For more than seventy years, courts have deferred to reasonable agency interpretations of ambiguous regulations. The Auer principle, as it is now called, has attracted academic criticism and some skepticism within the Supreme Court. But the principle is entirely correct. In the absence of a clear congressional direction, courts should assume that because of their specialized competence, and their greater accountability, agencies are in the best position to decide on the meaning of ambiguous terms. The recent challenges to the Auer principle rest on fragile foundations, including an anachronistic understanding of the nature of interpretation, an overheated argument about the separation of powers, and an empirically unfounded and logically weak argument about agency incentives, which exemplifies what we call “the sign fallacy.”


I.  Cases and Problems

To see the wide range of cases to which Auer is relevant,11 consider the following:12

1.  The facts of Auer itself: An agency must decide whether police captains and sergeants are eligible for overtime under the Fair Labor Standards Act. The statute contains an exemption from overtime for “executive, administrative, or professional” employees.13 Binding regulations, enacted with notice and comment, say that such employees must be paid on a “salary basis,” the antonym of piecework. The regulations further say that if pay is “subject to” being adjusted for quality, it is an indicator of piecework. The problem then becomes this: police captains might have their pay docked if, after internal affairs proceedings, they are found to have committed disciplinary infractions, such as maltreating a suspect.14 Does that possibility make them “subject to” adjustment for quality of work? What if the captains were never actually fined, but could be? How should the agency interpret the “subject to” language in the regulation?

2.  An agency issues a legislative rule requiring employers to report occupational diseases within two weeks after they are “diagnosed.” An employer asks the agency to clarify what counts as a “diagnosis.” The agency answers, in an interpretive rule, that chest x-rays that “score” above a specified level of opacity count as a diagnosis.15

3.  An agency issues a regulation that imposes federal tariffs on “diaries” and “bound books.” In an interpretive rule, it announces that “diaries” include daily journals, but not calendars, and that “bound books” refers only to formal bookbinding.16

4.  An agency issues a legislative rule that exempts “waiting time” from the requirements of its overtime regulations. In an interpretive rule, it says that certain emergency services employees are engaged in “waiting time,” rather than “working time,” if they are not required to perform job-related tasks, even if they have to be accessible by phone and available to come into work on ten minutes’ notice.17

5.  An agency issues a rule requiring swimming pools at hotels to be “fully accessible” to people who use wheelchairs through “lifts,” which make pools easy to use. In an interpretation, issued in a public document,18 the agency makes it clear that to qualify, a lift need not be a permanent part of the pool. It can be mobile and simply available to those who ask.19

At least since Bowles v Seminole Rock & Sand Co,20 decided in 1945, the Supreme Court has said that so long as the legislative regulation is genuinely ambiguous, courts should defer to reasonable agency interpretations.21 But if we emphasize a presumption in favor of private ordering, we should be able to see an immediate objection, which would move the law in significant new directions.22 The regulation sets out the law, and it is binding. But within the constraints of the law, regulated classes are authorized to do whatever they like. In the cases above, the agency’s interpretation is essentially irrelevant, because the legislative rule has given the regulated classes (some) room to maneuver. No additional constraints can come from the agency’s interpretation, which would effectively amend the legislative rule by imposing further restrictions.

Consider the swimming pool regulation. If the department has not banned hotels from using portable lifts, then they are entitled to use portable lifts. It is unlawful to add a requirement, through interpretation, that the legislative rule lacks. Perhaps many cases in which agencies seek to benefit from the Auer principle should be resolved against the government, on the theory that if agencies have not expressly regulated private conduct through a legislative rule, the matter is at an end. But suppose that a regulation is genuinely unclear. If so, the question remains: Who interprets it, court or agency? Does the agency’s view deserve some weight?

II.  Deference to Agencies: What Does Congress Want?

Suppose that Congress expressly grants HHS the power to interpret ambiguities in its own regulations—or expressly denies it that authority. That direction should be authoritative subject of course to any constitutional constraints, which we will discuss shortly. The resulting question is simple: Has Congress in fact exercised that authority, either globally or in particular statutes?

The justice who was both Auer’s author and (late in his career) its leading judicial critic, Justice Scalia, believed so. He pointed to § 706 of the Administrative Procedure Act23 (APA), which states that “the reviewing court shall . . . interpret constitutional and statutory provisions, and determine the meaning or applicability of the terms of an agency action.”24 In his view, the APA therefore “contemplates that courts, not agencies, will authoritatively resolve ambiguities in statutes and regulations.”25 But as Scalia was aware, that conclusion raises a serious problem, which is that it would require rejection of the well-established Chevron principle26 itself, on the ground that it
defies the APA. Scalia tried to rescue that principle as “in conformity with the long history of judicial review of executive action,”27 even though it is apparently in tension with the APA (at least on the view Scalia is endorsing). By contrast, his reasoning seems to run, Auer must be rejected on the ground that the APA forbids it—and no such long history justifies it.

But this argument moves far too quickly. Does the text of the APA really require independent judicial interpretations of law? If so, it would seem to repudiate any “long history,” which would therefore be irrelevant. On many approaches to interpretation, a long history, preceding enactment of a contrary text, cannot overcome that text, so long as the text is sufficiently clear. Perhaps Scalia believed that the APA is not clear (as we shall suggest); but he seemed to think that it is.

