# Kisor v. Wilkie

No. 18-15 · Decided June 26, 2019 · 588 U. S. 558

> Prepared by Supreme Court Report from the PDF published by the Supreme Court of the United States. The PDF is the authoritative text; this copy can contain errors of conversion. Source: <https://www.supremecourt.gov/opinions/preliminaryprint/588US2PP_final.pdf#page=220>

## Syllabus

Petitioner James Kisor, a Vietnam War veteran, first sought disability benefits from the Department of Veterans Affairs (VA) in 1982, alleging that he had developed post-traumatic stress disorder from his military service. The agency denied his initial request, but in 2006, Kisor moved to reopen his claim. The VA this time agreed he was eligible for benefits, but it granted those benefits only from the date of his motion to reopen, not (as Kisor had requested) from the date of his first application. The Board of Veterans’ Appeals—a part of the VA—affirmed that retroactivity decision, based on its interpretation of an agency rule governing such claims. The Court of Appeals for Veterans Claims affirmed.

The Federal Circuit also affirmed, but it did so by applying a doctrine called *Auer* (or sometimes, *Seminole Rock*) deference. See *Auer* v. *Robbins*, 519 U. S. 452; *Bowles* v. *Seminole Rock & Sand Co.*, 325 U. S. 410. Under that doctrine, this Court has long deferred to an agency’s reasonable reading of its own genuinely ambiguous regulations. The Court of Appeals concluded that the VA regulation at issue was ambiguous, and it therefore deferred to the Board’s interpretation of the rule. Kisor now asks the Court to overrule *Auer*, as well as its predecessor *Seminole Rock*, discarding the deference those decisions give to agencies.

*Held*: The judgment is vacated and remanded.

869 F. 3d 1360, vacated and remanded.

JUSTICE KAGAN delivered the opinion of the Court with respect to Parts I, II–B, III–B, and IV, holding that *Auer* and *Seminole Rock* are not overruled. Pp. 573–580, 586–590.

(a) This Court’s deference doctrine is rooted in a presumption that Congress intended for courts to defer to agencies when they interpret their own ambiguous rules. The Court adopts that presumption for a set of reasons related to the comparative attributes of courts and agencies in answering interpretive questions. But when the reasons for the presumption do not hold up, or when countervailing reasons outweigh them, courts should not give deference to an agency’s reading. The Court has thus cabined *Auer*’s scope in varied and critical ways.

[\*559] First and foremost, a court should not afford *Auer* deference unless, after exhausting all the “traditional tools” of construction, *Chevron U. S. A. Inc.* v. *Natural Resources Defense Council, Inc.*, 467 U. S. 837, 843, n. 9, the regulation is genuinely ambiguous. A court must carefully consider the text, structure, history, and purpose of a regulation before resorting to deference. If genuine ambiguity remains, the agency’s reading must still fall “within the bounds of reasonable interpretation.” *Arlington* v. *FCC*, 569 U. S. 290, 296.

And even then, not every reasonable agency reading of a genuinely ambiguous rule should receive *Auer* deference. Rather, a court must also make an independent inquiry into whether the character and context of the agency interpretation entitles it to controlling weight. See, *e. g., Christopher* v. *SmithKline Beecham Corp.*, 567 U. S. 142, 155. The inquiry along this dimension does not reduce to an exhaustive test, but the Court has laid out some especially important markers for identifying when *Auer* deference is and is not appropriate. To begin with, the regulatory interpretation must be the agency’s authoritative or official position, rather than any more *ad hoc* statement not reflecting the agency’s views. Next, the agency’s interpretation must in some way implicate its substantive expertise, as the basis for deference ebbs when the subject matter of a dispute is distant from the agency’s ordinary duties. Finally, an agency’s reading of a rule must reflect its “fair and considered judgment.” *Auer*, 519 U. S., at 462. A court should decline to defer, for example, to a merely “ ‘convenient litigating position,’ ” *Christopher*, 567 U. S., at 155*,* or to a new interpretation that creates “unfair surprise” to regulated parties, *Long Island Care at Home, Ltd.* v. *Coke*, 551 U. S. 158, 170. Pp. 573–580.

(b) *Stare decisis* cuts strongly against overruling *Auer.* Adherence to precedent is “a foundation stone of the rule of law,” *Michigan* v. *Bay Mills Indian Community*, 572 U. S. 782, 798, and any departure from the doctrine demands “special justification,” *Halliburton Co.* v. *Erica P. John Fund, Inc.*, 573 U. S. 258, 266. That is even more than usually so in the circumstances here. First, Kisor asks the Court to overrule a “long line of precedents”—each one reaffirming the rest and going back 75 years or more. *Bay Mills*, 572 U. S., at 798. Second, because *Auer* deference pervades the whole corpus of administrative law, abandoning it would cast doubt on many settled constructions of rules. And third, even if the Court is wrong about *Auer*, “Congress remains free to alter what \[the Court has\] done.” *Patterson* v. *McLean Credit Union*, 491 U. S. 164, 172–173. For approaching a century, Congress has let this deference regime work side-by-side with both the Administrative Procedure Act (APA) and the many statutes delegating rulemaking [\*560] power to agencies. This Court would thus need a particularly “special justification” to now reverse *Auer.*

Kisor offers nothing of that ilk. Nearly all of his arguments relate to whether the doctrine is wrong or poorly reasoned. He does not claim that *Auer* deference is “unworkable,” a traditional basis for overruling a case, *Patterson*, 491 U. S., at 173, or point to changes in legal rules that make *Auer* a “doctrinal dinosaur,” *Kimble* v. *Marvel Entertainment, LLC*, 576 U. S. 446, 458. Instead, his lone special justification is that the administrative state has evolved substantially since this Court decided *Seminole Rock* in 1945. It is true that agencies have far-reaching influence today; that is one reason the Court has taken care to reinforce the limits of *Auer* deference. But it is no answer to the growth of agencies for courts to take over their expertise-based, policymaking functions. Pp. 586–589.

(c) Turning to Kisor’s own case, a remand is necessary for two reasons. First, the Federal Circuit jumped the gun in declaring the VA’s regulation ambiguous before bringing all its interpretive tools to bear on the question. Second, the Federal Circuit assumed too fast that *Auer* deference should apply in the event of genuine ambiguity, rather than assessing whether the interpretation is of the sort that Congress would want to receive deference. On remand, the Court of Appeals must reconsider whether *Auer* deference is warranted, bearing in mind the principles outlined in this opinion. Pp. 589–590.

JUSTICE KAGAN, joined by JUSTICE GINSBURG, JUSTICE BREYER, and JUSTICE SOTOMAYOR, concluded in Parts II–A and III–A:

(a) *Auer* deference is rooted in 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. In part, the presumption arises because the agency that promulgated a rule is in the “better position \[to\] reconstruct” its original meaning. *Id.,* at 152. In still greater measure, the presumption stems from an awareness that resolving genuine regulatory ambiguities often “ ‘entail\[s\] the exercise of judgment grounded in policy concerns,’ ” an area where agencies have a comparative advantage over courts. *Thomas Jefferson Univ.* v. *Shalala*, 512 U. S. 504, 512. Finally, the presumption reflects the well-known benefits of uniformity in interpreting ambiguous rules. *Auer* deference promotes “resolving interpretive issues by uniform administrative decision, rather than piecemeal by litigation,” *Ford Motor Credit Co.* v. *Milhollin*, 444 U. S. 555, 568. Pp. 566–573.

