings, inter partes review.

Under inter partes review, anyone may file a petition challenging the patentability of an issued patent claim at almost any time. §§ 311(a), (c). The grounds for challenge are limited to the patentability of the claim under § 102 (which requires patent claims to be novel) and § 103 (which requires patent claims to be nonobvious). § 311(b).

The statute imposes other restrictions as well. A petition for inter partes review "may be considered only if" the petition satisfies certain requirements, including (as relevant here) that the petition "identif[y], in writing and with particularity, each claim challenged, the grounds on which the challenge to each claim is based, and the evidence that supports the grounds for the challenge to each claim." § 312(a)(3). Additionally, "inter partes review may not be instituted" if the party challenging the patent previously filed a civil action challenging the patent's validity or was sued for infringing the patent more than a year before seeking inter partes review. §§ 315(a)(1), (b). Finally, the Patent Office may not institute inter partes review "unless the Director [of the Patent Office] determines that the information presented in the [challenger's] petition ... and any response [by the patent owner] shows that there is a reasonable likelihood that the petitioner would prevail with respect to at least 1 of the claims challenged in the petition." § 314(a).2

The statute provides that "[t]he determination by the Director whether to institute an inter partes review under this section shall be final and nonappealable." § 314(d). If inter partes review is instituted, the Patent Office conducts a trial that culminates in a "final written decision" on the patentability of the challenged claims. § 318(a). Any patent owner or challenger that is "dissatisfied" with that decision may appeal to the Federal Circuit. § 319.

II

In this case, the Patent Office instituted inter partes review of claims 10 and 14 of Cuozzo's patent based on prior art that the challenger's petition did not cite with respect to those claims. After trial, the Patent Office issued a final written decision holding those claims unpatentable, and Cuozzo appealed that decision to the Federal Circuit. In its appeal, Cuozzo argued (among other things) that the Patent Office had violated the requirement that a *2150petition for inter partes review "may be considered only if" the petition identifies "the grounds on which the challenge to each claim is based, and the evidence that supports the grounds for the challenge," "with particularity." § 312(a)(3).

The Federal Circuit held that it could not entertain this argument because § 314(d) provides that the Patent Office's decision to institute an inter partes review is "final and nonappealable." See In re Cuozzo Speed Technologies, LLC, 793 F.3d 1268, 1273 (2015). This Court now affirms.

I disagree. We have long recognized that "Congress rarely intends to prevent courts from enforcing its directives to federal agencies. For that reason, this Court applies a 'strong presumption' favoring judicial review of administrative action." Mach Mining, LLC v. EEOC, 575 U.S. ----, ----, 135 S.Ct. 1645, 1651, 191 L.Ed.2d 607 (2015) (quoting Bowen v. Michigan Academy of Family Physicians, 476 U.S. 667, 670, 106 S.Ct. 2133, 90 L.Ed.2d 623 (1986) ). While the "presumption is rebuttable," "the agency bears a 'heavy burden' in attempting to show that Congress 'prohibit[ed] all judicial review' of the agency's compliance with a legislative mandate." Mach Mining, supra, at ---- - ----, 135 S.Ct., at 1651 (quoting Dunlop v. Bachowski, 421 U.S. 560, 567, 95 S.Ct. 1851, 44 L.Ed.2d 377 (1975) ). If a provision can reasonably be read to permit judicial review, it should be.

Our decision in Lindahl v. Office of Personnel Management, 470 U.S. 768, 105 S.Ct. 1620, 84 L.Ed.2d 674 (1985), illustrates the power of this presumption. The statute at issue there provided that agency " 'decisions ... concerning [questions of disability and dependency] are final and conclusive and are not subject to review.' " Id., at 771, 105 S.Ct. 1620. The Federal Circuit concluded that the statute cut off all judicial review of such decisions, stating that " '[i]t is difficult to conceive of a more clear-cut statement of congressional intent to preclude review than one in which the concept of finality is thrice repeated in a single sentence.' " Id., at 779, 105 S.Ct. 1620. We reversed. We acknowledged that the statute "plausibly c[ould] be read as imposing an absolute bar to judicial review," but we concluded that "it also quite naturally c[ould] be read as precluding review only of ... factual determinations" underlying the agency's decision, while permitting review of legal questions. Ibid. In light of the presumption of reviewability, we adopted the latter reading. We observed that "when Congress intends to bar judicial review altogether, it typically employs language far more unambiguous and comprehensive," giving as an example a statute that made an agency decision " 'final and conclusive for all purposes and with respect to all questions of law or fact' " and " 'not subject to review by another official of the United States or by a court by mandamus or otherwise.' " Id., at 779-780, and n. 13, 105 S.Ct. 1620.3