In other writings, however, Scalia offered a different justification of Chevron, one that (in his view, as expressed in those writings) does not violate any statute or threaten the separation of powers. The question is what the APA commands, and the statement that the court shall “interpret” questions of law is not decisive in favor of independent judicial review, if it is also the case that under organic statutes, the correct interpretation of law depends on the agency’s interpretation of law.28 If an organic statute says “source (as defined by the EPA),” then the law is what the agency says (so long as the agency’s interpretation is reasonable, and subject to constitutional constraints). But suppose that Congress has not said anything explicit but has generally given an agency the authority to issue regulations. Has it also given the agency the authority to interpret ambiguities in the underlying statute? If so, then the law is, to that extent, what the agency says it is—and in faithfully applying the APA, the reviewing court had better say so. On this view, courts do not violate the APA by deferring to reasonable agency interpretations; such deference just is, itself, part of the law that courts declare.29

But when exactly should judges conclude that Congress has indeed given agencies interpretive authority? Scalia saw that as a difficult question, answered in Chevron by a (good) legal fiction, one that is superior to the legal fiction that preceded it. His explanation warrants quotation at length:

An ambiguity in a statute committed to agency implementation can be attributed to either of two congressional desires: (1) Congress intended a particular result, but was not clear about it; or (2) Congress had no particular intent on the subject, but meant to leave its resolution to the agency. When the former is the case, what we have is genuinely a question of law, properly to be resolved by the courts. When the latter is the case, what we have is the conferral of discretion upon the agency, and the only question of law presented to the courts is whether the agency has acted within the scope of its discretion—i.e., whether its resolution of the ambiguity is reasonable. . . .

      . . . If the Chevron rule is not a 100% accurate estimation of modern congressional intent, the prior case-by-case evaluation was not so either—and was becoming less and less so, as the sheer volume of modern dockets made it less and less possible for the Supreme Court to police diverse application of an ineffable rule. And to tell the truth, the quest for the “genuine” legislative intent is probably a wild-goose chase anyway. In the vast majority of cases I expect that Congress neither (1) intended a single result, nor (2) meant to confer discretion upon the agency, but rather (3) didn’t think about the matter at all. If I am correct in that, then any rule adopted in this field represents merely a fictional, presumed intent, and operates principally as a background rule of law against which Congress can legislate.

      If that is the principal function to be served, Chevron is unquestionably better than what preceded it.30

We think that this argument, relying on “a fictional, presumed intent,” is essentially correct, so long as it is understood that the choice of fiction depends on the consequences of adopting one or another. And the reasons that such an instruction should be deemed best are principally consequentialist reasons Chevron itself gave, involving the agency’s comparative competence with respect to fact-finding and policy-making,31 and its comparative political accountability.32 When a statute is unclear, and especially when a complex modern regulatory statute is unclear, resolution of the ambiguity will inevitably require policy-making competence—which courts lack and which agencies have.33 In 1989, Scalia emphasized that precise point as well.34

III.  Auer’s Hour

Which brings us directly to Auer. Nothing in the APA either endorses or rejects Auer, at least in express terms. As we have noted, courts are instructed to “determine the meaning or applicability of the terms of an agency action,”35 but perhaps Congress has said, in general or in particular cases, that the meaning of a regulation turns on the agency’s interpretation of its meaning, where ambiguity exists. In that case, courts fulfill their duty to “determine the meaning” by deferring to that view. If so, courts might say that where legislative rules are ambiguous, the law is what the agency says it is (for example, through an interpretive rule).

It is true that just as in Chevron, Congress has not issued any such express instruction. But it has not issued a contrary instruction either. In this field, any rule again “represents merely a fictional, presumed intent”36 that must be defended as the best instruction to attribute to the national legislature. As others have noted, the best and most straightforward defense of Auer—as a rebuttable presumption about implicit legislative instructions as to the allocation of interpretive authority—rests on the same arguments that Justice Scalia adduced to justify Chevron.37

In that light, Auer itself might be defended in two different ways.38 The first points to the agency’s comparative epistemic advantage as an interpreter. Perhaps the agency has the best understanding of what the underlying legislative rule actually meant. (Of course this assumes a particular, controversial theory of interpretation; but let us bracket that question for the sake of discussion.) If an agency uses the word “diagnosis,” it is in the best position to know what it had in mind. In some cases, this rationale is indeed an exceptionally strong point for Auer, at least when the agency has issued an interpretive rule in relatively close temporal proximity to the legislative rule. (It might well work in the swimming pools case above, and it would work as well for a number of interpretations issued in the aftermath of legislative rules under the Affordable Care Act.) But when an agency has changed its interpretation, this particular argument will not work. And in many cases, an agency will not be uncovering an antecedent intention. It will be interpreting a term on which it had not previously focused. (That might well be true of the diagnosis case above.)

A second defense of Auer points not to the agency’s epistemic advantages as an interpreter, but to its comparative advantages as a policy-maker. On this view, interpretation of ambiguous regulations is really an exercise in policy-making, at least much of the time. A regulatory term like “subject to” calls for further specification in a diverse array of cases, an exercise that in turn requires judgments of policy. Agencies have technical expertise as well as political accountability, and so long as a regulation is ambiguous, it should be “interpreted” by them (policy should be made by them), not by courts, which lack those advantages.

To be sure, the “traditional tools of statutory construction”39 can be used to determine whether there is ambiguity at all. But where there is genuine ambiguity, the agency has comparative policy-making advantages—precisely parallel to its advantages in the Chevron setting. Just as, on Scalia’s view in 1989, Chevron is the best fictional default rule for statutory construction, so too Auer is the best fictional default rule for interpretation of agency regulations.

In 2013, Scalia objected that “the purpose of interpretation is to determine the fair meaning of the rule,” and “[n]ot to make policy.”40 In 1989, by contrast, Scalia had rightly observed “that the traditional tools of statutory construction include not merely text and legislative history but also, quite specifically, the consideration of policy consequences.”41 In his later account, agency enactments must be taken “as written,” thus giving “the Executive [ ] a stable background against which to write its rules and achieve the policy ends it thinks best.”42 We have argued that even if the 2013 view is correct, Auer is justifiable in a range of circumstances.

In any event, however, for the reasons stated by Scalia in 1989, the 2013 view is not correct. And on Scalia’s 1989 view, Auer is not at all inconsistent with the point, which it does not contest, that regulations must be taken as written. Scalia’s 1989 discussion shows that Auer is right for the same reason that Chevron is right: where Congress has not been clear, deference to the agency, in the face of genuine ambiguity, is the best instruction to attribute to it.