(b) None of Kisor’s arguments provide good reason to reconsider *Auer* deference. First, he claims that *Auer* is inconsistent with the APA’s judicial review provision, which instructs reviewing courts to [\*561] “determine the meaning” of an agency action. 5 U. S. C. §706. Even when a court defers to a regulatory reading, however, 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. This Court thus presumes that Congress would want courts to do so by reviewing agency interpretations for reasonableness. That is especially so because Section 706, when enacted, was understood to restate the present law of judicial review—which would have included deference under *Seminole Rock*. Nor does *Auer* circumvent the APA’s rulemaking requirements, which require regulations to go through notice and comment before they can bind third parties. Even though a court might defer to an agency’s interpretation of a regulation, the agency’s interpretation itself never forms the basis for an enforcement action. Rather, an agency bringing an enforcement action must always rely on a rule that went through notice and comment. And courts, in turn, always retain the final authority to approve—or not—an agency’s reading of that notice-and-comment rule. See *Perez* v. *Mortgage Bankers Assn.*, 575 U. S. 92, 104, n. 4.

Kisor’s policy and constitutional arguments fail just as roundly. As a policy matter, he contends that *Auer* encourages agencies to issue vague and open-ended regulations, confident that they can later impose whatever interpretation of those rules they prefer. But no real evidence backs up that assertion and strong incentives cut in the opposite direction. Finally, Kisor asserts that *Auer* deference violates “separationof-powers principles” by vesting both legislative and judicial functions in one branch. If that objection is to agencies’ usurping the interpretive role of courts, *Auer*—when properly understood and applied—does no such thing. And if the objection is instead to the supposed commingling of functions within an agency, this Court has explained that even when agency “activities take ‘legislative’ and ‘judicial’ forms,” they continue to be “exercises of the ‘executive Power,’ ” and thus raise no constitutional concerns. *Arlington*, 569 U. S., at 304–305, n. 4. Pp. 580–586. KAGAN, J., announced the judgment of the Court and delivered the opin

ion of the Court with respect to Parts I, II–B, III–B, and IV, in which ROBERTS, C. J., and GINSBURG, BREYER, and SOTOMAYOR, JJ., joined, and an opinion with respect to Parts II–A and III–A, in which GINSBURG, BREYER, and SOTOMAYOR, JJ., joined. ROBERTS, C. J., filed an opinion concurring in part, *post*, p. 590. GORSUCH, J., filed an opinion concurring in the judgment, in which THOMAS, J., joined, in which KAVANAUGH, J., joined as to Parts I, II, III, IV, and V, and in which ALITO, J., joined as to Parts I, II, and III, *post*, p. 592. KAVANAUGH, J., filed an opinion concurring in the judgment, in which ALITO, J., joined, *post*, p. 631.

\* Briefs of *amici curiae* urging reversal were filed for the State of Utah et al. by *Sean D*. *Reyes*, Attorney General of Utah, *Tyler R*. *Green*, Solicitor General, and *Stanford E*. *Purser*, Deputy Solicitor General, and by the Attorneys General for their respective States as follows: *Steve Marshall* of Alabama, *Kevin G*. *Clarkson* of Alaska, *Mark Brnovich* of Arizona, *Leslie Rutledge* of Arkansas, *Christopher M. Carr* of Georgia, *Curtis T. Hill, Jr.*, of Indiana, *Derek Schmidt* of Kansas, *Jeff Landry* of Louisiana, *Eric S*. *Schmitt* of Missouri, *Doug Peterson* of Nebraska, *Dave Yost* of Ohio, *Mike Hunter* of Oklahoma, *Alan Wilson* of South Carolina, *Ken Paxton* of Texas, and *Patrick Morrisey* of West Virginia; for the American Federation of Labor and Congress of Industrial Organizations by *Harold C*. *Becker* and *Matthew J*. *Ginsburg*; for the Atlantic Legal Foundation by *Martin S*. *Kaufman*; for the Cato Institute et al. by *Ilya Shapiro* and *John J*. *Vecchione*; for the Center for Constitutional Jurisprudence by *John C*. *Eastman* and *Anthony T*. *Caso*; for the Center for Workplace Compliance by *Rae T*. *Vann* and *Michael P*. *Bracken*; for the Chamber of Commerce of the United States of America et al. by *Mark W*. *Mosier*, *Kevin F*. *King*, and *Daniel Saphire*; for the National Association of Home Builders et al. by *Evan A*. *Young*, *Megan Berge*, *Jared Wigginton*, *Amy Chai*, *Thomas J*. *Ward*, *Karen R*. *Harned*, *Elizabeth Milito*, *Scott Yager*, *Ellen Steen*, *Travis Cushman*, *Peter C*. *Tolsdorf*, and *Richard Moskowitz*; for the National Immigrant Justice Center et al. by *Jean-Claude André* and *Charles Roth*; for the National Organization of Veterans’ Advocates, Inc., et al. by *Roman Martinez*, *Benjamin W*. *Snyder*, *Kevin J*. *Bartlett*, *Christopher J*. *Clay*, *Leonard J*. *Selfon*, and *Linda E*. *Blauhut*; for the National Right to Work Legal Defense Foundation, Inc., by *Raymond J*. *La-Jeunesse*, *Jr*.; for the National Veterans Legal Services Program et al. by *Amy Mason Saharia*, *Liam J*. *Montgomery*, *William T*. *Marks*, and *Barton F*. *Stichman*; for the New Civil Liberties Alliance by *Jonathan F*. *Mitchell* and *Steve Simpson*; for the Pacific Legal Foundation et al. by *Anthony L*. *François*, *Damien M*. *Schiff*, *Daniel M*. *Ortner*, and *Todd F*. *Gaziano*; for the Southeastern Legal Foundation et al. by *Kimberly S*. *Hermann*, *Braden Boucek*, *Robert Henneke*, *Jonathan Riches*, and *Aditya Dynar*; for State and Local Government Associations by *Allyson N*. *Ho*, *Kathryn Cherry*, *Bradley G*. *Hubbard*, and *Lisa E*. *Soronen*; for the Utility Air Regulatory Group by *Elbert Lin*, *F*. *William Brownell*, *Henry V*. *Nickel*, and *Nash E*. *Long III*; for the Washington Legal Foundation by *Corbin K*. *Barthold* and *Cory L*. *Andrews*; for Ronald A. Cass et al. by *Mr*. *Cass*, *pro se*; for Stephen C. Connaker et al. by *William J*. *Brown*, *Jr*., and *Matthew K*. *Wegner*; for Jeremy C. Doerre by *Mr. Doerre*, *pro se*; for Jeff S. Howard by *Shawn T*. *Welch*; and for Thomas Merrill by *William M*. *Jay* and *Mr*. *Merrill*, *pro se*.

Briefs of *amici curiae* urging affirmance were filed for Administrative Law Scholars by *Donald B*. *Verrilli*, *Jr*., and *Ronald M*. *Levin* and *Gillian E*. *Metzger,* both *pro se*; and for Sen. Sheldon Whitehouse by *Mr*. *Whitehouse*, *pro se*, and *Jason P*. *Steed*.

Briefs of *amici curiae* were filed for Professors of Administrative Law and Federal Regulation by *Vincent Levy*, *Daniel M*. *Sullivan*, *Gregory Dubinsky*, and *Samuel Estreicher*, *pro se*; and for Public Citizen by *Scott L*. *Nelson* and *Allison M*. *Zieve*.

## Counsel

*Paul W*. *Hughes* argued the cause for petitioner*.* With him on the briefs were *Michael B*. *Kimberly*, *Andrew J*. *Pincus*, *Charles A*. *Rothfeld*, *E*. *Brantley Webb*, *Kenneth M*. *Carpenter*, and *Eugene R*. *Fidell*.