This is a far easier case than Lindahl . There is no question that the statute now before us can naturally-perhaps most naturally-be read to permit judicial review *2151of issues bearing on the Patent Office's institution of inter partes review. Section 314(d) reads: "The determination by the Director whether to institute an inter partes review under this section shall be final and nonappealable." Unlike the statutes we addressed in Lindahl (including the one we found to permit review), § 314(d) does not say that an institution decision is "not subject to review." Instead, it makes the institution decision "nonappealable." This is fairly interpreted to bar only an appeal from the institution decision itself, while allowing review of institution-related issues in an appeal from the Patent Office's final written decision at the end of the proceeding. See § 319. Our cases have used the term "nonappealable" in just this way-to refer to matters that are not immediately or independently appealable, but which are subject to review at a later point.4 Thus, while the decision to institute inter partes review is "final and nonappealable" in the sense that a court cannot stop the proceeding from going forward,5 the question whether it was lawful to institute review will not escape judicial scrutiny. This approach is consistent with the normal rule that a party may challenge earlier agency rulings that are themselves "not directly reviewable" when seeking review of a final, appealable decision. 5 U.S.C. § 704. And it strikes a sensible balance: The Patent Office may proceed unimpeded with the inter partes review process (which must normally be completed within one year, see 35 U.S.C. § 316(a)(11) ), but it will be held to account for its compliance with the law at the end of the day.

In rejecting this commonsense interpretation, the Court gives short shrift to the presumption in favor of judicial review. Its primary reason for disregarding the presumption reduces to an assertion-devoid of any textual analysis-that surely § 314(d) must bar review of legal questions related to institution decisions. Ante, at 2139 - 2140. As I have explained, the statute's text does not require that conclusion.

Moving (further) away from the statutory text, the Court next objects that allowing judicial review "would undercut one important congressional objective, namely, giving the Patent Office significant power to revisit and revise earlier patent grants." Ante, at 2139. I am not sure that the Court appreciates how remarkable this assertion is. It would give us cause to do away with judicial review whenever we think that review makes it harder for an agency to carry out important work. In any event, the majority's logic is flawed. Judicial review enforces the limits that Congress has imposed on the agency's power. It thus serves to buttress, not "undercut," Congress's objectives. By asserting otherwise, the majority loses sight of the principle that "no legislation pursues its purposes at all costs." Rodriguez v. United States, 480 U.S. 522, 525-526, 107 S.Ct. 1391, 94 L.Ed.2d 533 (1987) (per curiam ). "Every statute purposes, not only to achieve certain ends, but also to achieve them by particular means-and *2152there is often a considerable legislative battle over what those means ought to be. The withholding of agency authority is as significant as the granting of it, and we have no right to play favorites between the two." Director, Office of Workers' Compensation Programs v. Newport News Shipbuilding & Dry Dock Co., 514 U.S. 122, 136, 115 S.Ct. 1278, 131 L.Ed.2d 160 (1995). The inter partes review statute is no exception. It empowers the Patent Office to clean up bad patents, but it expressly forbids the Patent Office to institute inter partes review-or even consider petitions for inter partes review-unless certain conditions are satisfied. Nothing in the statute suggests that Congress wanted to improve patent quality at the cost of fidelity to the law.

The Court also observes that the inter partes review appeal provision, § 319, "limit[s] appellate review to the 'final written decision.' " Ante, at 2140. The majority reads too much into this provision. Section 319 provides simply that "[a] party dissatisfied with the final written decision ... may appeal the decision." The statute does not restrict the issues that may be raised in such an appeal. As the Patent Office once explained (before having a change of heart), the "plain language of the statutory text" recognizes a "right of judicial review ... for any party 'dissatisfied' by the [Patent Office's] ultimate 'written [decision],' " and "[n]othing in the statutory scheme limits the reasons that a party might be so 'dissatisfied.' " Memorandum of Law in Support of Defendant's Motion to Dismiss in Versata Development Group, Inc. v. Rea, Civ. Action No. 1:13cv328 (ED Va., May 16, 2013), p. 16. A party may be dissatisfied with a final written decision in an inter partes review because the Patent Office lacked authority to institute the proceeding in the first place, or because the Office committed some other error in the lead-up to its final decision. Neither § 314(d) nor § 319 prevents a party from pressing such issues on an appeal from the final decision. This is familiar practice under 28 U.S.C. § 1291, which similarly limits appeals to "final decisions of the district courts" but allows appellants to challenge earlier rulings as part of those appeals. See Quackenbush v. Allstate Ins. Co., 517 U.S. 706, 712, 116 S.Ct. 1712, 135 L.Ed.2d 1 (1996) ("The general rule is that a party is entitled to a single appeal, to be deferred until final judgment has been entered, in which claims of district court error at any stage in the litigation may be ventilated" (internal quotation marks omitted)); 15A C. Wright, A. Miller, & E. Cooper, Federal Practice and Procedure § 3905.1, pp. 250, 252 (2d ed. 1992) (noting "the general rule that an appeal from final judgment ... permits review of all rulings that led up to the judgment" and observing that "[t]he variety of orders open to review on subsequent appeal from a final judgment is enormous"). And, as noted above, judicial review of "final agency action" likewise encompasses earlier rulings that are "not directly reviewable." 5 U.S.C. § 704 ; see supra, at 2151.