IV.  Concerns

Critics of Auer, including Justice Scalia in his later years, have several independent concerns. All of them raise significant questions about methodology.

A.    Incentives

On one view, Auer creates an unfortunate and even dangerous incentive for agencies, which “is to speak vaguely and broadly, so as to retain a ‘flexibility’ that will enable ‘clarification’ with retroactive effect.”43 Auer therefore encourages opportunistic behavior: agencies will issue vague, broad regulations, knowing full well that when the time comes, they will be able to impose the interpretation they prefer.

In the abstract, the concern is certainly intelligible. With Auer, agencies can know that they will have the benefit of being able to clarify ambiguities; without Auer, they would not have that benefit, and might therefore speak precisely. But the idea that Auer results in motivated and nefarious obscurity—“a dangerous permission slip for the arrogation of power”44 —strikes us as a phantasmal terror. Indeed, we are unaware of, and no one has pointed to, any regulation in American history that, because of Auer, was designed vaguely and broadly.45 There is no reason to believe that the magnitude of the posited incentive is substantial. We are dealing here, at best, with the sign fallacy.

No one should deny that such a thing might occur. Many things might occur. But in deciding on the optimal level of clarity and specificity, agencies have a wide range of incentives, cutting in different directions, and the most important of these have nothing at all to do with Auer. Internal pressures often create an incentive toward clarity, so that everyone inside government knows what the regulation means. External pressures often cut in exactly the same direction, with multiple requests for clarity from the regulated sector. To be sure, external and internal pressures might also call for deliberate ambiguity (though we believe that this is far less common). But when ambiguity exists, it is rarely, if ever, because of Auer.

After all, few people who are involved in writing regulations think a great deal about Auer; many of them have absolutely no idea what Auer is. A recent study finds that Auer was less well-known to agency drafters of regulations than Chevron U.S.A. Inc v Natural Resources Defense Council, Inc,46 Skidmore v Swift & Co,47 and United States v Mead Corp;48 drafters themselves knew about Auer only about half of the time.49 It is most unclear that even the half that know Auer think seriously about it when they are writing regulations.

There is a further point, which counts as a fundamental objection to the claim that Auer creates perverse incentives. In an important way, Auer actually incentivizes clarity, and eliminating it would eliminate that incentive. If an agency leaves a regulation ambiguous, it cannot be certain that a subsequent interpretation will be made by an administration with the same or similar values. For agencies, ambiguities are a threat at least as much as they are an opportunity. One administration might well want to ensure that its successor will not be allowed, with the aid of Auer, to shift from a prior position. We do not press this point, because doing so would commit the sign fallacy (with a different sign). Our only suggestion is that those who think Auer is wrong, because of the incentive problem, might have the sign wrong. There are multiple incentives cutting in multiple directions, and their net magnitude is at best unclear.

There is a palpable lack of realism, and a lack of empirical grounding, to the widespread concern that Auer is a significant part of the constellation of considerations that lead agencies to speak specifically or not. We do not believe that agencies often preserve ambiguity on purpose—in fact we think that that is highly unusual—but when they do, Auer is hardly ever, and possibly never, part of the picture. The critics speak abstractly of possible abuses, but present no empirical evidence to substantiate their fears.

B.    Separation of Powers

Auer’s critics have a more fundamental objection, one that involves heavy artillery,50 and that also has intuitive appeal: the decision produces a constitutionally suspect combination of the power to make law with the power to interpret law. Quoting Montesquieu, Scalia insisted that this is a serious problem, because when “legislative and executive powers are united in the same person . . . there can be no liberty.”51 He concluded: “He who writes a law must not adjudge its violation.”52 At least Chevron preserves that separation, because agencies interpret what Congress enacts, but Auer obliterates it, because agencies interpret what agencies enact. Or so the argument runs.

But this critique of Auer is both unsound and too sweeping. There are four critical points. First, the traditional and mainstream understanding in American public law is that when agencies—acting within a statutory grant of authority—make rules, interpret rules, and adjudicate violations, they exercise executive power, not legislative or judicial power. Executive power itself includes the power to make and interpret rules, in the course of carrying out statutory responsibilities.53 Hence there is no commingling of functions within agencies in the first place; any talk to the contrary is loose and imprecise. The Court recently and emphatically reiterated this point, through the pen of . . . Scalia:

[T]he dissent overstates when it claims that agencies exercise “legislative power” and “judicial power.” . . . The former is vested exclusively in Congress, U.S. Const., Art. I, § 1, the latter in the “one supreme Court” and “such inferior Courts as the Congress may from time to time ordain and establish,” Art. III, § 1. Agencies make rules (“Private cattle may be grazed on public lands X, Y, and Z subject to certain conditions”) and conduct adjudications (“This rancher’s grazing permit is revoked for violation of the conditions”) and have done so since the beginning of the Republic. These activities take “legislative” and “judicial” forms, but they are exercises of—indeed, under our constitutional structure they must be exercises of—the “executive Power.” Art. II, § 1, cl. 1.54

This understanding rests on a long-standing theory of the nature and scope of executive power—what has aptly been called the “completion” theory, in the context of presidential power.55 When agencies make rules and interpret law in the course of executing their statutory grants of authority, they are carrying out or completing a legislative plan—“carrying [it] into Execution,” to adapt the words of the Constitution.56

Nothing in that activity amounts to an exercise of legislative or judicial power, properly speaking. Of course there is a separate question here about the validity of the underlying grant of authority, which must contain an “intelligible principle” to avoid an invalid delegation of legislative authority.57 But the constitutional critique of Auer is different from, and tangential to, the delegation issue; the former is meant to apply even when the grant of statutory authority is straightforwardly valid under the nondelegation doctrine.