*Solicitor General Francisco* argued the cause for respondent*.* With him on the brief were *Assistant Attorney General Hunt*, *Deputy Solicitor General Wall*, *Deputy Assistant Attorney General Mooppan*, *Matthew Guarnieri*, *Mark B*. *Stern*, and *Daniel Aguilar*.\*

## Opinion of the Court

*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.*

This 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 (1997); *Bowles* v. *Seminole Rock & Sand Co.*, 325 U. S. 410 (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[\*564]  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 our 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 CFR § 3.156(c)(1) (2013). The Board acknowledged that Kisor had come up with two new service records, both confirming his participation[\*565]  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. 1050 (2018).

### [\*566] II

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 (1947). Or a “problem\[ \] may arise” that the agency, when drafting the rule, “could not \[have\] reasonably foresee\[n\].” *Id.,* at 202. 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 Disa

> bilities 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 CFR 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 [\*567] 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) re

> quires 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 truffie 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 CFR § 50.20(a) (1992). 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 ex

> clusive rights to drug products if they contain “no active moiety that has been approved by FDA in any other” new drug application. 21 CFR § 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.[^s3-1]

• 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 [\*568] if they are paid on a “salary basis.” 29 CFR § 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.

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.[^s3-2]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.[^s3-3]And *Seminole Rock* itself was not built on [\*569] 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 (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 (1991). Congress, we have pointed out, routinely delegates to agencies the power to implement statutes by issuing rules. See *id.,* at 151. In doing so, Congress knows (how could it not?) that regulations will sometimes contain ambiguities. See *supra,* at 566. 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 [\*570] agency’s delegated lawmaking powers.” *Martin*, 499 U. S., at 151. 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. 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 (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 567. 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 (1986). Want to know what a rule means? Ask its author.

In 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.” [\*571] *Thomas Jefferson Univ.* v. *Shalala*, 512 U. S. 504, 512 (1994) (internal quotation marks omitted). Return to our TSA example. See *supra,* at 567. 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 truffie 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 566–567. 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 567. 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; see *Thomas Jefferson*, 512 U. S., at 512. 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 (2007). And agencies (again unlike courts) have political accountability, because they are subject to the [\*572] 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 (2010); *Pauley* v. *BethEnergy Mines, Inc.*, 501 U. S. 680, 696 (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 through litigation.” *Ford Motor Credit Co.* v. *Milhollin*, 444 U. S. 555, 568 (1980). That preference may be strongest when the interpretive issue arises in the context of a “complex and highly technical regulatory program.” *Thomas Jefferson*, 512 U. S., at 512. 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 567–568. 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. Had the agency issued an interpretation before all those rulings [\*573] (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 (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 566–573. 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 (quoting *Skidmore* v. *Swift & Co.*, 323 U. S. 134, 140 (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. 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); cf. *United States* v. [\*574] *Mead Corp.*, 533 U. S. 218, 229–231 (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 (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. See, *e. g., Thorpe* v. *Housing Authority of Durham*, 393 U. S. 268, 276, and nn. 22–23 (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—may suggest a caricature of the doctrine, in which deference is “reflexive.” *Pereira* v. *Sessions*, 585 U. S. 198, 221 (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.[^s3-4]

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 (2000); *Seminole Rock*, 325 U. S., at 414 (deferring only “if the meaning of the words used is in doubt”). If uncertainty does not [\*575] 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 570–572. 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. *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 (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. 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 (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 [\*576] U. S., at 515. In other words, it 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 (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; see also *Mead*, 533 U. S., at 229–231, 236–237 (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 569–573. 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 (“tailor\[ing\] deference to \[the\] variety” of administrative action); *Arlington*, 569 U. S., at 309–310 (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[\*577]  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 (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 (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 “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 [\*578] agency.” *Martin*, 499 U. S., at 153. So the basis for deference ebbs when “\[t\]he subject matter of the \[dispute is\] dis-tan\[t\] from the agency’s ordinary” duties or “fall\[s\] within the scope of another agency’s authority.” *Arlington*, 569 U. S., at 309 (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. 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. 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 567. 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 *supra*, at 566–567. 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. High-lands Conservancy, Inc.* v. *Norton*, 343 F. 3d 239 (CA4 2003). Cf. *Adams Fruit Co.* v. *Barrett*, 494 U. S. 638, 649–650 (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.[^s3-5]

[\*579] Finally, an agency’s reading of a rule must reflect “fair and considered judgment” to receive *Auer* deference. *Christopher*, 567 U. S., at 155 (quoting *Auer*, 519 U. S., at 462). 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 (quoting *Bowen* v. *Georgetown Univ. Hospital*, 488 U. S. 204, 213 (1988), and *Auer*, 519 U. S., at 462).[^s3-6]And a court may not 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. 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. 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. Here too the lack of “fair warning” outweighed the reasons to apply *Auer. Id.,* at 156 (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 [\*581] 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” of 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 574–579. 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[\*582]  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 569–573. 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 (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 (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 APA not to “significantly alter the common law of judicial review of agency action.” *Heckler* v. *Chaney*, 470 U. S. 821, 832 (1985) (internal quotation marks omitted). [\*583] That pre-APA common law included *Seminole Rock* itself (decided the year before) along with prior decisions foretelling that ruling. See *supra,* at 568–569. 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, 97 (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 97.

[\*584] 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 104, 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 574–579. 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 104, 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. Remember that a court may defer to only an agency’s authoritative and considered judgments. See *supra,* at 576–579. 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 579. 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.

[\*585] 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 (citing Manning, *supra,* at 655–668); *Decker* v. *Northwest Environmental Defense Center*, 568 U. S. 597, 620–621 (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 (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 [\*586] 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 fieetingly) 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 574–579; *Mortgage Bankers,* 575 U. S., at 104, 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, *e. g., Withrow* v. *Larkin*, 421 U. S. 35, 54 (1975) (permitting such a combination of functions); *FTC* v. *Cement Institute*, 333 U. S. 683, 702 (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. 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. 446, 455 (2015). Adherence to precedent is “a foundation stone of the rule of law.” *Michigan* v. *Bay Mills Indian Community*, 572 U. S. 782, 798 (2014). “\[I\]t promotes [\*587] 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 (1991). To be sure, *stare decisis* is “not an inexorable command.” *Id.,* at 828. 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 (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; 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 (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 [\*588] as that branch elects.” *Kimble*, 576 U. S., at 456. 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 that 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; *supra*, at 569–570. 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 (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. Nor does he point to changes in legal rules that make *Auer* a “doctrinal dinosaur.” *Kimble*, 576 U. S., at 458. 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, [\*589] 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 567. 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 564–565. 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 574–575. 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 574–575. Rather, the court must make a conscientious effort to determine, based on indicia like text, structure, history, and purpose, whether the regulation really has more [\*590] 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 576–579. 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 (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.*

[^s3-1]: In case you’re wondering, the regulatory definition of active moiety is “\[t\]he molecule or ion, excluding those appended portions of the molecule that cause the drug to be an ester, salt (including a salt with hydrogen or coordination bonds), or the noncovalent derivative (such as a complex, chelate, or clathrate) of the molecule, responsible for the physiological or pharmacological action of the drug substance.” 21 CFR § 314.3(b) (2018).

[^s3-2]: See, *e. g.*, *PLIVA, Inc.* v. *Mensing*, 564 U. S. 604, 613 (2011); *Chase Bank USA, N. A.* v. *McCoy*, 562 U. S. 195, 208–210 (2011); *Coeur Alaska, Inc.* v. *Southeast Alaska Conservation Council*, 557 U. S. 261, 274–275 (2009); *Riegel* v. *Medtronic, Inc.*, 552 U. S. 312, 328 (2008); *Long Island Care at Home, Ltd.* v. *Coke*, 551 U. S. 158, 171 (2007); *Washington State Dept. of Social and Health Servs.* v. *Guardianship Estate of Keffeler*, 537 U. S. 371, 387–388 (2003).