The Court next contends that my interpretation renders 35 U.S.C. § 314(d)"superfluous." Ante, at 2140. Reading the statute to defer review of institution decisions is "unnecessary," the Court says, because the "Administrative Procedure Act already limits review to final agency decisions" and a "decision to initiate inter partes review is 'preliminary,' not 'final.' " Ibid. But Congress reasonably may have thought that the matter needed clarifying, given that § 314(d) itself calls such a decision "final" (albeit in a different sense, see supra, at 2150 - 2151). Language is not superfluous when it "remove[s] any doubt" about a point that might otherwise be unclear.

*2153Ali v. Federal Bureau of Prisons, 552 U.S. 214, 226, 128 S.Ct. 831, 169 L.Ed.2d 680 (2008). More important, my reading prevents an appeal from a decision not to institute inter partes review, which is plainly final agency action and so-absent § 314(d) -might otherwise trigger immediate review. The Court asserts that this too is unnecessary because, in its view, a decision to deny inter partes review is "committed to agency discretion by law" and so unreviewable under normal principles of administrative law. 5 U.S.C. § 701(a)(2) ; see ante, at 2153. I agree that one can infer from the statutory scheme that the Patent Office has discretion to deny inter partes review even if a challenger satisfies the threshold requirements for review. But the law does not say so directly and Congress may not have thought the point self-evident. Again, 35 U.S.C. § 314(d) plays a clarifying role. This gives the provision plenty of work to do. There is no need to read it more broadly.6

III

A

None of this is to say that courts must-or should-throw out an inter partes review decision whenever there is some technical deficiency in the challenger's petition or in the Patent Office's institution decision. Although § 314(d) does not preclude review of issues bearing on institution, normal limits on judicial review still apply. For example, errors that do not cause a patent owner prejudice may not warrant relief. See 5 U.S.C. § 706 ("[D]ue account shall be taken of the rule of prejudicial error"). Some errors may also be superseded by later developments. Most notably, once the Patent Office issues its final written decision, the probabilistic question whether a challenger is "reasonabl[y] likel[y]" to prevail on the merits, 35 U.S.C. § 314(a), will be subsumed by the ultimate question whether the challenger should in fact prevail.7 And while I have no occasion here to decide the matter, it may be that courts owe some degree of deference to the Patent Office's application of the statutory prerequisites to inter partes review.

I would leave these considerations for the Court of Appeals to address in the first instance. But I must confess doubts that Cuozzo could ultimately prevail. As noted above, Cuozzo argues that the Patent Office improperly granted inter partes review of claims 10 and 14 on grounds not asserted in the petition for inter partes review, in violation of the statutory requirement *2154that a petition must state the grounds for challenge "with particularity." § 312(a)(3). The problem for Cuozzo is that claim 17-which the petition properly challenged-incorporates all of the elements of claims 10 and 14. Accordingly, an assertion that claim 17 is unpatentable in light of certain prior art is necessarily an assertion that claims 10 and 14 are unpatentable as well. Assuming that Cuozzo must show prejudice from the error it alleges, it is hard to see how Cuozzo could do so here.

B

But any perceived weakness in the merits of Cuozzo's appeal does not mean that such issues are unworthy of judicial review. Section 312(a)(3)'s particularity requirement is designed, at least in part, to ensure that a patent owner has sufficient notice of the challenge against which it must defend. Once inter partes review is instituted, the patent owner's response-its opening brief, essentially-is filed as an opposition to the challenger's petition. See § 316(a)(8) ; 37 CFR § 42.120. Thus, if a petition fails to state its challenge with particularity-or if the Patent Office institutes review on claims or grounds not raised in the petition-the patent owner is forced to shoot into the dark. The potential for unfairness is obvious.

Other problems arise if the Patent Office fails to enforce the prohibitions against instituting inter partes review at the behest of challengers that have already sued to invalidate the patent or that were sued for infringement more than a year before seeking inter partes review. 35 U.S.C. §§ 315(a)(1), (b). Allowing such a challenge exposes the patent owner to the burden of multiplicative proceedings-including discovery in both forums, see § 316(a)(5) -while permitting the challenger to exploit inter partes review's lower standard of proof and more favorable claim construction standard. Congress understandably thought that the Patent Office's power should not be wielded in this way. Yet, according to the Court, Congress made courts powerless to correct such abuses.

Even more striking are the consequences that today's decision portends for the AIA's other patent review mechanisms, post-grant review and CBM review, see supra, at 2148 - 2149 which are subject to a "no appeal" provision virtually identical to § 314(d). See § 324(e) ("The determination by the Director whether to institute a post-grant review under this section shall be final and nonappealable"); see AIA § 18(a)(1), 125 Stat. 329, note following 35 U.S.C. § 321, p. 1442 (CBM review generally "shall be regarded as, and shall employ the standards and procedures of, a post-grant review"). Post-grant review and CBM review allow for much broader review than inter partes review. While inter partes review is limited to assessing patentability under § 102 and § 103, in post-grant review and CBM review, patent claims can also be scrutinized (and canceled) on any invalidity ground that may be raised as a defense to infringement, including such grounds as ineligible subject matter under § 101, indefiniteness under § 112, and improper enlargement of reissued claims under § 251. See § 321(b) ; §§ 282(b)(2), (3). But this broader review comes with its own strict limits. A petition for post-grant review must be filed within nine months after a patent is granted. § 321(c). And while CBM review is not subject to this time limit, Congress imposed a subject-matter restriction: The Patent Office "may institute a [CBM review] proceeding only for a patent that is a covered business method patent," which Congress defined to cover certain patents with claims relating to "a financial product or service." AIA