Our second point is that the separation-of-powers critique of Auer, and of the combination of rulemaking and rule-interpreting functions, is pitched at the wrong level. The separation of powers is fully satisfied so long as the principal institutions set out in the Constitution—Congress, president, and judiciary—exercising their prescribed functions, devise and approve the scheme of agency authority that combines rulemaking and rule-interpreting power in the agency’s hands. Whatever reasons make the constitutional separation of powers attractive in turn support that combination of functions. If the constitutional institutions, operating as they were set up to operate, have decided that such an arrangement is both valid and wise, then respect for the separation of powers counsels approval of the arrangement. Conversely, there is no constitutional rule that each and every subordinate body set up by the constitutional institutions must itself have the same internal structure as the Constitution of 1789, in some oddly fractal way.

Third, bracketing our earlier points, there is a severe mismatch between the sweeping constitutional critique, on the one hand, and on the other the exceedingly narrow context of Auer, where agencies are merely sorting out ambiguities in their own rules. If the combination of lawmaking and law-interpreting functions in agencies really is constitutionally suspect as such, then there are much larger problems than Auer to discuss. The combination of functions in agencies is a hallmark of the administrative state, so the FCC, FTC, SEC, and a myriad of other agencies would seem to be constitutionally suspect as well; all of these agencies write binding rules, bring enforcement actions, and adjudicate violations, in the course of which they interpret the very rules that they themselves have made.

Justice Thomas seems to think that these agencies have to go. But it is clear that Scalia, the “faint-hearted originalist,”58 would never have gone so far. And the Court as a body has repeatedly said that the combination of functions is not in itself a constitutional problem.59 In the Auer setting the quite modest combination (for example, in defining the term “diagnosis”), if that is what it is, seems especially trivial, de minimis.

If correct, therefore, the constitutional critique actually amounts to an indictment not merely of Auer, which seems at most a minor detail, but of much of the contemporary administrative state. Its proponents should have the candor to argue for it on those terms. Perhaps those proponents, or some of them, believe that the constitutionality of agencies that combine lawmaking with law interpretation is too entrenched to deserve rethinking, while Auer is fair game. But is it really a good or even intelligible use of the separation-of-powers principle to insist that judges must, entirely on their own, interpret the meaning of words like “diagnosis” or “diaries”? Does constitutional liberty depend on an affirmative answer?

Fourth and finally, it is a simple confusion to suggest an agency could ever “delegate power to itself.” Agencies just have whatever quantum of power they have, under relevant statutory grants of authority; whether they exercise that power through legislative rulemaking, adjudication, guidances, interpretations, or whatever, the quantum of power itself is unaffected. Judges can always enforce the outer boundaries of the agency’s grant of authority, however it is exercised, whether through statutory interpretation or through arbitrariness review.

What is really at stake in the Auer setting is not agency self-delegation of power, and certainly not the expansion of power, but rather timing—the timing of the exercise of whatever statutory power the agency otherwise has.60 When an agency makes valid legislative rules, those rules bind the agency itself as well as all the world. The more specific the rule, the less future discretion the agency has when interpreting the rule; the less specific the rule, the more future discretion the agency enjoys to flesh out the rule by means of guidances, interpretations, and adjudicative orders. Again, the overall quantum of statutory power is not expanded but instead allocated between present and future.

Put differently—and this is the way administrative law puts it—the agency’s choice is to allocate its authority between more general rulemaking now and more specific interpretation or adjudication later.61 The more content the agency supplies through legislative rulemaking now, the less content it will have to supply (or indeed be able to supply) through issue-specific interpretation or case-specific adjudication later.

It then becomes clear that the Auer issue that Scalia attempts to describe as an issue of self-delegation is really just a version of the familiar administrative law question of agency discretion to choose between policy-making forms or policy instruments.62 And the law’s answer—at least since Securities & Exchange Commission v Chenery Corp63 (“Chenery II”) in 1947, and continuing throughout the modern era—has been that agency discretion to make procedural choices is extremely broad and subject only to the constraints of arbitrariness.64 In the standard Auer case, there is nothing at all arbitrary about the agency’s decision to specify, through interpretation, what a legislative rule means, not least because the agency is often answering a question that it did not anticipate.

C.    The Thomas Critique

Thomas has recently offered a somewhat different objection to Auer, also constitutionally grounded—and as we will see, its implications are even more sweeping. In his view, Auer “subjects regulated parties to precisely the abuses that the Framers sought to prevent,” and so it is “constitutionally suspect.”65 It is, after all, the power of courts to issue authoritative interpretations in judicial proceedings. The critical problem with Auer is that it substitutes the agency’s power of interpretation for that of the courts: “Because the agency is thus not properly constituted to exercise the judicial power under the Constitution, the transfer of interpretive judgment raises serious separation-of-powers concerns.”66 Congress may not delegate “binding” interpretive power to agencies construing their own regulations, because as a constitutional matter Congress does not possess the “judicial power” in the first place, and therefore cannot give it away.

Thomas’s theory at least has the virtue of novelty. But the problems that afflict it are so many, and so transparent, that one stares puzzled at Thomas’s opinion—can it really mean what it seems to be saying? Agency interpretations are only as “binding” as the underlying legislative rules or statutes themselves; their force is entirely derivative. Thomas thus fails to explain why agencies’ interpretation of their own regulations should be thought to count as “binding” at all.

As the Court emphasized in Perez v Mortgage Bankers Association,67 courts retain the ultimate authority to decide whether the agency’s interpretation is correct;68 in that sense the agency binds no one. And if the agency’s interpretation really has binding legal effect, then it may be vulnerable on procedural grounds if created without notice and comment.69 Thomas seems to be working with some unarticulated, and perhaps indefensible, conception of the distinction between interpretation and lawmaking. Like the academic work on which Thomas draws,70 his view “largely elides the line-drawing problem posed by any attempt to distinguish interpretation from legislation.”71

In any event Thomas’s theory sweeps far beyond agencies’ interpretation of their own regulations, to include any “binding” agency interpretation at all, including agency interpretations of organic statutes themselves when those interpretations are deemed binding. The theory would therefore bar not only Auer deference, but also Chevron deference itself, even if explicitly required by Congress—here too an outcome that is congenial to Thomas, but almost certainly not to Scalia.