[^s3-3]: Our (pre-*Auer*) decisions applying *Seminole Rock* deference are legion. See, *e. g.*, *Shalala* v. *Guernsey Memorial Hospital*, 514 U. S. 87, 94–95 (1995); *Thomas Jefferson Univ*. v. *Shalala*, 512 U. S. 504, 512 (1994); *Stinson* v. *United States*, 508 U. S. 36, 44–45 (1993); *INS* v. *National Center for Immigrants’ Rights, Inc.*, 502 U. S. 183, 189–190 (1991); *Robertson* v. *Methow Valley Citizens Council*, 490 U. S. 332, 358–359 (1989); *Mullins Coal Co. of Va.* v. *Director, Office of Workers’ Compensation Programs*, 484 U. S. 135, 159 (1987); *Lyng* v. *Payne*, 476 U. S. 926, 939 (1986); *Fidelity Fed. Sav. & Loan Assn.* v. *De la Cuesta*, 458 U. S. 141, 158, n. 13 (1982); *Blanding* v. *DuBose*, 454 U. S. 393, 401 (1982) (*per curiam*); *Ford Motor Credit Co.* v. *Milhollin*, 444 U. S. 555, 566 (1980); *United States* v. *Larionoff*, 431 U. S. 864, 872 (1977); *Northern Indiana Public Service Co.* v. *Porter County Chapter of Izaak Walton League of America, Inc.*, 423 U. S. 12, 15 (1975) (*per curiam*); *Ehlert* v. *United States*, 402 U. S. 99, 105 (1971); *INS* v. *Stanisic*, 395 U. S. 62, 72 (1969); *Thorpe* v. *Housing Authority of Durham*, 393 U. S. 268, 276 (1969); *Udall* v. *Tallman*, 380 U. S. 1, 16–17 (1965).

[^s3-4]: The proper understanding of the scope and limits of the *Auer* doctrine is, of course, not set out in any of the opinions that concur only in the judgment.

[^s3-5]: For a similar reason, this Court has denied *Auer* deference when an agency interprets a rule that parrots the statutory text. See *Gonzales* v. *Oregon*, 546 U. S. 243, 257 (2006). An agency, we explained, gets no “spe cial authority to interpret its own words when, instead of using its expertise and experience to formulate a regulation, it has elected merely to paraphrase the statutory language.” *Ibid.*

[^s3-6]: The general rule, then, is not to give deference to agency interpretations advanced for the first time in legal briefs. See *Bowen*, 488 U. S., at 212–213. But we have not entirely foreclosed that practice. *Auer* itself deferred to a new regulatory interpretation presented in an *amicus curiae* brief in this Court. There, the agency was not a party to the litigation, and had expressed its views only in response to the Court’s request. “\[I\]n the circumstances,” the Court explained, “\[t\]here \[was\] simply no reason to suspect that the interpretation \[did\] not reflect the agency’s fair and considered judgment on the matter in question.” *Auer*, 519 U. S., at 462.

## Chief Justice Roberts, concurring in part

*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 (1997), and *Bowles* v. *Seminole Rock & Sand Co.*, 325 U. S. 410 (1945). For the reasons the Court discusses in Part III–B, I agree that overruling those [\*591] 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 (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 cases 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 (1984). I do not regard the Court’s decision today to touch upon the latter question.

## Justice Gorsuch, concurring in the judgment

*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*.[^s5-1]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.” [^s5-2]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 [\*593] 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 advanced 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 [\*594] 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.[^s5-3]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.” [^s5-4]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.[^s5-5]But courts respected executive interpretations only because and to the extent “they embodied understandings made roughly contemporaneously with . . . enactment [\*595] and stably maintained and practiced since that time,” not “because they were executive as such.” [^s5-6]

Writing 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*.[^s5-7]*Ante*, at 569. 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.” [^s5-8]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.” [^s5-9]*Eaton* thus simply followed the well-worn path of acknowledging that an agency’s interpretation of a regulation can supply *evidence* of its meaning.[^s5-10]Nowhere did the Court even hint that it would have deferred to the State Department’s[\*596]  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.[^s5-11]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.” [^s5-12]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.[^s5-13]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.” [^s5-14]At the time, the influential administrative law scholar Kenneth Culp Davis considered this “\[a\]n entirely reliable statement” of the law.[^s5-15]

### [\*597] B

In truth, the seeds of the *Auer* doctrine were first planted only in 1945, in *Bowles* v. *Seminole Rock & Sand Co.*[^s5-16]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.” [^s5-17]

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.[^s5-18]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.” [^s5-19]

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.[^s5-20]The dictum, too, was readily explained as reflecting the unusual factual context in which the case [\*598] arose, involving an emergency government program created to deal with “unique circumstances of war and economic depression.” [^s5-21]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.[^s5-22]

No 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.” [^s5-23]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.[^s5-24]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.” [^s5-25]The letter, the court stressed, [\*599] 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.” [^s5-26]

### C

This Court did not cite *Seminole Rock*’s “controlling weight” dictum again until 1965, in *Udall* v. *Tallman*.[^s5-27]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.” [^s5-28]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*.[^s5-29]

*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.” [^s5-30]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.

[\*600] 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, the courts at least do not.” [^s5-31]In the real world the judge uses his traditional interpretive toolkit, full of canons and tie-breaking 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.” [^s5-32]

If that were not troubling enough, *Auer* has also become “a doctrine of uncertain scope and application.” [^s5-33]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.[^s5-34]Deepening the confusion, this Court and lower courts have, over time, tried to soften *Auer*’s rigidity by declaring that it “might” not apply [\*601] 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.[^s5-35]All this has resulted in “widespread confusion” about when and how to apply *Auer* deference.[^s5-36]

In light of *Auer*’s many problems, it should come as no surprise that several Members of this Court,[^s5-37]along with a great many lower court judges[^s5-38]and members of the legal [\*602] academy,[^s5-39]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 [\*603] choice but to treat the Board’s interpretation as “ ‘controlling’ ” unless that interpretation was “ ‘plainly erroneous or inconsistent with the regulatio\[n\].’ ” [^s5-40]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.” [^s5-41]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.[^s5-42]Mr. Kisor then asked us to grant certiorari to reconsider *Auer*. Thinking it past time to do so, we granted the petition.[^s5-43]

### II. 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-lawmaking 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.[^s5-44]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.

### [\*604] 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.” [^s5-45]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.[^s5-46]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.[^s5-47]

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 581–582.

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 [\*605] 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.[^s5-48]But when it comes to the business of interpreting 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.” [^s5-49]

[\*606] 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 decisionmaking process. See *supra*, at 597. But in light of *National Cable & Telecommunications Assn.* v. *Brand X Internet Services*,[^s5-50]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.[^s5-51]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.[^s5-52]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 case “is in the public interest.” [^s5-53]If a court thought the proposed judgment harmful to the public but decided to defer to the government’s “reasonable” contrary [\*607] 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.[^s5-54]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.[^s5-55]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, [\*608] 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.” [^s5-56]*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.[^s5-57]

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.[^s5-58]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 rule 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.[^s5-59]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.[^s5-60]Whether an agency issues[\*609]  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 583–584. 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.” [^s5-61]To suggest that *Auer* does not make an agency’s interpretive guidance “binding o\[n\] anyone,” *ante*, at 584, 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 569. 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 569–570. But against [\*610] 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.[^s5-62]

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 courts to afford agencies *Auer* deference. *Ante*, at 582–583. 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 582. 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.” [^s5-63]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.” [^s5-64]

[\*611] 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.” [^s5-65]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.” [^s5-66]

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,[^s5-67]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.[^s5-68]In truth, when Congress passed [\*612] the APA the law of judicial review of agency action was in a confused state. During the 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.” [^s5-69]