*2155§§ 18(a)(1)(E), (d)(1), at 1442; see § 18(a)(1)(A), ibid.8

Congress thus crafted a three-tiered framework for Patent Office review of issued patents: broad post-grant review in a patent's infancy, followed by narrower inter partes review thereafter, with a limited exception for broad review of older covered business method patents. Today's decision threatens to undermine that carefully designed scheme. Suppose that the Patent Office instituted post-grant review on a petition filed 12 months (or even 12 years) after a patent was issued, and then invalidated a patent claim as indefinite under § 112 -a ground available in post-grant review but not in inter partes review. This would grossly exceed the Patent Office's authority and would be manifestly prejudicial to the patent owner. Can Congress really have intended to shield such shenanigans from judicial scrutiny? The Court answers with a non sequitur: Of course the Patent Office cannot cancel a patent under § 112"in inter partes review." Ante, at 2154. The Court seems to think that we could overturn the Patent Office's decision to institute "post-grant review" based on an untimely petition and declare that the agency has really instituted only "inter partes review." But how is that possible under today's opinion? After all, the petition's timeliness, no less than the particularity of its allegations, is "closely tied to the application and interpretation of statutes related to the Patent Office's decision to initiate ... review," and the Court says that such questions are unreviewable. Ibid. ; see § 321(c) ; § 312(a)(3).

To take things a step further, suppose that the Patent Office purported to forgive the post-grant review petition's tardiness by declaring the challenged patent a "covered business method patent," even though the patent has nothing to do with financial products or services (it claims, say, a new kind of tempered glass). Again, this involves the application of statutes related to the Patent Office's institution decision. See AIA § 18(a)(1)(E), at 1442 (Patent Office "may institute a [CBM review] proceeding only for a patent that is a covered business method patent"). So is this specious determination immune from judicial scrutiny under the Court's reasoning?

If judicial review of these issues is unavailable, then nothing would prevent the Patent Office from effectively collapsing Congress's three-tiered review structure and subjecting all patents to broad post-grant review at all times. Congress cannot have intended that.

I take the Court at its word that today's opinion will not permit the Patent Office "to act outside its statutory limits" in these ways. Ante, at 2141. But how to get there from the Court's reasoning-and how to determine which "statutory limits" we should enforce and which we should not-remains a mystery. I would avoid the suspense and hold that 35 U.S.C. § 314(d) does not bar judicial review of the Patent Office's compliance with any of the limits Congress imposed on the institution of patent review proceedings. That includes the statutory limit, § 312(a)(3), that Cuozzo alleges was violated here.

* * *

In enacting the AIA, Congress entrusted the Patent Office with a leading role in combating the detrimental effect that bad patents can have on innovation. But Congress did not give the agency unbridled authority. The principles I have set forth afford the Patent Office plenty of latitude to carry out its charge, while ensuring that *2156the Office's actions-no less than the patents it reviews-stay within the bounds of the law.

I would vacate the Federal Circuit's judgment and remand for that court to consider whether the Patent Office exceeded its authority to institute inter partes review with respect to claims 10 and 14 of Cuozzo's patent. With respect to claim 17, I agree with the Court that the judgment below must be affirmed. See n. 1, supra ; Part III, ante .

Chief Justice ROBERTS delivered the opinion of the Court.

The Endangered Species Act directs the Secretary of the Interior, upon listing a species as endangered, to also designate the "critical habitat" of the species. A group of landowners whose property was designated as critical habitat for an endangered frog challenged the designation. The landowners urge that their land cannot be critical habitat because it is not habitat, which they contend refers only to areas where the frog could currently survive. The court below ruled that the Act imposed no such limitation on the scope of critical habitat.

The Act also authorizes the Secretary to exclude an area that would otherwise be included as critical habitat, if the benefits of exclusion outweigh the benefits of designation. The landowners challenged the decision of the Secretary not to exclude their property, but the court below held that the Secretary's action was not subject to judicial review.

We granted certiorari to review both rulings.

I

A

The amphibian Rana sevosa is popularly known as the "dusky gopher frog"-*365"dusky" because of its dark coloring and "gopher" because it lives underground. The dusky gopher frog is about three inches long, with a large head, plump body, and short legs. Warts dot its back, and dark spots cover its entire body. Final Rule To List the Mississippi Gopher Frog Distinct Population Segment of Dusky Gopher Frog as Endangered, 66 Fed.Reg. 62993 (2001) (Final Listing). It is noted for covering its eyes with its front legs when it feels threatened, peeking out periodically until danger passes. Markle Interests, LLC v. United States Fish and Wildlife Serv., 827 F.3d 452, 458, n. 2 (C.A.5 2016). Less endearingly, it also secretes a bitter, milky substance to deter would-be diners. Brief for Intervenor-Respondents 6, n. 1.