Let us end with the Court’s recent, simple, emphatic, and powerful response to criticisms of Auer: when the underlying regulation is clear, the agency must comply, and it is ultimately up to the judges to decide when the regulation is clear.72 It is instructive in this regard to compare the critiques of Auer with parallel arguments in City of Arlington, Texas v Federal Communications Commission73 —written by Scalia in 2013. City of Arlington upheld the authority of agencies to determine, through statutory interpretation, the scope of their own “jurisdiction,” within the bounds of statutory ambiguity74 —an abomination to the traditional legal mind, and a holding that prompted vehement objections from the chief justice in dissent.75

For agencies to do as City of Arlington authorized, the dissent argued, would strengthen the “potent brew of executive, legislative, and judicial power”76 that agencies already mix together. (Again, however, let us be clear that the Court has consistently held that agencies implementing statutory grants of authority always and only exercise executive power, which includes subsidiary powers to make and interpret rules.)77

Scalia’s reply was strong and straightforward:

The fox-in-the-henhouse syndrome is to be avoided not by establishing an arbitrary and undefinable category of agency decisionmaking that is accorded no deference, but by taking seriously, and applying rigorously, in all cases, statutory limits on agencies’ authority. Where Congress has established a clear line, the agency cannot go beyond it; and where Congress has established an ambiguous line, the agency can go no further than the ambiguity will fairly allow.78

But this argument applies, mutatis mutandis, to Auer deference equally well. The putative separation-of-powers problem with the combination of lawmaking and law-interpreting power in the same hands—the fox put in charge of the henhouse—is exactly the same in both the Auer setting and the setting of agency self-determination of “jurisdiction.” And the remedy Scalia proposes, judicial enforcement of clear texts, is the same as well.

It is not as though there are no checks on agencies built into the Auer framework. Indeed there are several:79

1.   Judicial enforcement of clear regulations and statutes.

First and foremost, the regulation that is being interpreted, if otherwise valid, provides the law, and any interpretation must comply with it. The regulation itself must also comply with the underlying statute, and its enactment must be procedurally valid. Judges, not anyone else, decide whether these requirements are satisfied.

2.   Arbitrariness review.

As the Court also emphasized in Perez, the “most notable” constraint on agency decision-making is “the arbitrary and capricious standard,” which serves to promote “procedural fairness” by requiring agencies to give good reasons for their procedural choices—and, of course, for their interpretations.80 But this is not a general objection to Auer deference as such; it is an inquiry to be carried out in particular cases.

3.   Reliance, consistency, and arbitrariness.

An agency interpretation might be arbitrary and capricious because it defeats reliance interests without adequate explanation.81 Indeed, interpretations that are inconsistent over time82 may be disqualified from receiving Auer deference at all.83 Here too, however, the notions of reliance and consistency support no general objection to Auer. They are qualifications that do not entail or presuppose any doubt about the validity of deference in the normal case.

4.   The antiparroting canon.

Where an agency issues a binding regulation (perhaps through notice and comment) that merely “parrots” the language of the underlying statute, and then interprets the regulation rather than the statute, Auer deference does not apply.84 This is best understood as a corollary of the completion theory of executive power;85 an agency engaged in completion should add some specification, some content, to the underlying statute, or its action has accomplished nothing and counts for nothing.

* * *

So there are a range of qualifications and checks on Auer deference, under current law. The alternative to full judicial review of agency interpretations is hardly unchecked administrative power; it is a cabined regime of partial and qualified deference.

D.    The Skidmore Alternative?

Scalia’s own view seems to have been that the alternative to Auer deference was no deference at all—de novo judicial interpretation of agency regulations.86 On a different view, there is an appealing compromise: independent judicial interpretation of agency regulations would be a bad idea (for reasons we have sketched), but Auer confers too much discretion on agencies, and so the right approach is based on “Skidmore deference.” In theory, Skidmore opts for “persuasive” rather than “authoritative” deference. It looks to “the thoroughness evident in [the agency’s] consideration, the validity of its reasoning, its consistency with earlier and later pronouncements, and all those factors which give it power to persuade, if lacking power to control”87 —a kind of intermediate type or level of deference, which does not amount to the abdication that is Auer.88

The simplest answer to this question is that there is no problem to be fixed. Nothing is broken. However, assuming there is a problem, Skidmore is not a better solution, for several reasons. Scalia’s view, which rejects the Skidmore alternative altogether, is theoretically uncompromising, whereas the Skidmore alternative seems to flinch from the logic of the very arguments the critics adduce to undermine Auer. After all, if agencies interpreting their own regulations are engaged in a kind of constitutionally illegitimate self-dealing that results from the fusion of lawmaking and law-interpreting power, why should their claims be given any sort of deference at all? The more consistent approach would be to treat such claims as tainted by their illegitimate origin, and to ignore them altogether.

Furthermore, the line between Auer deference and Skidmore deference is thin even in principle and often invisible in operation; in general, it is of far more interest to administrative law teachers than to actual judicial practice. How often would courts strike down an agency interpretation under Skidmore but uphold it under Auer?89 We suspect not very often. In any case, is it really worthwhile to renovate a doctrine that has worked reasonably well for decades in order to substitute a standard of review that is, at most, marginally less deferential than current law? Increases in legal complexity have their place, but this is not one of them.