### 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.’ ” [^s5-70]As Chief Justice Marshall put it, “\[i\]t is emphatically the province and duty of the judicial department to say what the law is.” [^s5-71]And never, this Court has warned, should the “judicial power . . . be shared with \[the\] Executive Branch.” [^s5-72]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 decisionmaking. One of the abuses of royal power that led to the [\*613] American Revolution was King George’s attempt to gain influence over colonial judges.[^s5-73]Colonial legislatures, too, had interfered with the courts’ independence “at the behest of private interests and factions.” [^s5-74]These experiences had taught the founders that “ ‘there is no liberty if the power of judging be not separated from the legislative and executive powers.’ ” [^s5-75]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.[^s5-76]Maybe 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. [\*614] 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,” [^s5-77]they designed a judiciary that would be able to interpret the laws “free from potential domination by other branches of government.” [^s5-78]To that end, they resisted proposals that would have subjected judicial decisions to review by political actors.[^s5-79]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.” [^s5-80]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.[^s5-81]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.’ ” [^s5-82]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 [\*615] power, or the Congress share with the Judiciary the power to override a Presidential veto.” [^s5-83]

*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’ ” [^s5-84]and are part of the body of federal law, 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.” [^s5-85]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.[^s5-86]*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 [\*616] own views.’ ” [^s5-87]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.” [^s5-88]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.[^s5-89]

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.[^s5-90]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.[^s5-91]

### [\*617] B

What do our colleagues have to say about these concerns? A majority has nothing 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 586. 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.” [^s5-92]

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* [\*618] of cases they may decide,” or what relief they may supply, but a forbidden attempt “to prescribe or superintend *how* they decide those cases.” [^s5-93]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.” [^s5-94]

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 585. 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 its 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.

### [\*619] 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 570. 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.” *Ibid*. 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.” [^s5-95]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.” [^s5-96]If the best reading of the regulation turns out to be something other than what [\*620] 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 controlling one.[^s5-97]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 571. 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 [\*621] 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 572, 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.” [^s5-98]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 [\*622] public meaning. Today it is even said that we judges are, to one degree or another, “all textualists now.” [^s5-99]

Pursuing a more modest tack, JUSTICE KAGAN next suggests that *Auer* is justified by the respect due agencies’ “technical” expertise. *Ante*, at 571. 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.[^s5-100]The fact remains, however, that even agency experts “can be wrong; even Homer nodded.” [^s5-101]*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 572–573. 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 [\*623] whether deference is warranted, see *supra*, at 600–601, 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 588.

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 574. 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 575–576. 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 576–577. 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 576–579. The only certainty in all this is that the majority isn’t really much [\*624] 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 all. 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*.[^s5-102]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 [\*625] that the Court has never engaged,” it may well “exceed the limits of stare decisis.” [^s5-103]

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 587–588) 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\[s\].” [^s5-104]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 573. Congress can, of course, *rebut* the presumption on a statuteby-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.[^s5-105][\*626] [^s5-105]We should not be in the business of tossing “ ‘balls . . . into Congress’s court,’ ” *ante*, at 587, 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,’ ” [^s5-106]and this Court should not always remain bound to decisions whose “rationale no longer withstands ‘careful analysis.’ ” [^s5-107]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.” [^s5-108]As applied to *Auer*, all of these considerations[\*627]  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.[^s5-109]Today’s decision attempts just such a “refinement” by hedging *Auer* with new qualifications and limitations. See *ante*, at 573–579. This shifting ground “undermin\[es\] the force of *stare decisis*.” [^s5-110]

*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 open-ended standard like *Skidmore*.” [^s5-111]See *supra*, at 600–601. 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 576, seems destined only to compound the confusion. See *supra*, at 625–626. Many words come to mind to describe the tasks we assign lower court judges today, but “workable” is not among them.

[\*628] *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.[^s5-112]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.’ ” [^s5-113]Affording “controlling weight” to regulators’ post-promulgation views about the meaning of an ambiguous regulation is hard to square with these usual judicial practices.[^s5-114]

*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 [\*629] 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.” [^s5-115]Among other things, it produces “ ‘reams of regulations’ ” [^s5-116]—so many 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.[^s5-117]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 [\*630] countervailing interest’ ” of all citizens “ ‘in having their constitutional rights fully protected.’ ” [^s5-118]

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 587. 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.” [^s5-119]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 deference that they could have spent interpreting [\*631] 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 575. 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.

[^s5-1]: 519 U. S. 452 (1997).

[^s5-2]: Larkin & Slattery, The World After *Seminole Rock* and *Auer*, 42 Harv. J. L. & Pub. Pol’y 625, 641 (2019) (internal quotation marks omitted).

[^s5-3]: See Knudsen & Wildermuth, Unearthing the Lost History of *Seminole Rock*, 65 Emory L. J. 47, 55, 65, 68 (2015) (Lost History).

[^s5-4]: *St. Joseph Stock Yards Co.* v. *United States*, 298 U. S. 38, 73 (1936) (concurring opinion). See also *FTC* v. *Gratz*, 253 U. S. 421, 427 (1920); *ICC* v. *Union Pacific R. Co.*, 222 U. S. 541, 547 (1912); *Belden* v. *Chase*, 150 U. S. 674, 698 (1893); *Decatur* v. *Paulding*, 14 Pet. 497, 515 (1840); accord, Woolhandler, Judicial Deference to Administrative Action—A Revisionist History, 43 Admin. L. Rev. 197, 206–207 (1991).

[^s5-5]: Bamzai, The Origins of Judicial Deference to Executive Interpretation, 126 Yale L. J. 908, 930–947 (2017) (Origins).

[^s5-6]: *Id.*, at 943, 962; cf. *NLRB* v. *Noel Canning*, 573 U. S. 513, 572–573 (2014) (Scalia, J., concurring in judgment) (an “open, widespread, and unchallenged” governmental practice can “guide \[courts’\] interpretation” of an ambiguous text, but it cannot “alter” the meaning of that text); *Edwards’ Lessee* v. *Darby*, 12 Wheat. 206, 210 (1827) (“In the construction of a doubtful and ambiguous law, the cotemporaneous construction of those who were called upon to act under the law, and were appointed to carry its provisions into effect, is entitled to very great respect”).

[^s5-7]: 169 U. S. 331.

[^s5-8]: *Id.*, at 342.

[^s5-9]: *Id.*, at 342–343.

[^s5-10]: Cf. Newman, How Courts Interpret Regulations, 35 Cal. L. Rev. 509, 521, and n. 78 (1947) (noting that *Eaton* suggested administrative interpretations could be “ ‘persuasive’ but not binding”).

[^s5-11]: 323 U. S. 134.

[^s5-12]: *Id.*, at 136–137. Much of the legal analysis supporting this conclusion was contained in the companion case, *Armour & Co.* v. *Wantock*, 323 U. S. 126 (1944), which made no mention of any administrative interpretations. *Id.*, at 129–134; see *Skidmore*, 323 U. S., at 136 (citing the “reasons set forth in the *Armour* case decided herewith”).

[^s5-13]: *Id.*, at 139.

[^s5-14]: *Id.*, at 140; see also *id.*, at 139 (the agency’s views “are not, of course, conclusive, even in the cases with which they directly deal” and do not “bin\[d\] a district court’s processes, as an authoritative pronouncement of a higher court might do”).

[^s5-15]: Davis, Administrative Rules—Interpretative, Legislative, and Retroactive, 57 Yale L. J. 919, 936–939, and n. 86 (1948); see also K. Davis, Administrative Law § 249, p. 901 (1951) (“\[S\]ubstitution of judicial judg ment on the content of interpretative rules is always permissible, even though the reviewing court may give ‘weight’ or ‘great weight’ to the rule. The best guide may be the Court’s formula in *Skidmore* . . . ”).