The frog spends most of its time in burrows and stump holes located in upland longleaf pine forests. In such forests, frequent fires help maintain an open canopy, which in turn allows vegetation to grow on the forest floor. The vegetation supports the small insects that the frog eats and provides a place for the frog's eggs to attach when it breeds. The frog breeds in "ephemeral" ponds that are dry for part of the year. Such ponds are safe for tadpoles because predatory fish cannot live in them. Designation of Critical Habitat for Dusky Gopher Frog, 77 Fed.Reg. 35129-35131 (2012) (Designation).

The dusky gopher frog once lived throughout coastal Alabama, Louisiana, and Mississippi, in the longleaf pine forests that used to cover the southeast. But more than 98% of those forests have been removed to make way for urban development, agriculture, and timber plantations. The timber plantations consist of fast-growing loblolly pines planted as close together as possible, resulting in a closed-canopy forest inhospitable to the frog. The near eradication of the frog's habitat sent the species into severe decline. By 2001, the known wild population of the dusky gopher frog had dwindled to a group of 100 at a single pond in southern Mississippi. That year, the Fish and Wildlife Service, which administers the Endangered Species Act of 1973 on behalf of the Secretary of the Interior, listed the dusky gopher frog as an endangered species. Final Listing 62993-62995; see 87 Stat. 886, 16 U.S.C. § 1533(a)(1).

B

When the Secretary lists a species as endangered, he must also designate the critical habitat of that species. § 1533(a)(3)(A)(i). The ESA defines "critical habitat" as:

"(i) the specific areas within the geographical area occupied by the species ... on which are found those physical or biological features (I) essential to the conservation of the species and (II) which may require special management considerations or protection; and

"(ii) specific areas outside the geographical area occupied by the species ... upon a determination by the Secretary that such areas are essential for the conservation of the species." § 1532(5)(A).

Before the Secretary may designate an area as critical habitat, the ESA requires him to "tak[e] into consideration the economic impact" and other relevant impacts of the designation. § 1533(b)(2). The statute goes on to authorize him to "exclude any area from critical habitat if he determines that the benefits of such exclusion outweigh the benefits of [designation]," unless exclusion would result in extinction of the species. Ibid.

A critical-habitat designation does not directly limit the rights of private landowners. It instead places conditions on *366the Federal Government's authority to effect any physical changes to the designated area, whether through activities of its own or by facilitating private development. Section 7 of the ESA requires all federal agencies to consult with the Secretary to "[e]nsure that any action authorized, funded, or carried out by such agency" is not likely to adversely affect a listed species' critical habitat. 16 U.S.C. § 1536(a)(2). If the Secretary determines that an agency action, such as issuing a permit, would harm critical habitat, then the agency must terminate the action, implement an alternative proposed by the Secretary, or seek an exemption from the Cabinet-level Endangered Species Committee. See National Assn. of Home Builders v. Defenders of Wildlife, 551 U.S. 644, 652, 127 S.Ct. 2518, 168 L.Ed.2d 467 (2007) ; 50 C.F.R. 402.15 (2017).

Due to resource constraints, the Service did not designate the frog's critical habitat in 2001, when it listed the frog as endangered. Designation, at 35118-35119. In the following years, the Service discovered two additional naturally occurring populations and established another population through translocation. The first population nonetheless remains the only stable one and by far the largest. Dept. of Interior, U.S. Fish and Wildlife Serv., Dusky Gopher Frog (Rana sevosa ) Recovery Plan iv, 6-7 (2015).

In 2010, in response to litigation by the Center for Biological Diversity, the Service published a proposed critical-habitat designation. Designation, at 35119. The Service proposed to designate as occupied critical habitat all four areas with existing dusky gopher frog populations. The Service found that each of those areas possessed the three features that the Service considered "essential to the conservation" of the frog and that required special protection: ephemeral ponds; upland open-canopy forest containing the holes and burrows in which the frog could live; and open-canopy forest connecting the two. But the Service also determined that designating only those four sites would not adequately ensure the frog's conservation. Because the existing dusky gopher frog populations were all located in two adjacent counties on the Gulf Coast of Mississippi, local events such as extreme weather or an outbreak of an infectious disease could jeopardize the entire species. Designation of Critical Habitat for Mississippi Gopher Frog, 75 Fed.Reg. 31394 (2010) (proposed 50 C.F.R. Part 17).

To protect against that risk, the Service proposed to designate as unoccupied critical habitat a 1,544-acre site in St. Tammany Parish, Louisiana. The site, dubbed "Unit 1" by the Service, had been home to the last known population of dusky gopher frogs outside of Mississippi. The frog had not been seen in Unit 1 since 1965, and a closed-canopy timber plantation occupied much of the site. But the Service found that the site retained five ephemeral ponds "of remarkable quality," and determined that an open-canopy forest could be restored on the surrounding uplands "with reasonable effort." Although the uplands in Unit 1 lacked the open-canopy forests (and, of course, the frogs) necessary for designation as occupied critical habitat, the Service concluded that the site met the statutory definition of unoccupied critical habitat because its rare, high-quality breeding ponds and its distance from existing frog populations made it essential for the conservation of the species. Designation, at 35118, 35124, 35133, 35135.