E.    A Pragmatic Perspective

There is something overheated, wildly disproportionate, about the separation-of-powers critique of Auer. Return to the cases with which we began. Is constitutional liberty really at risk if an agency is allowed to interpret the phrase “subject to” or the word “diagnosis,” within the bounds of textual meaning? “Bound books”? “Diaries”? Is liberty less at risk if, in the face of ambiguity, courts, composed of generalist judges, interpret such terms on their own? Does it matter that agency interpretations often increase, rather than confine, the freedom of the regulated class, by telling its members that they may in fact do what they want to do? Does it matter that in hard cases, judicial interpretation of ambiguities often entails political judgments, as reflected in the conspicuously and predictably different views of Republican and Democratic appointees, even under Chevron?90 Does it matter that we are typically speaking of interstitial and highly technical judgments, in which agencies understand an ambiguous term (“diaries”) in a linguistically permissible way?

In our view, the strongest objections to Auer are not large-scale constitutional abstractions. The strongest objections are instead pragmatic. They suggest, far more modestly, that when an agency is interpreting its own regulations, the best legal fiction is that ambiguities are for courts, not administrators, to resolve. On this view, it would be possible to approve of Chevron but to disapprove of Auer, contending that if those who write laws (regulations) can also interpret them, there is a risk of bias. If we distrusted the agency’s competence or expertise, or believed that it was systematically biased, and if we thought that judges could be better trusted to resolve questions that often involve policy judgments, we might deliberately choose to indulge the fiction that Congress means to have courts, rather than agencies, resolve ambiguities in agency regulations. Some version of that essentially pragmatic view may underlie the opposition to Auer, which nonetheless is expressed in constitutional terms.

But return to the cases with which we began, and ask whether a judgment about institutional competence really justifies the conclusion that the relevant ambiguities are always best resolved through independent judicial judgment. Law is full of situations in which judges review the work of potentially biased experts (in part because bias and expertise sometimes come as a package); but the risk of bias is just one factor among many, and will sometimes or even often be outweighed by the advantages of deference. Because of the need to resolve technical issues, and because of the plain advantages of accountability, the balance cuts hard in the direction of Auer.

The law’s usual response to such situations is not to substitute judicial judgments for agency judgments wholesale, but to examine at retail the reasons that experts give for their policy choices—which is why the Perez Court both reaffirmed Auer, and yet also emphasized the role of arbitrariness review in monitoring agency choices, in particular cases.91 In any event, as we have suggested, the fear that agencies will seek to expand their own authority depends on a mistaken factual predicate, indeed a naïve picture of what agency interpretations of their own rules are for. Indeed, one of the primary functions of such interpretations has been to confine agency authority rather than to expand it, often in response to questions and concerns from the regulated community.

Auer in effect allows agencies to clarify how they intend to exercise their discretion, without fear of judicial second-guessing (within the limits of relevant ambiguity). It accommodates the many agencies that respond to the questions and concerns of regulated parties by saying that the agency does not mean to invade, and has no intention of invading, the private sphere in the relevant way. So long as the underlying regulation is ambiguous, and the agency is interpreting it, it would hardly be better for courts to second-guess agency judgments on this count.


For more than seventy years, courts have deferred to agency interpretations of ambiguous regulations. In the absence of a clear congressional direction, courts have assumed that because of their specialized competence, and their greater accountability, agencies are in a better position to decide on the meaning of ambiguous terms. That assumption is correct.

The contemporary challenges to Auer rest on three weak foundations: a question-begging argument about the APA; unhelpfully abstract and overly sweeping rhetoric about separation of powers; and an unrealistic (and somewhat fearful) claim about agency opportunism, exemplifying the sign fallacy. The best approach to agency regulations is simple: Use the conventional tools of interpretation, and if ambiguity remains, the agency’s interpretation prevails.92