[^s5-16]: 325 U. S. 410.

[^s5-17]: *Id.*, at 414.

[^s5-18]: *Id.*, at 414–417.

[^s5-19]: *Id.*, at 417.

[^s5-20]: See Davis, Scope of Review of Federal Administrative Action, 50 Colum. L. Rev. 559, 597 (1950).

[^s5-21]: Lost History 60; see also Anthony, The Supreme Court and the APA: Sometimes They Just Don’t Get It, 10 Admin. L. J. Am. U. 1, 12 (1996).

[^s5-22]: 325 U. S., at 417–418; see Pojanowski, Revisiting *Seminole Rock*, 16 Geo. J. L. & Pub. Pol’y 87, 88 (2018) (“A closer look at *Seminole Rock* suggests an unremarkable application of the less-deferential standard of review of *Skidmore*”).

[^s5-23]: Lost History 94–97; see Pojanowski, *supra*, at 92–96.

[^s5-24]: Lost History 65–68.

[^s5-25]: *Southern Goods Corp.* v. *Bowles*, 158 F. 2d 587, 590 (1946).

[^s5-26]: *Ibid.*

[^s5-27]: 380 U. S. 1, 4, 17–18 (accepting a regulatory interpretation by the Secretary of the Interior that was consistent, widely disseminated, and heavily relied upon, while not suggesting any disagreement with the Secretary’s interpretation).

[^s5-28]: Lost History 80.

[^s5-29]: See generally *id.*, at 68–92, 98.

[^s5-30]: *Id.*, at 53.

[^s5-31]: Scalia, Judicial Deference to Administrative Interpretations of Law, 1989 Duke L. J. 511, 520.

[^s5-32]: *Decker* v. *Northwest Environmental Defense Center*, 568 U. S. 597, 613 (2013); see *Pauley* v. *BethEnergy Mines*, *Inc.*, 501 U. S. 680, 702 (1991) (the agency’s interpretation “need not be the best or most natural one by grammatical or other standards”).

[^s5-33]: Hickman & Thomson, The *Chevron*ization of *Auer*, 103 Minn. L. Rev. Headnotes 103, 105 (2019).

[^s5-34]: See Kavanaugh, Fixing Statutory Interpretation, 129 Harv. L. Rev. 2118, 2134–2144 (2016).

[^s5-35]: *Christopher* v. *SmithKline Beecham Corp.*, 567 U. S. 142, 155 (2012) (alterations and internal quotation marks omitted).

[^s5-36]: Leske, Splits in the *Rock*: The Conflicting Interpretations of the *Seminole Rock* Deference Doctrine by the U. S. Courts of Appeals, 66 Admin. L. Rev. 787, 832 (2014); see Hickman & Thomson, *supra*, at 111 (noting a “glut of recent cases in which members of the same court are openly divided on the proper application of *Auer*”).

[^s5-37]: See *Perez* v. *Mortgage Bankers Assn.*, 575 U. S. 92, 107–108 (2015) (ALITO, J., concurring in part and concurring in judgment); *id.*, at 108–112 (Scalia, J., concurring in judgment); *id.*, at 119–133 (THOMAS, J., concurring in judgment); *Decker*, 568 U. S., at 615–616 (ROBERTS, C. J., joined by ALITO, J., concurring); *id.*, at 616–621 (Scalia, J., concurring in part and dissenting in part); *Talk America*, *Inc.* v. *Michigan Bell Telephone Co.*, 564 U. S. 50, 67–69 (2011) (Scalia, J., concurring); see also Kavanaugh, Keynote Address: Justice Scalia and Deference 19:06 (June 2, 2016), http://vimeo.com/169758593 (predicting “that *Auer* will someday be overruled and that Justice Scalia’s dissent in *Decker* will be the law of the land”).

[^s5-38]: See, *e. g.*, *Forrest Gen. Hospital* v. *Azar*, 926 F. 3d 221, 229–230 (CA5 2019); *San Diego Gas & Elec. Co.* v. *FERC*, 913 F. 3d 127, 145, n. 4 (CADC 2019) (Randolph, J., dissenting); *United States* v. *Havis*, 907 F. 3d 439, 450–452 (CA6 2018) (Thapar, J., concurring), vacated, 921 F. 3d 628, on reh’g en banc, 927 F. 3d 382 (CA6 2019); *Marsh* v. *J. Alexander’s LLC*, 905 F. 3d 610, 652–653 (CA9 2018) (Ikuta, J., dissenting); *Egan* v. *Delaware River Port Auth.*, 851 F. 3d 263, 279 (CA3 2017) (Jordan, J., concurring in judgment); *Perez* v. *Loren Cook Co.*, 803 F. 3d 935, 938, n. 2 (CA8 2015) (en banc); *Johnson* v. *McDonald*, 762 F. 3d 1362, 1366–1368 (CA Fed. 2014) (O’Malley, J., concurring); *Exelon Generation Co.* v. *Local 15*, *Int’l Brotherhood of Elec. Workers*, *AFL–CIO*, 676 F. 3d 566, 576, n. 5 (CA7 2012).

[^s5-39]: See, *e. g.*, Hickman & Thomson, *supra*, at 111–113; Adler, *Auer* Evasions, 16 Geo. J. L. & Pub. Pol’y 1, 26 (2018); Pojanowski, 16 Geo. J. L. & Pub. Pol’y, at 99; Knudsen & Wildermuth, Lessons From the Lost History of *Seminole Rock*, 22 Geo. Mason L. Rev. 647, 667 (2015); Leske, *supra*, at 789–793; Molot, The Judicial Perspective in the Administrative State: Reconciling Modern Doctrines of Deference with the Judiciary’s Structural Role, 53 Stan. L. Rev. 1, 108–110 (2000); Anthony, 10 Admin. L. J., at 4–12; Manning, Constitutional Structure and Judicial Deference to Agency Interpretations of Agency Rules, 96 Colum. L. Rev. 612, 696 (1996).

[^s5-40]: *Kisor* v. *Shulkin*, 869 F. 3d 1360, 1367 (2017).

[^s5-41]: *Id.*, at 1368.

[^s5-42]: *Kisor* v. *Shulkin*, 880 F. 3d 1378, 1379 (CA Fed. 2018) (opinion of O’Malley, J.).

[^s5-43]: 586 U. S. 1050 (2018).

[^s5-44]: *Abbott Laboratories* v. *Gardner*, 387 U. S. 136, 140 (1967).

[^s5-45]: 5 U. S. C. §706.

[^s5-46]: *Ibid.*; see § 551(13) (defining “agency action”).

[^s5-47]: The case before us doesn’t arise under the APA, but the statute that governs here is plainly modeled on the APA and contains essentially the same commands. It directs a reviewing court to “decide all relevant questions of law” and to “set aside any regulation or any interpretation thereof ” that is “not in accordance with law.” 38 U. S. C. §7292(d)(1).

[^s5-48]: See, *e. g.*, § 706(2)(A) (arbitrary and capricious, abuse of discretion); § 706(2)(E) (substantial evidence); see also *Universal Camera Corp.* v. *NLRB*, 340 U. S. 474, 482, n. 14 (1951) (noting that as originally proposed, the APA’s judicial review provision would have included an explicit requirement for courts to accord “due weight” to “the experience, technical competence, specialized knowledge, and legislative policy of the agency involved as well as the discretionary authority conferred upon it” (internal quotation marks omitted)).