After issuing its proposal, the Service commissioned a report on the probable economic impact of designating each area, including Unit 1, as critical habitat for the dusky gopher frog. See 16 U.S.C. § 1533(b)(2) ; App. 63. Petitioner Weyerhaeuser *367Company, a timber company, owns part of Unit 1 and leases the remainder from a group of family landowners. Brief for Petitioner 16. While the critical-habitat designation has no direct effect on the timber operations, St. Tammany Parish is a fast-growing part of the New Orleans metropolitan area, and the landowners have already invested in plans to more profitably develop the site. App. 80-83. The report recognized that anyone developing the area may need to obtain Clean Water Act permits from the Army Corps of Engineers before filling any wetlands on Unit 1. 33 U.S.C. § 1344(a). Because Unit 1 is designated as critical habitat, Section 7 of the ESA would require the Corps to consult with the Service before issuing any permits.

According to the report, that consultation process could result in one of three outcomes. First, it could turn out that the wetlands in Unit 1 are not subject to the Clean Water Act permitting requirements, in which case the landowners could proceed with their plans unimpeded. Second, the Service could ask the Corps not to issue permits to the landowners to fill some of the wetlands on the site, in effect prohibiting development on 60% of Unit 1. The report estimated that this would deprive the owners of $20.4 million in development value. Third, by asking the Corps to deny even more of the permit requests, the Service could bar all development of Unit 1, costing the owners $33.9 million. The Service concluded that those potential costs were not "disproportionate" to the conservation benefits of designation. "Consequently," the Service announced, it would not "exercis[e][its] discretion to exclude" Unit 1 from the dusky gopher frog's critical habitat. App. 188-190.

C

Weyerhaeuser and the family landowners sought to vacate the designation in Federal District Court. They contended that Unit 1 could not be critical habitat for the dusky gopher frog because the frog could not survive there: Survival would require replacing the closed-canopy timber plantation encircling the ponds with an open-canopy longleaf pine forest. The District Court nonetheless upheld the designation. Markle Interests, LLC v. United States Fish and Wildlife Serv ., 40 F.Supp.3d 744 (E.D.La.2014). The court determined that Unit 1 satisfied the statutory definition of unoccupied critical habitat, which requires only that the Service deem the land "essential for the conservation [of] the species." Id., at 760.

Weyerhaeuser also challenged the Service's decision not to exclude Unit 1 from the dusky gopher frog's critical habitat, arguing that the Service had failed to adequately weigh the benefits of designating Unit 1 against the economic impact. In addition, Weyerhaeuser argued that the Service had used an unreasonable methodology for estimating economic impact and, regardless of methodology, had failed to consider several categories of costs. Id., at 759. The court approved the Service's methodology and declined to consider Weyerhaeuser's challenge to the decision not to exclude. See id ., at 763-767, and n. 29.

The Fifth Circuit affirmed. 827 F.3d 452. The Court of Appeals rejected the suggestion that the definition of critical habitat contains any "habitability requirement." Id., at 468. The court also concluded that the Service's decision not to exclude Unit 1 was committed to agency discretion by law and was therefore unreviewable. Id., at 473-475. Judge Owen dissented. She wrote that Unit 1 could not be "essential for the conservation of the species" because it lacked the open-canopy forest that the Service itself had *368determined was "essential to the conservation" of the frog. Id., at 480-481.

The Fifth Circuit denied rehearing en banc. Markle Interests, LLC v. United States Fish and Wildlife Serv ., 848 F.3d 635 (2017). Judge Jones dissented, joined by Judges Jolly, Smith, Clement, Owen, and Elrod. They reasoned that critical habitat must first be habitat, and Unit 1 in its present state could not be habitat for the dusky gopher frog. Id., at 641. The dissenting judges also concluded that the Service's decision not to exclude Unit 1 was reviewable for abuse of discretion. Id., at 654, and n. 21.

We granted certiorari to consider two questions: (1) whether "critical habitat" under the ESA must also be habitat; and (2) whether a federal court may review an agency decision not to exclude a certain area from critical habitat because of the economic impact of such a designation. 583 U.S. ----, --- S.Ct. ----, --- L.Ed.2d ---- (2018).1

II

A

Our analysis starts with the phrase "critical habitat." According to the ordinary understanding of how adjectives work, "critical habitat" must also be "habitat." Adjectives modify nouns-they pick out a subset of a category that possesses a certain quality. It follows that "critical habitat" is the subset of "habitat" that is "critical" to the conservation of an endangered species.

Of course, "[s]tatutory language cannot be construed in a vacuum," Sturgeon v. Frost, 577 U.S. ----, ----, 136 S.Ct. 1061, 1070, 194 L.Ed.2d 108 (2016) (internal quotation marks omitted), and so we must also consider "critical habitat" in its statutory context. Section 4(a)(3)(A)(i), which the lower courts did not analyze, is the sole source of authority for critical-habitat designations. That provision states that when the Secretary lists a species as endangered he must also "designate any habitat of such species which is then considered to be critical habitat." 16 U.S.C. § 1533(a)(3)(A)(i) (emphasis added). Only the "habitat" of the endangered species is eligible for designation as critical habitat. Even if an area otherwise meets the statutory definition of unoccupied critical habitat because the Secretary finds the area essential for the conservation of the species, Section 4(a)(3)(A)(i) does not authorize the Secretary to designate the area as critical habitat unless it is also habitat for the species.