  • 11Here and throughout, we assume that the agency’s interpretation of its own legislative regulation either was enacted through notice-and-comment procedures in its own right, or else falls within a valid exception to notice-and-comment rulemaking under § 553(b)(3)(A) or (B) of the Administrative Procedure Act (the exceptions for interpretive rules, procedural rules, general statements of policy, or legislative rules as to which there is “good cause” for dispensing with notice and comment). 5 USC § 553(b)(3)(A)–(B). We thereby bracket and set aside any claim that the agency’s interpretation was procedurally defective; our focus is strictly on the substantive validity of the interpretation. Although Chevron deference to agency interpretation of statutes is partly tied to the procedures agencies use to enact their interpretations, see United States v Mead Corp, 533 US 218, 229–30 (2001), in the Auer setting the law rejects that approach. Indeed, agencies may receive deference even for interpretations set forth in appellate briefs. See, for example, Talk America, Inc v Michigan Bell Telephone Co, 564 US 50, 59 (2011).
  • 12In these hypotheticals, we draw on actual cases but deviate from them in minor ways.
  • 1329 USC § 213(a)(1).
  • 14See Auer, 519 US at 455–56.
  • 15See American Mining Congress v Mine Safety & Health Administration, 995 F2d 1106, 1107–08 (DC Cir 1993).
  • 16See Mead, 533 US at 221–25.
  • 17See Skidmore v Swift & Co, 323 US 134, 135–36, 139 (1944).
  • 18In accordance with existing law, we are understanding Auer to apply not only to “interpretative” rules within the meaning of the Administrative Procedure Act, 5 USC § 553(b)(3)(A), but also to a broad class of agency interpretations, so long as they are authoritative and do not fall within certain exceptions we will discuss later. See Perez v Mortgage Bankers Association, 135 S Ct 1199, 1208 n 4 (2015) (listing limited qualifications and exceptions to Auer deference). Six justices (including the chief justice and Justice Anthony Kennedy) joined the opinion in Perez and its important footnote 4, suggesting that a clear supermajority of the Court is, as of now, willing to accept Auer subject to qualifications. Perez, 135 S Ct at 1202. Further evidence is supplied by United Student Aid Funds, Inc v Bible, 136 S Ct 1607 (2016), a certiorari petition that asked the Court to overturn Auer, but that was denied, with Justice Thomas the lone dissenter. See id at 1608–09 (Thomas dissenting from denial of certiorari).
  • 19But see Questions and Answers: Accessibility Requirements for Existing Swimming Pools at Hotels and Other Public Accommodations (DOJ, May 24, 2012), archived at (discussing this issue but coming out the other way via “Questions and Answers”).
  • 20325 US 410 (1945).
  • 21Id at 413–14.
  • 22This approach is a close cousin to that defended in Frank H. Easterbrook, Statutes’ Domains, 50 U Chi L Rev 533 (1983), which, in our view, similarly rests on a presumption in favor of private ordering. See id at 552.
  • 2360 Stat 237 (1946), codified as amended in various sections of Title 5.
  • 245 USC § 706.
  • 25Perez v Mortgage Bankers Association, 135 S Ct 1199, 1211 (2015) (Scalia concurring in the judgment).
  • 26See Chevron U.S.A. Inc v Natural Resources Defense Council, Inc, 467 US 837, 842–45 (1984).
  • 27Perez, 135 S Ct at 1212 (Scalia concurring in the judgment).
  • 28See Henry P. Monaghan, Marbury and the Administrative State, 83 Colum L Rev 1, 6 (1983).
  • 29See id. Here, too, we do not necessarily mean to endorse this justification for Chevron, only to elicit its implications for the Auer question.
  • 30Antonin Scalia, Judicial Deference to Administrative Interpretations of Law, 1989 Duke L J 511, 516–17.
  • 31See Cass R. Sunstein, The Most Knowledgeable Branch, 164 U Pa L Rev 1607, 1608–11 (2016).
  • 32See Chevron, 467 US at 865–66. There are other considerations in support of Chevron. See Peter L. Strauss, One Hundred Fifty Cases per Year: Some Implications of the Supreme Court’s Limited Resources for Judicial Review of Agency Action, 87 Colum L Rev 1093, 1121 (1987).
  • 33See Sunstein, 164 U Pa L Rev at 1608–09 (cited in note 31). Note, however, the interesting qualification in King v Burwell, 135 S Ct 2480 (2015), stressing that in extraordinary cases, involving issues of great economic and social significance, the presumption should be reversed, and Congress should be presumed to want independent judicial judgment. See id at 2488–89. For discussion, see Cass R. Sunstein, Chevron Step Zero, 92 Va L Rev 187, 236–42 (2006).
  • 34See Scalia, 1989 Duke L J at 516–18 (cited in note 30). There is a competing view, which would invoke background principles of separation of powers, and perhaps concerns about agency self-dealing, to generate a different presumption. See Cynthia R. Farina, Statutory Interpretation and the Balance of Power in the Administrative State, 89 Colum L Rev 452, 456 (1989). See also Stephen Breyer, Judicial Review of Questions of Law and Policy, 38 Admin L Rev 363, 394–95 (1986) (suggesting a multifactor balancing test). In our view, Scalia’s position in 1989 is essentially correct. For discussion, see Adrian Vermeule, Introduction: Mead in the Trenches, 71 Geo Wash L Rev 347, 358 (2003); Sunstein, 92 Va L Rev at 202–05 (cited in note 33).
  • 355 USC § 706.
  • 36Scalia, 1989 Duke L J at 517 (cited in note 30).
  • 37See Separation of Powers Restoration Act of 2016, Hearing on HR 4768 before the Subcommittee on Regulatory Reform, Commercial and Antitrust Law of the House Committee on the Judiciary, 114th Cong, 2d Sess 13 (May 17, 2016) (“Testimony of Ronald M. Levin”) (statement of Ronald M. Levin, William R. Orthwein Distinguished Professor of Law, Washington University in St. Louis), archived at (“[T]he strongest justifications [for Auer] run parallel to the pragmatic justifications for Chevron.”).
  • 38For helpful discussion, see Matthew C. Stephenson and Miri Pogoriler, Seminole Rock’s Domain, 79 Geo Wash L Rev 1449, 1454–57 (2011).
  • 39Chevron U.S.A. Inc v Natural Resources Defense Council, Inc, 467 US 837, 843 n 9 (1984).
  • 40Decker v Northwest Environmental Defense Center, 133 S Ct 1326, 1340 (2013) (Scalia concurring in part and dissenting in part).
  • 41Scalia, 1989 Duke L J at 515 (cited in note 30) (quotation marks omitted).
  • 42Decker, 133 S Ct at 1340 (Scalia concurring in part and dissenting in part).
  • 43Decker v Northwest Environmental Defense Center, 133 S Ct 1326, 1341 (2013) (Scalia concurring in part and dissenting in part).
  • 44Id (Scalia concurring in part and dissenting in part).
  • 45In nearly four years in the federal government, one of us (Sunstein) dealt with well over two thousand rules, and he never heard even a single person suggest, or come close to suggesting, that a regulation should be written vaguely or ambiguously in light of Auer, or so that the agency could later interpret it as it saw fit.