[^s5-49]: Duffy, Administrative Common Law in Judicial Review, 77 Texas L. Rev. 113, 194–195 (1998); see Merrill, Capture Theory and the Courts: 1967–1983, 72 Chi.-Kent L. Rev. 1039, 1085–1086 (1997) (noting the “embarrassing” fact that “the APA appears to compel th\[e\] conclusion” that “courts should decide all questions of law de novo”). See also, *e. g.*, Origins 985; Mashaw, Rethinking Judicial Review of Administrative Action: A Nineteenth Century Perspective, 32 Cardozo L. Rev. 2241, 2243 (2011); Garrett, Legislating *Chevron*, 101 Mich. L. Rev. 2637, 2640 (2003); Molot, Reexamining *Marbury* in the Administrative State: A Structural and Institutional Defense of Judicial Power over Statutory Interpretation, 96 Nw. U. L. Rev. 1239, 1249 (2002); Anthony, 10 Admin. L. J. Am. U., at 9–10; Farina, Statutory Interpretation and the Balance of Power in the Administrative State, 89 Colum. L. Rev. 452, 473, and n. 85 (1989); Starr, Sunstein, Willard, & Morrison, Judicial Review of Administrative Action in a Conservative Era, 39 Admin. L. Rev. 353, 368 (1987) (remarks of Prof. Sunstein); Pierce & Shapiro, Political and Judicial Review of Agency Action, 59 Texas L. Rev. 1175, 1182 (1981); 4 K. Davis, Administrative Law § 30.01, pp. 190–191 (1958).

[^s5-50]: 545 U. S. 967 (2005).

[^s5-51]: See, *e. g.*, *In re Lovin*, 652 F. 3d 1349, 1353–1354 (CA Fed. 2011); *Levy* v. *Sterling Holding Co.*, 544 F. 3d 493, 502–503 (CA3 2008).

[^s5-52]: 18 U. S. C. § 3553(a).

[^s5-53]: 15 U. S. C. § 16(e)(1).

[^s5-54]: See *Perez*, 575 U. S., at 95–96.

[^s5-55]: *United States* v. *Nixon*, 418 U. S. 683, 695–696 (1974).

[^s5-56]: *Perez*, 575 U. S., at 127 (THOMAS, J., concurring in judgment).

[^s5-57]: *Ibid.*; see *id.*, at 110 (Scalia, J., concurring in judgment) (*Auer* lets agencies “use \[interpretive\] rules not just to advise the public, but also to bind them”).

[^s5-58]: See *INS* v. *Chadha*, 462 U. S. 919, 951, 954 (1983).

[^s5-59]: Brief for Administrative Law Scholars as *Amici Curiae* 9–10, n. 4.

[^s5-60]: See *Perez*, 575 U. S., at 101; *Marseilles Land & Water Co.* v. *FERC*, 345 F. 3d 916, 920 (CADC 2003).

[^s5-61]: *Perez*, 575 U. S., at 110–111 (Scalia, J., concurring in judgment).

[^s5-62]: *Milner* v. *Department of Navy*, 562 U. S. 562, 572 (2011).

[^s5-63]: McCarran, Improving “Administrative Justice”: Hearings and Evidence; Scope of Judicial Review, 32 A. B. A. J. 827, 893 (1946).

[^s5-64]: H. R. Rep. No. 1980, 79th Cong., 2d Sess., 44 (1946); accord, S. Rep. No. 752, 79th Cong., 1st Sess., 28 (1945); 92 Cong. Rec. 5654 (1946) (statement of Rep. Walter). See also Shepherd, Fierce Compromise: The Ad ministrative Procedure Act Emerges from New Deal Politics, 90 Nw. U. L. Rev. 1557, 1662–1666 (1996).

[^s5-65]: *Universal Camera*, 340 U. S., at 490 (emphasis added).

[^s5-66]: Dickinson, Administrative Procedure Act: Scope and Grounds of Broadened Judicial Review, 33 A. B. A. J. 434, 516 (1947). See also Origins 990–991 (critiquing the Attorney General’s characterization of the APA as “inherently question begging” and unsupported by any analysis).

[^s5-67]: Adler, 16 Geo. J. L. & Pub. Pol’y, at 7; see Lost History 63; Pojanowski, 16 Geo. J. L. & Pub. Pol’y, at 95–96.

[^s5-68]: Davis, 50 Colum. L. Rev., at 597–598; see also Davis, 57 Yale L. J., at 936, n. 72; Newman, 35 Cal. L. Rev., at 521–522.

[^s5-69]: Hearings on H. R. 184 et al. before the House Committee on the Judiciary, 79th Cong., 1st Sess., 38 (1945); see Origins 988–989.

[^s5-70]: *Patchak* v. *Zinke*, 583 U. S. 244, 249 (2018) (plurality opinion) (quoting *Massachusetts* v. *Mellon*, 262 U. S. 447, 488 (1923)).

[^s5-71]: *Marbury* v. *Madison*, 1 Cranch 137, 177 (1803); see also *Wayman* v. *Southard*, 10 Wheat. 1, 46 (1825) (“\[T\]he legislature makes, the executive executes, and the judiciary construes the law”); The Federalist No. 78, p. 467 (C. Rossiter ed. 1961) (A. Hamilton).

[^s5-72]: *Miller* v. *Johnson*, 515 U. S. 900, 922 (1995).

[^s5-73]: See Declaration of Independence ¶11.

[^s5-74]: *Plaut* v. *Spendthrift Farm*, *Inc.*, 514 U. S. 211, 220–221 (1995).

[^s5-75]: The Federalist No. 78, at 466.

[^s5-76]: *Palmore* v. *United States*, 411 U. S. 389, 412 (1973) (Douglas, J., dissenting); see *Oil States Energy Services*, *LLC* v. *Greene’s Energy Group*, *LLC*, 584 U. S. 325, 348 (2018) (GORSUCH, J., dissenting) (“\[W\]hen an independent Judiciary gives ground to bureaucrats in the adjudication of cases, the losers will often prove the unpopular and vulnerable”); *United States* v. *Hatter*, 532 U. S. 557, 568–569 (2001) (quoting John Marshall’s admonition that a judge who may be called on to decide a dispute “ ‘between the most powerful individual in the community, and the poorest and most unpopular,’ ” must be “ ‘perfectly and completely independent, with nothing to influence or control him but God and his conscience’ ” (alterations omitted)); Jackson, The Meaning of Statutes: What Congress Says or What the Court Says, 34 A. B. A. J. 535, 536 (1948) (“\[T\]he interpretation of \[the laws’\] fair meaning . . . should be made by judges as independent of politics as humanly possible and not serving the interests of the class for whom, or a majority by whom, legislation is enacted”).

[^s5-77]: 2 Records of the Federal Convention of 1787, p. 75 (M. Farrand ed. 1911); see also Manning, 96 Colum. L. Rev., at 640–648.

[^s5-78]: *United States* v. *Will*, 449 U. S. 200, 218 (1980).

[^s5-79]: See The Federalist No. 81, at 482 (A. Hamilton).

[^s5-80]: *Id.*, at 483.

[^s5-81]: *Northern Pipeline Constr. Co.* v. *Marathon Pipe Line Co.*, 458 U. S. 50, 60 (1982) (plurality opinion).

[^s5-82]: *Bank Markazi* v. *Peterson*, 578 U. S. 212, 225 (2016) (alterations omitted).

[^s5-83]: *Stern* v. *Marshall*, 564 U. S. 462, 483 (2011) (internal quotation marks omitted).

[^s5-84]: *Perez*, 575 U. S., at 96; see *Chrysler Corp.* v. *Brown*, 441 U. S. 281, 295–296 (1979). To be sure, our precedent allowing executive agencies to issue legally binding regulations to govern private conduct may raise constitutional questions of its own. See, *e. g.*, *Department of Transportation* v. *Association of American Railroads*, 575 U. S. 43, 70–76 (2015) (THOMAS, J., concurring in judgment).

[^s5-85]: Cox, Judge Learned Hand and the Interpretation of Statutes, 60 Harv. L. Rev. 370, 390 (1947).