The Center for Biological Diversity contends that the statutory definition of critical habitat is complete in itself and does not require any independent inquiry into the meaning of the term "habitat," which the statute leaves undefined. Brief for Intervenor-Respondents 43-49. But the statutory definition of "critical habitat" tells us what makes habitat "critical," not what makes it "habitat." Under the statutory definition, critical habitat comprises areas occupied by the species "on which *369are found those physical or biological features (I) essential to the conservation of the species and (II) which may require special management considerations or protection," as well as unoccupied areas that the Secretary determines to be "essential for the conservation of the species." 16 U.S.C. § 1532(5)(A). That is no baseline definition of habitat-it identifies only certain areas that are indispensable to the conservation of the endangered species. The definition allows the Secretary to identify the subset of habitat that is critical, but leaves the larger category of habitat undefined.

The Service does not now dispute that critical habitat must be habitat, see Brief for Federal Respondents 23, although it made no such concession below. Instead, the Service argues that habitat includes areas that, like Unit 1, would require some degree of modification to support a sustainable population of a given species. Id ., at 27. Weyerhaeuser, for its part, urges that habitat cannot include areas where the species could not currently survive. Brief for Petitioner 25. (Habitat can, of course, include areas where the species does not currently live, given that the statute defines critical habitat to include unoccupied areas.) The Service in turn disputes Weyerhaeuser's premise that the administrative record shows that the frog could not survive in Unit 1. Brief for Federal Respondents 22, n. 4.

The Court of Appeals concluded that "critical habitat" designations under the statute were not limited to areas that qualified as habitat. See 827 F.3d, at 468 ("There is no habitability requirement in the text of the ESA or the implementing regulations."). The court therefore had no occasion to interpret the term "habitat" in Section 4(a)(3)(A)(i) or to assess the Service's administrative findings regarding Unit 1. Accordingly, we vacate the judgment below and remand to the Court of Appeals to consider these questions in the first instance.2

B

Weyerhaeuser also contends that, even if Unit 1 could be properly classified as critical habitat for the dusky gopher frog, the Service should have excluded it from designation under Section 4(b)(2) of the ESA. That provision requires the Secretary to "tak[e] into consideration the economic impact ... of specifying any particular area as critical habitat" and authorizes him to "exclude any area from critical habitat if he determines that the benefits of such exclusion outweigh the benefits of specifying such area as part of the critical habitat." 16 U.S.C. § 1533(b)(2). To satisfy its obligation to consider economic impact, the Service commissioned a report estimating the costs of its proposed critical-habitat designation. The Service concluded that the costs of designating the proposed areas, including Unit 1, were not "disproportionate" to the conservation benefits and, "[c]onsequently," declined to make any exclusions.

Weyerhaeuser claims that the Service's conclusion rested on a faulty assessment of the costs and benefits of designation and that the resulting decision not to exclude should be set aside. Specifically, Weyerhaeuser contends that the Service improperly *370weighed the costs of designating Unit 1 against the benefits of designating all proposed critical habitat, rather than the benefits of designating Unit 1 in particular. Weyerhaeuser also argues that the Service did not fully account for the economic impact of designating Unit 1 because it ignored, among other things, the costs of replacing timber trees with longleaf pines, maintaining an open canopy through controlled burning, and the tax revenue that St. Tammany Parish would lose if Unit 1 were never developed. Brief for Petitioner 53-54. The Court of Appeals did not consider Weyerhaeuser's claim because it concluded that a decision not to exclude a certain area from critical habitat is unreviewable.

The Administrative Procedure Act creates a "basic presumption of judicial review [for] one 'suffering legal wrong because of agency action.' " Abbott Laboratories v. Gardner, 387 U.S. 136, 140, 87 S.Ct. 1507, 18 L.Ed.2d 681 (1967) (quoting 5 U.S.C. § 702 ). As we explained recently, "legal lapses and violations occur, and especially so when they have no consequence. That is why this Court has so long applied a strong presumption favoring judicial review of administrative action." Mach Mining, LLC v. EEOC, 575 U.S. ----, ---- - ----, 135 S.Ct. 1645, 1652-1653, 191 L.Ed.2d 607 (2015). The presumption may be rebutted only if the relevant statute precludes review, 5 U.S.C. § 701(a)(1), or if the action is "committed to agency discretion by law," § 701(a)(2). The Service contends, and the lower courts agreed, that Section 4(b)(2) of the ESA commits to the Secretary's discretion decisions not to exclude an area from critical habitat.