  • 46467 US 837 (1984).
  • 47323 US 134 (1944).
  • 48533 US 218 (2001).
  • 49See Christopher J. Walker, Inside Agency Statutory Interpretation, 67 Stan L Rev 999, 1061–66 (2015).
  • 50See Manning, 96 Colum L Rev at 654 (cited in note 4).
  • 51Decker, 133 S Ct at 1341 (Scalia concurring in part and dissenting in part) (ellipsis in original).
  • 52Id at 1342 (Scalia concurring in part and dissenting in part).
  • 53See United States v Grimaud, 220 US 506, 521 (1911) (noting that statutory authority to make administrative rules is a grant of executive power, not legislative power).
  • 54City of Arlington, Texas v Federal Communications Commission, 133 S Ct 1863, 1873 n 4 (2013).
  • 55See Jack Goldsmith and John F. Manning, The President’s Completion Power, 115 Yale L J 2280, 2282 (2006).
  • 56US Const Art I, § 8. The idea behind the clause is actually helpful, by analogy. Just as Congress makes laws to “carry into execution” its powers granted by higher (constitutional) law, so too agencies make rules to carry into execution their powers granted by higher (statutory) law.
  • 57J.W. Hampton, Jr., & Co v United States, 276 US 394, 409 (1928).
  • 58Antonin Scalia, Originalism: The Lesser Evil, 57 U Cin L Rev 849, 864 (1989).
  • 59See, for example, Withrow v Larkin, 421 US 35, 54 (1975); Marcello v Bonds, 349 US 302, 311 (1955); Federal Trade Commission v Cement Institute, 333 US 683, 702 (1948).
  • 60For valuable discussion, see Aaron L. Nielson, Beyond Seminole Rock, 105 Georgetown L J *4–7 (forthcoming 2017), archived at
  • 61For a discussion of the close connection between Auer and agency choice between rulemaking and adjudication, see Testimony of Ronald M. Levin, 114th Cong, 2d Sess at 16–18 (cited in note 37).
  • 62See M. Elizabeth Magill, Agency Choice of Policymaking Form, 71 U Chi L Rev 1383, 1384–85 (2004).
  • 63332 US 194, 201–03 (1947).
  • 64See, for example, National Labor Relations Board v Bell Aerospace Co, 416 US 267, 293–95 (1974).
  • 65Perez v Mortgage Bankers Association, 135 S Ct 1199, 1213, 1215 (2015) (Thomas concurring in the judgment).
  • 66Id at 1220 (Thomas concurring in the judgment).
  • 67135 S Ct 1199 (2015).
  • 68See id at 1208 n 4.
  • 69See id at 1204.
  • 70See generally Hamburger, Is Administrative Law Unlawful? (cited in note 6).
  • 71Gary Lawson, Book Review, The Return of the King: The Unsavory Origins of Administrative Law, 93 Tex L Rev 1521, 1545 (2015). See also Adrian Vermeule, Book Review, No, 93 Tex L Rev 1547, 1560–61 (2015).
  • 72See Perez, 135 S Ct at 1208 n 4 (“Even in cases where an agency’s interpretation receives Auer deference, however, it is the court that ultimately decides whether a given regulation means what the agency says.”).
  • 73133 S Ct 1863 (2013).
  • 74See id at 1868.
  • 75Note, however, that the chief justice took a narrower approach, asking exactly what Congress had delegated, and did not argue in favor of carving out jurisdictional judgments as such. See id at 1877 (Roberts dissenting).
  • 76Id at 1886 (Roberts dissenting).
  • 77See Grimaud, 220 US at 517–21; City of Arlington, 133 S Ct at 1873 n 4.
  • 78City of Arlington, 133 S Ct at 1874.
  • 79For a discussion of possible limitations to Auer deference, partly (but only partly) overlapping with the limitations in current law, see Stephenson and Pogoriler, 79 Geo Wash L Rev at 1466–1503 (cited in note 38).
  • 80Perez, 135 S Ct at 1209.
  • 81See id. See also Encino Motorcars, LLC v Navarro, 136 S Ct 2117, 2125–26 (2016); Federal Communications Commission v Fox Television Stations, Inc, 556 US 502, 515–16 (2009).
  • 82Thomas Jefferson University v Shalala, 512 US 504, 515 (1994) (“[A]n agency’s interpretation of a . . . regulation that conflicts with a prior interpretation is entitled to considerably less deference than a consistently held agency view.”) (quotation marks omitted).
  • 83See Perez, 135 S Ct at 1208 n 4 (collecting qualifications). The same note also alludes to the possibility that Auer deference might not apply “when there is reason to suspect that the agency’s interpretation does not reflect the agency’s fair and considered judgment.” Id, quoting Christopher v SmithKline Beecham Corp, 132 S Ct 2156, 2166 (2012). This last point is a well-established qualification to the Court’s periodic practice of affording Auer deference even to agency interpretations contained in briefs and other litigation-related documents. See, for example, Talk America, Inc v Michigan Bell Telephone Co, 564 US 50, 59 (2011) (stating that the Court “defer[s] to an agency’s interpretation of its regulations, even in a legal brief,” unless there is some “reason to suspect that the interpretation does not reflect the agency’s fair and considered judgment on the matter in question”). The issue of deference to agency litigating positions is tangential to the questions we address. The examples we have given all involve interpretations issued during agency proceedings, not post hoc rationalizations generated during litigation, and the major criticisms of Auer, based on the separation-of-powers and agency incentives, are not cabined to the latter class of situations.
  • 84See Gonzales v Oregon, 546 US 243, 257 (2006).
  • 85See Goldsmith and Manning, 115 Yale L J at 2282 (cited in note 55).
  • 86See Testimony of Ronald M. Levin, 114th Cong, 2d Sess at 16–17 (cited in note 37). See also Decker, 133 S Ct at 1342–44 (Scalia concurring in part and dissenting in part) (rejecting Auer and then interpreting a regulation de novo, without reference to Skidmore).
  • 87Skidmore, 323 US at 140.
  • 88See Manning, 96 Colum L Rev at 686–90 (cited in note 4).
  • 89In Christopher, the Court concluded that in a case involving legitimate reliance interests, Skidmore, and not Auer, would provide the appropriate standard. However, the Court deemed the agency interpretation unpersuasive even under Skidmore, which reinforces our impression that in most cases the two standards yield the same conclusions; the difference between Auer and Skidmore usually makes no difference. See Christopher, 132 S Ct at 2168–70. See also Gonzales, 546 US at 268–69 (rejecting Auer deference and then finding an agency interpretation unpersuasive under Skidmore). And in any event Perez, which lays out a framework for approaching questions of deference to agency interpretations of regulations, makes no mention of Skidmore. See Perez, 135 S Ct at 1208 n 4.
  • 90See Miles and Sunstein, 73 U Chi L Rev at 825–26 (cited in note 8); Cass R. Sunstein and Thomas J. Miles, Depoliticizing Administrative Law, 58 Duke L J 2193, 2202 (2009).
  • 91See Perez, 135 S Ct at 1208 n 4, 1209.
  • 92Subject to various qualifications. See id.