[^s5-86]: *Id.*, at 390–391, and n. 58; see also Kavanaugh, 129 Harv. L. Rev., at 2151 (in pursuing their policy goals, “\[e\]xecutive branch agencies often think they can take a particular action unless it is *clearly forbidden*”).

[^s5-87]: *Plaut*, 514 U. S., at 225 (quoting T. Cooley, Constitutional Limitations 95 (1868)).

[^s5-88]: *Id.*, at 95; see also *Bank Markazi*, 578 U. S., at 225, n. 17.

[^s5-89]: *Ogden* v. *Blackledge*, 2 Cranch 272, 277.

[^s5-90]: Cooley, *supra*, at 95.

[^s5-91]: Cf. *Cary* v. *Curtis*, 3 How. 236, 253, 257 (1845) (Story, J., dissenting) (if the “right to interpret the laws” is taken away from courts and “confided to an executive functionary,” then “the judicial power, designed by the Constitution to be the final and appellate jurisdiction to interpret our laws, is superseded in its most vital and important functions”).

[^s5-92]: *Marbury*, 1 Cranch, at 177.

[^s5-93]: *Arlington* v. *FCC*, 569 U. S. 290, 297 (2013) (emphasis added).

[^s5-94]: *Clinton* v. *City of New York*, 524 U. S. 417, 452 (1998) (Kennedy, J., concurring).

[^s5-95]: *Marbury*, 1 Cranch, at 163.

[^s5-96]: Holmes, The Theory of Legal Interpretation, 12 Harv. L. Rev. 417, 417–418 (1899); see *INS* v. *Cardoza-Fonseca*, 480 U. S. 421, 452–453 (1987) (Scalia, J., concurring in judgment) (“Judges interpret laws rather than reconstruct legislators’ intentions”); H. Hart & A. Sacks, The Legal Process 1375 (1994) (“Unenacted intentions or wishes cannot be given effect as law”).

[^s5-97]: *United States* v. *Woods*, 571 U. S. 31, 48 (2013).

[^s5-98]: O’Scannlain, “We Are All Textualists Now”: The Legacy of Justice Antonin Scalia, 91 St. John’s L. Rev. 303, 304–305 (2017) (contesting the radical indeterminacy of legal texts).

[^s5-99]: *Id.*, at 313; see Siegel, Textualism and Contextualism in Administrative Law, 78 B. U. L. Rev. 1023, 1057 (1998).

[^s5-100]: Larkin & Slattery, 42 Harv. J. L. & Pub. Pol’y, at 647.

[^s5-101]: *Ibid.*

[^s5-102]: See Criddle & Staszewski, Against Methodological Stare Decisis, 102 Geo. L. J. 1573, 1577, and n. 12 (2014); C. Oldfather, Methodological Stare Decisis and Constitutional Interpretation, in Precedent in the United States Supreme Court 135, 135–136 (C. Peters ed. 2013).

[^s5-103]: Kozel, Statutory Interpretation, Administrative Deference, and the Law of Stare Decisis, 97 Texas L. Rev. 1125, 1159 (2019); see Raso & Eskridge, *Chevron* as a Canon, Not a Precedent: An Empirical Study of What Motivates Justices in Agency Deference Cases, 110 Colum. L. Rev. 1727, 1765–1766 (2010) (concluding that in practice, this Court has not treated administrative-deference regimes such as *Chevron* and *Auer* as binding precedents).

[^s5-104]: *South Dakota* v. *Wayfair*, *Inc.*, 585 U. S. 162, 183 (2018).

[^s5-105]: See, *e. g.*, Alexander & Prakash, Mother May I? Imposing Mandatory Prospective Rules of Statutory Interpretation, 20 Const. Comment. 97 (2003); Elhauge, Preference-Estimating Statutory Default Rules, 102 Colum. L. Rev. 2027, 2109–2110, and nn. 231–233 (2002).

[^s5-106]: *Pearson* v. *Callahan*, 555 U. S. 223, 233 (2009).

[^s5-107]: *Arizona* v. *Gant*, 556 U. S. 332, 348 (2009) (quoting *Lawrence* v. *Texas*, 539 U. S. 558, 577 (2003)).

[^s5-108]: *Janus* v. *State*, *County*, *and Municipal Employees*, 585 U. S. 878, 917 (2018).

[^s5-109]: Brief for Administrative Law Scholars as *Amici Curiae* 13.

[^s5-110]: *Knick* v. *Township of Scott*, 588 U. S. 180, 204 (2019); see *Janus*, 585 U. S., at 928. See also Lost History 54–92; Knudsen & Wildermuth, 22 Geo. Mason L. Rev., at 658–664.

[^s5-111]: Hickman & Thomson, 103 Minn. L. Rev. Headnotes, at 110.

[^s5-112]: *Green* v. *Brennan*, 578 U. S. 547, 553–554 (2016); see, *e. g.*, *National Assn. of Home Builders* v. *Defenders of Wildlife*, 551 U. S. 644, 668–669 (2007) (construing regulation in light of text, history, and canon against surplusage).

[^s5-113]: *Woods*, 571 U. S., at 48; see also *Bruesewitz* v. *Wyeth LLC*, 562 U. S. 223, 242 (2011); *Jones* v. *United States*, 526 U. S. 227, 238 (1999); *United States* v. *Mine Workers*, 330 U. S. 258, 281–282 (1947).

[^s5-114]: To be sure, under *Chevron U. S. A. Inc.* v. *Natural Resources Defense Council*, *Inc.*, 467 U. S. 837 (1984), we sometimes defer to an agency’s construction of a *statute*. But there are serious questions, too, about whether *that* doctrine comports with the APA and the Constitution. See, *e. g.*, *Pereira* v. *Sessions*, 585 U. S. 198, 219–221 (2018) (Kennedy, J., concurring); *Michigan* v. *EPA*, 576 U. S. 743, 760–764 (2015) (THOMAS, J., concurring); *Perez*, 575 U. S., at 109–110 (Scalia, J., concurring in judgment). Regardless, it would be a mistake to suppose that *Auer* is in any way a “logical corollary to *Chevron*.” *Decker*, 568 U. S., at 620 (Scalia, J., concurring in part and dissenting in part).

[^s5-115]: *Arlington*, 569 U. S., at 313 (ROBERTS, C. J., dissenting) (internal quotation marks omitted).

[^s5-116]: *Federal Maritime Comm’n* v. *South Carolina Ports Authority*, 535 U. S. 743, 755 (2002).

[^s5-117]: See Office of the Federal Register, Code of Federal Regulations: Total Pages 1938–1949, and Total Volumes and Pages 1950–2018, http://www. federalregister.gov/uploads/2019/04/cfrTotalPages2018.pdf; *United States* v. *Secretary*, *Fla. Dept. of Corrections*, 778 F. 3d 1223, 1225 (CA11 2015).

[^s5-118]: *Janus*, 585 U. S., at 927 (quoting *Gant*, 556 U. S., at 349).

[^s5-119]: *Gonzales* v. *Oregon*, 546 U. S. 243, 269 (2006); see *Christopher*, 567 U. S., at 159 (applying *Skidmore* after concluding that agency’s interpretation did not merit *Auer* deference).

## Justice Kavanaugh, concurring in the judgment

*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 [\*632] as it may initially appear.” *Ante,* at 591 (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 575 (quoting *Chevron U. S. A. Inc.* v. *Natural Resources Defense Council, Inc.*, 467 U. S. 837, 843, n. 9 (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, and 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 (1983).

[\*633] 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 591. Like THE CHIEF JUSTICE, “I do not regard the Court’s decision” not to formally overrule *Auer “*to touch upon the latter question.” *Ibid.*