This Court has noted the "tension" between the prohibition of judicial review for actions "committed to agency discretion" and the command in § 706(2)(A) that courts set aside any agency action that is "arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law." Heckler v. Chaney, 470 U.S. 821, 829, 105 S.Ct. 1649, 84 L.Ed.2d 714 (1985). A court could never determine that an agency abused its discretion if all matters committed to agency discretion were unreviewable. To give effect to § 706(2)(A) and to honor the presumption of review, we have read the exception in § 701(a)(2) 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." Lincoln v. Vigil, 508 U.S. 182, 191, 113 S.Ct. 2024, 124 L.Ed.2d 101 (1993). The Service contends that Section 4(b)(2) of the ESA is one of those rare statutory provisions.

There is, at the outset, reason to be skeptical of the Service's position. The few cases in which we have applied the § 701(a)(2) exception involved agency decisions that courts have traditionally regarded as unreviewable, such as the allocation of funds from a lump-sum appropriation, Lincoln, 508 U.S., at 191, 113 S.Ct. 2024 or a decision not to reconsider a final action, ICC v. Locomotive Engineers, 482 U.S. 270, 282, 107 S.Ct. 2360, 96 L.Ed.2d 222 (1987). By contrast, this case involves the sort of routine dispute that federal courts regularly review: An agency issues an order affecting the rights of a private party, and the private party objects that the agency did not properly justify its determination under a standard set forth in the statute.

Section 4(b)(2) states that the Secretary

"shall designate critical habitat ... after taking into consideration the economic impact, the impact on national security, and any other relevant impact, of specifying any particular area as critical habitat.

*371The Secretary may exclude any area from critical habitat if he determines that the benefits of such exclusion outweigh the benefits of specifying such area ... unless he determines ... that the failure to designate such area as critical habitat will result in the extinction of the species concerned." 16 U.S.C. § 1533(b)(2).

Although the text meanders a bit, we recognized in Bennett v. Spear, 520 U.S. 154, 117 S.Ct. 1154, 137 L.Ed.2d 281 (1997), that the provision describes a unified process for weighing the impact of designating an area as critical habitat. The first sentence of Section 4(b)(2) imposes a "categorical requirement" that the Secretary "tak[e] into consideration" economic and other impacts before such a designation. Id., at 172, 117 S.Ct. 1154 (emphasis deleted). The second sentence authorizes the Secretary to act on his consideration by providing that he may exclude an area from critical habitat if he determines that the benefits of exclusion outweigh the benefits of designation. The Service followed that procedure here (albeit in a flawed manner, according to Weyerhaeuser). It commissioned a report to estimate the costs of designating the proposed critical habitat, concluded that those costs were not "disproportionate" to the benefits of designation, and "[c]onsequently" declined to "exercis[e] [its] discretion to exclude any areas from [the] designation of critical habitat." App. 190.

Bennett explained that the Secretary's "ultimate decision" to designate or exclude, which he "arriv[es] at" after considering economic and other impacts, is reviewable "for abuse of discretion." 520 U.S., at 172, 117 S.Ct. 1154. The Service dismisses that language as a "passing reference ... not necessarily inconsistent with the Service's understanding," which is that the Secretary's decision not to exclude an area is wholly discretionary and therefore unreviewable. Brief for Federal Respondents 50. The Service bases its understanding on the second sentence of Section 4(b)(2), which states that the Secretary "may exclude [an] area from critical habitat if he determines that the benefits of such exclusion outweigh the benefits of [designation]."

The use of the word "may" certainly confers discretion on the Secretary. That does not, however, segregate his discretionary decision not to exclude from the procedure mandated by Section 4(b)(2), which directs the Secretary to consider the economic and other impacts of designation when making his exclusion decisions. Weyerhaeuser's claim is the familiar one in administrative law that the agency did not appropriately consider all of the relevant factors that the statute sets forth to guide the agency in the exercise of its discretion. Specifically, Weyerhaeuser contends that the Service ignored some costs and conflated the benefits of designating Unit 1 with the benefits of designating all of the proposed critical habitat. This is the sort of claim that federal courts routinely assess when determining whether to set aside an agency decision as an abuse of discretion under § 706(2)(A). See Judulang v. Holder, 565 U.S. 42, 53, 132 S.Ct. 476, 181 L.Ed.2d 449 (2011) ("When reviewing an agency action, we must assess ... whether the decision was based on a consideration of the relevant factors and whether there has been a clear error of judgment." (internal quotation marks omitted)).

Section 4(b)(2) requires the Secretary to consider economic impact and relative benefits before deciding whether to exclude an area from critical habitat or to proceed with designation. The statute is, therefore, not "drawn so that a court would have no meaningful standard against which *372to judge the [Secretary's] exercise of [his] discretion" not to exclude. Lincoln, 508 U.S., at 191, 113 S.Ct. 2024.

Because it determined that the Service's decisions not to exclude were committed to agency discretion and therefore unreviewable, the Court of Appeals did not consider whether the Service's assessment of the costs and benefits of designation was flawed in a way that rendered the resulting decision not to exclude Unit 1 arbitrary, capricious, or an abuse of discretion. Accordingly, we remand to the Court of Appeals to consider that question, if necessary, in the first instance.

* * *

The judgment of the Court of Appeals for the Fifth Circuit is vacated, and the case is remanded for further proceedings consistent with this opinion.

It is so ordered.

Justice KAVANAUGH took no part in the consideration or decision of this case.